Alois Knoll

Shangding Gu, Long Yang, Yali Du et al.

Reinforcement Learning (RL) has achieved tremendous success in many complex decision-making tasks. However, safety concerns are raised during deploying RL in real-world applications, leading to a growing demand for safe RL algorithms, such as in autonomous driving and robotics scenarios. While safe control has a long history, the study of safe RL algorithms is still in the early stages. To establish a good foundation for future safe RL research, in this paper, we provide a review of safe RL from the perspectives of methods, theories, and applications. Firstly, we review the progress of safe RL from five dimensions and come up with five crucial problems for safe RL being deployed in real-world applications, coined as "2H3W". Secondly, we analyze the algorithm and theory progress from the perspectives of answering the "2H3W" problems. Particularly, the sample complexity of safe RL algorithms is reviewed and discussed, followed by an introduction to the applications and benchmarks of safe RL algorithms. Finally, we open the discussion of the challenging problems in safe RL, hoping to inspire future research on this thread. To advance the study of safe RL algorithms, we release an open-sourced repository containing the implementations of major safe RL algorithms at the link: https://github.com/chauncygu/Safe-Reinforcement-Learning-Baselines.git.

Ekim Yurtsever, Emeç Erçelik, Mingyu Liu et al.

State-of-the-art lidar-based 3D object detection methods rely on supervised learning and large labeled datasets. However, annotating lidar data is resource-consuming, and depending only on supervised learning limits the applicability of trained models. Self-supervised training strategies can alleviate these issues by learning a general point cloud backbone model for downstream 3D vision tasks. Against this backdrop, we show the relationship between self-supervised multi-frame flow representations and single-frame 3D detection hypotheses. Our main contribution leverages learned flow and motion representations and combines a self-supervised backbone with a supervised 3D detection head. First, a self-supervised scene flow estimation model is trained with cycle consistency. Then, the point cloud encoder of this model is used as the backbone of a single-frame 3D object detection head model. This second 3D object detection model learns to utilize motion representations to distinguish dynamic objects exhibiting different movement patterns. Experiments on KITTI and nuScenes benchmarks show that the proposed self-supervised pre-training increases 3D detection performance significantly. https://github.com/emecercelik/ssl-3d-detection.git

Haotian Liu, Guang Chen, Sanqing Qu et al.

Event cameras have the ability to record continuous and detailed trajectories of objects with high temporal resolution, thereby providing intuitive motion cues for optical flow estimation. Nevertheless, most existing learning-based approaches for event optical flow estimation directly remould the paradigm of conventional images by representing the consecutive event stream as static frames, ignoring the inherent temporal continuity of event data. In this paper, we argue that temporal continuity is a vital element of event-based optical flow and propose a novel Temporal Motion Aggregation (TMA) approach to unlock its potential. Technically, TMA comprises three components: an event splitting strategy to incorporate intermediate motion information underlying the temporal context, a linear lookup strategy to align temporally fine-grained motion features and a novel motion pattern aggregation module to emphasize consistent patterns for motion feature enhancement. By incorporating temporally fine-grained motion information, TMA can derive better flow estimates than existing methods at early stages, which not only enables TMA to obtain more accurate final predictions, but also greatly reduces the demand for a number of refinements. Extensive experiments on DSEC-Flow and MVSEC datasets verify the effectiveness and superiority of our TMA. Remarkably, compared to E-RAFT, TMA achieves a 6\% improvement in accuracy and a 40\% reduction in inference time on DSEC-Flow. Code will be available at \url{https://github.com/ispc-lab/TMA}.

Ruiqi Zhang, Jing Hou, Florian Walter et al. · berkeley

Reinforcement Learning (RL) is a potent tool for sequential decision-making and has achieved performance surpassing human capabilities across many challenging real-world tasks. As the extension of RL in the multi-agent system domain, multi-agent RL (MARL) not only need to learn the control policy but also requires consideration regarding interactions with all other agents in the environment, mutual influences among different system components, and the distribution of computational resources. This augments the complexity of algorithmic design and poses higher requirements on computational resources. Simultaneously, simulators are crucial to obtain realistic data, which is the fundamentals of RL. In this paper, we first propose a series of metrics of simulators and summarize the features of existing benchmarks. Second, to ease comprehension, we recall the foundational knowledge and then synthesize the recently advanced studies of MARL-related autonomous driving and intelligent transportation systems. Specifically, we examine their environmental modeling, state representation, perception units, and algorithm design. Conclusively, we discuss open challenges as well as prospects and opportunities. We hope this paper can help the researchers integrate MARL technologies and trigger more insightful ideas toward the intelligent and autonomous driving.

Mert Keser, Alois Knoll

Reliable object detection is critical for automated driving, yet even state-of-the-art detectors inevitably make errors that can compromise safety. Introspection methods that predict detector failures enable safer deployment by triggering fallback mechanisms or alerting human operators. However, existing approaches rely solely on last-layer features or hand-crafted statistics, discarding valuable information from earlier layers that capture different levels of visual abstraction. We propose Layer Feature Attention (LFA), a lightweight introspection method that learns to aggregate features from multiple backbone layers through an attention mechanism. Our key insight is that detection errors manifest differently across feature hierarchies: low-level layers capture fine-grained details essential for detecting small or occluded objects, while high-level layers encode semantic information for scene understanding. LFA learns layer importance weights end-to-end, enabling both improved error prediction and interpretable analysis of which feature levels are most indicative of detector failures. Extensive experiments on KITTI and BDD100K demonstrate that LFA achieves state-of-the-art introspection performance, outperforming single-layer baselines across multiple detector architectures.

Xiangtong Yao, Hongkuan Zhou, Oier Mees et al. · cmu

Language-conditioned robot manipulation is an emerging field aimed at enabling seamless communication and cooperation between humans and robotic agents by teaching robots to comprehend and execute instructions conveyed in natural language. This interdisciplinary area integrates scene understanding, language processing, and policy learning to bridge the gap between human instructions and robot actions. In this comprehensive survey, we systematically explore recent advancements in language-conditioned robot manipulation. We categorize existing methods based on the primary ways language is integrated into the robot system, namely language for state evaluation, language as a policy condition, language for cognitive planning and reasoning, and language in unified vision-language-action models. Specifically, we further analyze state-of-the-art techniques from five axes of action granularity, data and supervision regimes, system cost and latency, environments and evaluations, and cross-modal task specification. Additionally, we highlight the key debates in the field. Finally, we discuss open challenges and future research directions, focusing on potentially enhancing generalization capabilities and addressing safety issues in language-conditioned robot manipulators.

Shangding Gu, Alap Kshirsagar, Yali Du et al.

Deployment of Reinforcement Learning (RL) algorithms for robotics applications in the real world requires ensuring the safety of the robot and its environment. Safe Robot RL (SRRL) is a crucial step towards achieving human-robot coexistence. In this paper, we envision a human-centered SRRL framework consisting of three stages: safe exploration, safety value alignment, and safe collaboration. We examine the research gaps in these areas and propose to leverage interactive behaviors for SRRL. Interactive behaviors enable bi-directional information transfer between humans and robots, such as conversational robot ChatGPT. We argue that interactive behaviors need further attention from the SRRL community. We discuss four open challenges related to the robustness, efficiency, transparency, and adaptability of SRRL with interactive behaviors.

Tianpei Zou, Sanqing Qu, Zhijun Li et al.

3D point cloud segmentation has received significant interest for its growing applications. However, the generalization ability of models suffers in dynamic scenarios due to the distribution shift between test and training data. To promote robustness and adaptability across diverse scenarios, test-time adaptation (TTA) has recently been introduced. Nevertheless, most existing TTA methods are developed for images, and limited approaches applicable to point clouds ignore the inherent hierarchical geometric structures in point cloud streams, i.e., local (point-level), global (object-level), and temporal (frame-level) structures. In this paper, we delve into TTA in 3D point cloud segmentation and propose a novel Hierarchical Geometry Learning (HGL) framework. HGL comprises three complementary modules from local, global to temporal learning in a bottom-up manner.Technically, we first construct a local geometry learning module for pseudo-label generation. Next, we build prototypes from the global geometry perspective for pseudo-label fine-tuning. Furthermore, we introduce a temporal consistency regularization module to mitigate negative transfer. Extensive experiments on four datasets demonstrate the effectiveness and superiority of our HGL. Remarkably, on the SynLiDAR to SemanticKITTI task, HGL achieves an overall mIoU of 46.91\%, improving GIPSO by 3.0\% and significantly reducing the required adaptation time by 80\%. The code is available at https://github.com/tpzou/HGL.

Fengyi Shen, Akhil Gurram, Ziyuan Liu et al.

Domain adaptive semantic segmentation methods commonly utilize stage-wise training, consisting of a warm-up and a self-training stage. However, this popular approach still faces several challenges in each stage: for warm-up, the widely adopted adversarial training often results in limited performance gain, due to blind feature alignment; for self-training, finding proper categorical thresholds is very tricky. To alleviate these issues, we first propose to replace the adversarial training in the warm-up stage by a novel symmetric knowledge distillation module that only accesses the source domain data and makes the model domain generalizable. Surprisingly, this domain generalizable warm-up model brings substantial performance improvement, which can be further amplified via our proposed cross-domain mixture data augmentation technique. Then, for the self-training stage, we propose a threshold-free dynamic pseudo-label selection mechanism to ease the aforementioned threshold problem and make the model better adapted to the target domain. Extensive experiments demonstrate that our framework achieves remarkable and consistent improvements compared to the prior arts on popular benchmarks. Codes and models are available at https://github.com/fy-vision/DiGA

Rui Song, Dai Liu, Dave Zhenyu Chen et al.

In federated learning, all networked clients contribute to the model training cooperatively. However, with model sizes increasing, even sharing the trained partial models often leads to severe communication bottlenecks in underlying networks, especially when communicated iteratively. In this paper, we introduce a federated learning framework FedD3 requiring only one-shot communication by integrating dataset distillation instances. Instead of sharing model updates in other federated learning approaches, FedD3 allows the connected clients to distill the local datasets independently, and then aggregates those decentralized distilled datasets (e.g. a few unrecognizable images) from networks for model training. Our experimental results show that FedD3 significantly outperforms other federated learning frameworks in terms of needed communication volumes, while it provides the additional benefit to be able to balance the trade-off between accuracy and communication cost, depending on usage scenario or target dataset. For instance, for training an AlexNet model on CIFAR-10 with 10 clients under non-independent and identically distributed (Non-IID) setting, FedD3 can either increase the accuracy by over 71% with a similar communication volume, or save 98% of communication volume, while reaching the same accuracy, compared to other one-shot federated learning approaches.

Xiangtong Yao, Zhenshan Bing, Genghang Zhuang et al.

Meta-reinforcement learning (meta-RL) is a promising approach that enables the agent to learn new tasks quickly. However, most meta-RL algorithms show poor generalization in multi-task scenarios due to the insufficient task information provided only by rewards. Language-conditioned meta-RL improves the generalization capability by matching language instructions with the agent's behaviors. While both behaviors and language instructions have symmetry, which can speed up human learning of new knowledge. Thus, combining symmetry and language instructions into meta-RL can help improve the algorithm's generalization and learning efficiency. We propose a dual-MDP meta-reinforcement learning method that enables learning new tasks efficiently with symmetrical behaviors and language instructions. We evaluate our method in multiple challenging manipulation tasks, and experimental results show that our method can greatly improve the generalization and learning efficiency of meta-reinforcement learning. Videos are available at https://tumi6robot.wixsite.com/symmetry/.

Christian Creß, Walter Zimmer, Leah Strand et al.

Data-intensive machine learning based techniques increasingly play a prominent role in the development of future mobility solutions - from driver assistance and automation functions in vehicles, to real-time traffic management systems realized through dedicated infrastructure. The availability of high quality real-world data is often an important prerequisite for the development and reliable deployment of such systems in large scale. Towards this endeavour, we present the A9-Dataset based on roadside sensor infrastructure from the 3 km long Providentia++ test field near Munich in Germany. The dataset includes anonymized and precision-timestamped multi-modal sensor and object data in high resolution, covering a variety of traffic situations. As part of the first set of data, which we describe in this paper, we provide camera and LiDAR frames from two overhead gantry bridges on the A9 autobahn with the corresponding objects labeled with 3D bounding boxes. The first set includes in total more than 1000 sensor frames and 14000 traffic objects. The dataset is available for download at https://a9-dataset.com.

Hu Cao, Zehua Zhang, Yan Xia et al.

In frame-based vision, object detection faces substantial performance degradation under challenging conditions due to the limited sensing capability of conventional cameras. Event cameras output sparse and asynchronous events, providing a potential solution to solve these problems. However, effectively fusing two heterogeneous modalities remains an open issue. In this work, we propose a novel hierarchical feature refinement network for event-frame fusion. The core concept is the design of the coarse-to-fine fusion module, denoted as the cross-modality adaptive feature refinement (CAFR) module. In the initial phase, the bidirectional cross-modality interaction (BCI) part facilitates information bridging from two distinct sources. Subsequently, the features are further refined by aligning the channel-level mean and variance in the two-fold adaptive feature refinement (TAFR) part. We conducted extensive experiments on two benchmarks: the low-resolution PKU-DDD17-Car dataset and the high-resolution DSEC dataset. Experimental results show that our method surpasses the state-of-the-art by an impressive margin of $\textbf{8.0}\%$ on the DSEC dataset. Besides, our method exhibits significantly better robustness (\textbf{69.5}\% versus \textbf{38.7}\%) when introducing 15 different corruption types to the frame images. The code can be found at the link (https://github.com/HuCaoFighting/FRN).

Mingyang Wang, Zhenshan Bing, Xiangtong Yao et al.

Meta-reinforcement learning enables artificial agents to learn from related training tasks and adapt to new tasks efficiently with minimal interaction data. However, most existing research is still limited to narrow task distributions that are parametric and stationary, and does not consider out-of-distribution tasks during the evaluation, thus, restricting its application. In this paper, we propose MoSS, a context-based Meta-reinforcement learning algorithm based on Self-Supervised task representation learning to address this challenge. We extend meta-RL to broad non-parametric task distributions which have never been explored before, and also achieve state-of-the-art results in non-stationary and out-of-distribution tasks. Specifically, MoSS consists of a task inference module and a policy module. We utilize the Gaussian mixture model for task representation to imitate the parametric and non-parametric task variations. Additionally, our online adaptation strategy enables the agent to react at the first sight of a task change, thus being applicable in non-stationary tasks. MoSS also exhibits strong generalization robustness in out-of-distributions tasks which benefits from the reliable and robust task representation. The policy is built on top of an off-policy RL algorithm and the entire network is trained completely off-policy to ensure high sample efficiency. On MuJoCo and Meta-World benchmarks, MoSS outperforms prior works in terms of asymptotic performance, sample efficiency (3-50x faster), adaptation efficiency, and generalization robustness on broad and diverse task distributions.

Chih-Hong Cheng, Saddek Bensalem, Yu-Fang Chen et al.

We present algorithms to synthesize component-based systems that are safe and deadlock-free using priorities, which define stateless-precedence between enabled actions. Our core method combines the concept of fault-localization (using safety-game) and fault-repair (using SAT for conflict resolution). For complex systems, we propose three complementary methods as preprocessing steps for priority synthesis, namely (a) data abstraction to reduce component complexities, (b) alphabet abstraction and #-deadlock to ignore components, and (c) automated assumption learning for compositional priority synthesis.

Rui Song, Runsheng Xu, Andreas Festag et al.

Bird's eye view (BEV) perception is becoming increasingly important in the field of autonomous driving. It uses multi-view camera data to learn a transformer model that directly projects the perception of the road environment onto the BEV perspective. However, training a transformer model often requires a large amount of data, and as camera data for road traffic are often private, they are typically not shared. Federated learning offers a solution that enables clients to collaborate and train models without exchanging data but model parameters. In this paper, we introduce FedBEVT, a federated transformer learning approach for BEV perception. In order to address two common data heterogeneity issues in FedBEVT: (i) diverse sensor poses, and (ii) varying sensor numbers in perception systems, we propose two approaches -- Federated Learning with Camera-Attentive Personalization (FedCaP) and Adaptive Multi-Camera Masking (AMCM), respectively. To evaluate our method in real-world settings, we create a dataset consisting of four typical federated use cases. Our findings suggest that FedBEVT outperforms the baseline approaches in all four use cases, demonstrating the potential of our approach for improving BEV perception in autonomous driving.

Tianhang Wang, Guang Chen, Kai Chen et al.

Multi-agent collaborative perception (MCP) has recently attracted much attention. It includes three key processes: communication for sharing, collaboration for integration, and reconstruction for different downstream tasks. Existing methods pursue designing the collaboration process alone, ignoring their intrinsic interactions and resulting in suboptimal performance. In contrast, we aim to propose a Unified Collaborative perception framework named UMC, optimizing the communication, collaboration, and reconstruction processes with the Multi-resolution technique. The communication introduces a novel trainable multi-resolution and selective-region (MRSR) mechanism, achieving higher quality and lower bandwidth. Then, a graph-based collaboration is proposed, conducting on each resolution to adapt the MRSR. Finally, the reconstruction integrates the multi-resolution collaborative features for downstream tasks. Since the general metric can not reflect the performance enhancement brought by MCP systematically, we introduce a brand-new evaluation metric that evaluates the MCP from different perspectives. To verify our algorithm, we conducted experiments on the V2X-Sim and OPV2V datasets. Our quantitative and qualitative experiments prove that the proposed UMC greatly outperforms the state-of-the-art collaborative perception approaches.

Fengyi Shen, Zador Pataki, Akhil Gurram et al.

Due to the lack of training labels and the difficulty of annotating, dealing with adverse driving conditions such as nighttime has posed a huge challenge to the perception system of autonomous vehicles. Therefore, adapting knowledge from a labelled daytime domain to an unlabelled nighttime domain has been widely researched. In addition to labelled daytime datasets, existing nighttime datasets usually provide nighttime images with corresponding daytime reference images captured at nearby locations for reference. The key challenge is to minimize the performance gap between the two domains. In this paper, we propose LoopDA for domain adaptive nighttime semantic segmentation. It consists of self-loops that result in reconstructing the input data using predicted semantic maps, by rendering them into the encoded features. In a warm-up training stage, the self-loops comprise of an inner-loop and an outer-loop, which are responsible for intra-domain refinement and inter-domain alignment, respectively. To reduce the impact of day-night pose shifts, in the later self-training stage, we propose a co-teaching pipeline that involves an offline pseudo-supervision signal and an online reference-guided signal `DNA' (Day-Night Agreement), bringing substantial benefits to enhance nighttime segmentation. Our model outperforms prior methods on Dark Zurich and Nighttime Driving datasets for semantic segmentation. Code and pretrained models are available at https://github.com/fy-vision/LoopDA.

Rui Song, Liguo Zhou, Venkatnarayanan Lakshminarasimhan et al.

Deep learning is a key approach for the environment perception function of Cooperative Intelligent Transportation Systems (C-ITS) with autonomous vehicles and smart traffic infrastructure. In today's C-ITS, smart traffic participants are capable of timely generating and transmitting a large amount of data. However, these data can not be used for model training directly due to privacy constraints. In this paper, we introduce a federated learning framework coping with Hierarchical Heterogeneity (H2-Fed), which can notably enhance the conventional pre-trained deep learning model. The framework exploits data from connected public traffic agents in vehicular networks without affecting user data privacy. By coordinating existing traffic infrastructure, including roadside units and road traffic clouds, the model parameters are efficiently disseminated by vehicular communications and hierarchically aggregated. Considering the individual heterogeneity of data distribution, computational and communication capabilities across traffic agents and roadside units, we employ a novel method that addresses the heterogeneity of different aggregation layers of the framework architecture, i.e., aggregation in layers of roadside units and cloud. The experiment results indicate that our method can well balance the learning accuracy and stability according to the knowledge of heterogeneity in current communication networks. Comparing to other baseline approaches, the evaluation on federated datasets shows that our framework is more general and capable especially in application scenarios with low communication quality. Even when 90% of the agents are timely disconnected, the pre-trained deep learning model can still be forced to converge stably, and its accuracy can be enhanced from 68% to over 90% after convergence.

Josip Josifovski, Sayantan Auddy, Mohammadhossein Malmir et al.

Domain Randomization (DR) is commonly used for sim2real transfer of reinforcement learning (RL) policies in robotics. Most DR approaches require a simulator with a fixed set of tunable parameters from the start of the training, from which the parameters are randomized simultaneously to train a robust model for use in the real world. However, the combined randomization of many parameters increases the task difficulty and might result in sub-optimal policies. To address this problem and to provide a more flexible training process, we propose Continual Domain Randomization (CDR) for RL that combines domain randomization with continual learning to enable sequential training in simulation on a subset of randomization parameters at a time. Starting from a model trained in a non-randomized simulation where the task is easier to solve, the model is trained on a sequence of randomizations, and continual learning is employed to remember the effects of previous randomizations. Our robotic reaching and grasping tasks experiments show that the model trained in this fashion learns effectively in simulation and performs robustly on the real robot while matching or outperforming baselines that employ combined randomization or sequential randomization without continual learning. Our code and videos are available at https://continual-dr.github.io/.

Rui Song, Anupama Hegde, Numan Senel et al.

Sensor data sharing in vehicular networks can significantly improve the range and accuracy of environmental perception for connected automated vehicles. Different concepts and schemes for dissemination and fusion of sensor data have been developed. It is common to these schemes that measurement errors of the sensors impair the perception quality and can result in road traffic accidents. Specifically, when the measurement error from the sensors (also referred as measurement noise) is unknown and time varying, the performance of the data fusion process is restricted, which represents a major challenge in the calibration of sensors. In this paper, we consider sensor data sharing and fusion in a vehicular network with both, vehicle-to-infrastructure and vehicle-to-vehicle communication. We propose a method, named Bidirectional Feedback Noise Estimation (BiFNoE), in which an edge server collects and caches sensor measurement data from vehicles. The edge estimates the noise and the targets alternately in double dynamic sliding time windows and enhances the distributed cooperative environment sensing at each vehicle with low communication costs. We evaluate the proposed algorithm and data dissemination strategy in an application scenario by simulation and show that the perception accuracy is on average improved by around 80 % with only 12 kbps uplink and 28 kbps downlink bandwidth.

Rui Song, Liguo Zhou, Lingjuan Lyu et al.

Federated learning enables cooperative training among massively distributed clients by sharing their learned local model parameters. However, with increasing model size, deploying federated learning requires a large communication bandwidth, which limits its deployment in wireless networks. To address this bottleneck, we introduce a residual-based federated learning framework (ResFed), where residuals rather than model parameters are transmitted in communication networks for training. In particular, we integrate two pairs of shared predictors for the model prediction in both server-to-client and client-to-server communication. By employing a common prediction rule, both locally and globally updated models are always fully recoverable in clients and the server. We highlight that the residuals only indicate the quasi-update of a model in a single inter-round, and hence contain more dense information and have a lower entropy than the model, comparing to model weights and gradients. Based on this property, we further conduct lossy compression of the residuals by sparsification and quantization and encode them for efficient communication. The experimental evaluation shows that our ResFed needs remarkably less communication costs and achieves better accuracy by leveraging less sensitive residuals, compared to standard federated learning. For instance, to train a 4.08 MB CNN model on CIFAR-10 with 10 clients under non-independent and identically distributed (Non-IID) setting, our approach achieves a compression ratio over 700X in each communication round with minimum impact on the accuracy. To reach an accuracy of 70%, it saves around 99% of the total communication volume from 587.61 Mb to 6.79 Mb in up-streaming and to 4.61 Mb in down-streaming on average for all clients.

Soubarna Banik, Patricia Gschoßmann, Alejandro Mendoza Garcia et al.

Occlusion is an omnipresent challenge in 3D human pose estimation (HPE). In spite of the large amount of research dedicated to 3D HPE, only a limited number of studies address the problem of occlusion explicitly. To fill this gap, we propose to combine exploitation of spatio-temporal features with synthetic occlusion augmentation during training to deal with occlusion. To this end, we build a spatio-temporal 3D HPE model, StridedPoseGraphFormer based on graph convolution and transformers, and train it using occlusion augmentation. Unlike the existing occlusion-aware methods, that are only tested for limited occlusion, we extensively evaluate our method for varying degrees of occlusion. We show that our proposed method compares favorably with the state-of-the-art (SoA). Our experimental results also reveal that in the absence of any occlusion handling mechanism, the performance of SoA 3D HPE methods degrades significantly when they encounter occlusion.

Josip Josifovski, Mohammadhossein Malmir, Noah Klarmann et al.

Randomization is currently a widely used approach in Sim2Real transfer for data-driven learning algorithms in robotics. Still, most Sim2Real studies report results for a specific randomization technique and often on a highly customized robotic system, making it difficult to evaluate different randomization approaches systematically. To address this problem, we define an easy-to-reproduce experimental setup for a robotic reach-and-balance manipulator task, which can serve as a benchmark for comparison. We compare four randomization strategies with three randomized parameters both in simulation and on a real robot. Our results show that more randomization helps in Sim2Real transfer, yet it can also harm the ability of the algorithm to find a good policy in simulation. Fully randomized simulations and fine-tuning show differentiated results and translate better to the real robot than the other approaches tested.

Brian Hsuan-Cheng Liao, Chih-Hong Cheng, Hasan Esen et al.

In this work, we consider the safety-oriented performance of 3D object detectors in autonomous driving contexts. Specifically, despite impressive results shown by the mass literature, developers often find it hard to ensure the safe deployment of these learning-based perception models. Attributing the challenge to the lack of safety-oriented metrics, we hereby present uncompromising spatial constraints (USC), which characterize a simple yet important localization requirement demanding the predictions to fully cover the objects when seen from the autonomous vehicle. The constraints, as we formulate using the perspective and bird's-eye views, can be naturally reflected by quantitative measures, such that having an object detector with a higher score implies a lower risk of collision. Finally, beyond model evaluation, we incorporate the quantitative measures into common loss functions to enable safety-oriented fine-tuning for existing models. With experiments using the nuScenes dataset and a closed-loop simulation, our work demonstrates such considerations of safety notions at the perception level not only improve model performances beyond accuracy but also allow for a more direct linkage to actual system safety.

Syed Qutub, Florian Geissler, Yang Peng et al.

Object detection neural network models need to perform reliably in highly dynamic and safety-critical environments like automated driving or robotics. Therefore, it is paramount to verify the robustness of the detection under unexpected hardware faults like soft errors that can impact a systems perception module. Standard metrics based on average precision produce model vulnerability estimates at the object level rather than at an image level. As we show in this paper, this does not provide an intuitive or representative indicator of the safety-related impact of silent data corruption caused by bit flips in the underlying memory but can lead to an over- or underestimation of typical fault-induced hazards. With an eye towards safety-related real-time applications, we propose a new metric IVMOD (Image-wise Vulnerability Metric for Object Detection) to quantify vulnerability based on an incorrect image-wise object detection due to false positive (FPs) or false negative (FNs) objects, combined with a severity analysis. The evaluation of several representative object detection models shows that even a single bit flip can lead to a severe silent data corruption event with potentially critical safety implications, with e.g., up to (much greater than) 100 FPs generated, or up to approx. 90% of true positives (TPs) are lost in an image. Furthermore, with a single stuck-at-1 fault, an entire sequence of images can be affected, causing temporally persistent ghost detections that can be mistaken for actual objects (covering up to approx. 83% of the image). Furthermore, actual objects in the scene are continuously missed (up to approx. 64% of TPs are lost). Our work establishes a detailed understanding of the safety-related vulnerability of such critical workloads against hardware faults.

Sondos Mohamed, Walter Zimmer, Ross Greer et al.

Accurately detecting 3D objects from monocular images in dynamic roadside scenarios remains a challenging problem due to varying camera perspectives and unpredictable scene conditions. This paper introduces a two-stage training strategy to address these challenges. Our approach initially trains a model on the large-scale synthetic dataset, RoadSense3D, which offers a diverse range of scenarios for robust feature learning. Subsequently, we fine-tune the model on a combination of real-world datasets to enhance its adaptability to practical conditions. Experimental results of the Cube R-CNN model on challenging public benchmarks show a remarkable improvement in detection performance, with a mean average precision rising from 0.26 to 12.76 on the TUM Traffic A9 Highway dataset and from 2.09 to 6.60 on the DAIR-V2X-I dataset when performing transfer learning. Code, data, and qualitative video results are available on the project website: https://roadsense3d.github.io.

Mohammadhossein Malmir, Josip Josifovski, Noah Klarmann et al.

Delayed Markov decision processes (DMDPs) fulfill the Markov property by augmenting the state space of agents with a finite time window of recently committed actions. In reliance on these state augmentations, delay-resolved reinforcement learning algorithms train policies to learn optimal interactions with environments featuring observation or action delays. Although such methods can be directly trained on the real robots, due to sample inefficiency, limited resources, or safety constraints, a common approach is to transfer models trained in simulation to the physical robot. However, robotic simulations rely on approximated models of the physical systems, which hinders the sim2real transfer. In this work, we consider various uncertainties in modeling the robot or environment dynamics as unknown intrinsic disturbances applied to the system input. We introduce the disturbance-augmented Markov decision process (DAMDP) in delayed settings as a novel representation to incorporate disturbance estimation in training on-policy reinforcement learning algorithms. The proposed method is validated across several metrics on learning robotic reaching and pushing tasks and compared with disturbance-unaware baselines. The results show that the disturbance-augmented models can achieve higher stabilization and robustness in the control response, which in turn improves the prospects of successful sim2real transfer.

Chang Liu, Liguo Zhou, Yanliang Huang et al.

Vehicle perception systems strive to achieve comprehensive and rapid visual interpretation of their surroundings for improved safety and navigation. We introduce YOLO-BEV, an efficient framework that harnesses a unique surrounding cameras setup to generate a 2D bird's-eye view of the vehicular environment. By strategically positioning eight cameras, each at a 45-degree interval, our system captures and integrates imagery into a coherent 3x3 grid format, leaving the center blank, providing an enriched spatial representation that facilitates efficient processing. In our approach, we employ YOLO's detection mechanism, favoring its inherent advantages of swift response and compact model structure. Instead of leveraging the conventional YOLO detection head, we augment it with a custom-designed detection head, translating the panoramically captured data into a unified bird's-eye view map of ego car. Preliminary results validate the feasibility of YOLO-BEV in real-time vehicular perception tasks. With its streamlined architecture and potential for rapid deployment due to minimized parameters, YOLO-BEV poses as a promising tool that may reshape future perspectives in autonomous driving systems.

Soubarna Banik, Edvard Avagyan, Sayantan Auddy et al.

Existing skeleton-based 3D human pose estimation methods only predict joint positions. Although the yaw and pitch of bone rotations can be derived from joint positions, the roll around the bone axis remains unresolved. We present PoseGraphNet++ (PGN++), a novel 2D-to-3D lifting Graph Convolution Network that predicts the complete human pose in 3D including joint positions and bone orientations. We employ both node and edge convolutions to utilize the joint and bone features. Our model is evaluated on multiple datasets using both position and rotation metrics. PGN++ performs on par with the state-of-the-art (SoA) on the Human3.6M benchmark. In generalization experiments, it achieves the best results in position and matches the SoA in orientation, showcasing a more balanced performance than the current SoA. PGN++ exploits the mutual relationship of joints and bones resulting in significantly \SB{improved} position predictions, as shown by our ablation results.

Haitao Meng, Changcai Li, Gang Chen et al.

The proposal of Pseudo-Lidar representation has significantly narrowed the gap between visual-based and active Lidar-based 3D object detection. However, current researches exclusively focus on pushing the accuracy improvement of Pseudo-Lidar by taking the advantage of complex and time-consuming neural networks. Seldom explore the profound characteristics of Pseudo-Lidar representation to obtain the promoting opportunities. In this paper, we dive deep into the pseudo Lidar representation and argue that the performance of 3D object detection is not fully dependent on the high precision stereo depth estimation. We demonstrate that even for the unreliable depth estimation, with proper data processing and refining, it can achieve comparable 3D object detection accuracy. With this finding, we further show the possibility that utilizing fast but inaccurate stereo matching algorithms in the Pseudo-Lidar system to achieve low latency responsiveness. In the experiments, we develop a system with a less powerful stereo matching predictor and adopt the proposed refinement schemes to improve the accuracy. The evaluation on the KITTI benchmark shows that the presented system achieves competitive accuracy to the state-of-the-art approaches with only 23 ms computing, showing it is a suitable candidate for deploying to real car-hold applications.

Kejia Chen, Zheng Shen, Yue Zhang et al.

Large Language Models (LLMs) have gained popularity in task planning for long-horizon manipulation tasks. To enhance the validity of LLM-generated plans, visual demonstrations and online videos have been widely employed to guide the planning process. However, for manipulation tasks involving subtle movements but rich contact interactions, visual perception alone may be insufficient for the LLM to fully interpret the demonstration. Additionally, visual data provides limited information on force-related parameters and conditions, which are crucial for effective execution on real robots. In this paper, we introduce an in-context learning framework that incorporates tactile and force-torque information from human demonstrations to enhance LLMs' ability to generate plans for new task scenarios. We propose a bootstrapped reasoning pipeline that sequentially integrates each modality into a comprehensive task plan. This task plan is then used as a reference for planning in new task configurations. Real-world experiments on two different sequential manipulation tasks demonstrate the effectiveness of our framework in improving LLMs' understanding of multi-modal demonstrations and enhancing the overall planning performance.

Haichuan Li, Liguo Zhou, Zhenshan Bing et al.

Several autonomous driving strategies have been applied to autonomous vehicles, especially in the collision avoidance area. The purpose of collision avoidance is achieved by adjusting the trajectory of autonomous vehicles (AV) to avoid intersection or overlap with the trajectory of surrounding vehicles. A large number of sophisticated vision algorithms have been designed for target inspection, classification, and other tasks, such as ResNet, YOLO, etc., which have achieved excellent performance in vision tasks because of their ability to accurately and quickly capture regional features. However, due to the variability of different tasks, the above models achieve good performance in capturing small regions but are still insufficient in correlating the regional features of the input image with each other. In this paper, we aim to solve this problem and develop an algorithm that takes into account the advantages of CNN in capturing regional features while establishing feature correlation between regions using variants of attention. Finally, our model achieves better performance in the test set of L5Kit compared to the other vision models. The average number of collisions is 19.4 per 10000 frames of driving distance, which greatly improves the success rate of collision avoidance.

Yuning Cui, Syed Waqas Zamir, Salman Khan et al.

In the image acquisition process, various forms of degradation, including noise, haze, and rain, are frequently introduced. These degradations typically arise from the inherent limitations of cameras or unfavorable ambient conditions. To recover clean images from degraded versions, numerous specialized restoration methods have been developed, each targeting a specific type of degradation. Recently, all-in-one algorithms have garnered significant attention by addressing different types of degradations within a single model without requiring prior information of the input degradation type. However, these methods purely operate in the spatial domain and do not delve into the distinct frequency variations inherent to different degradation types. To address this gap, we propose an adaptive all-in-one image restoration network based on frequency mining and modulation. Our approach is motivated by the observation that different degradation types impact the image content on different frequency subbands, thereby requiring different treatments for each restoration task. Specifically, we first mine low- and high-frequency information from the input features, guided by the adaptively decoupled spectra of the degraded image. The extracted features are then modulated by a bidirectional operator to facilitate interactions between different frequency components. Finally, the modulated features are merged into the original input for a progressively guided restoration. With this approach, the model achieves adaptive reconstruction by accentuating the informative frequency subbands according to different input degradations. Extensive experiments demonstrate that the proposed method achieves state-of-the-art performance on different image restoration tasks, including denoising, dehazing, deraining, motion deblurring, and low-light image enhancement. Our code is available at https://github.com/c-yn/AdaIR.

Syed Sha Qutub, Neslihan Kose, Rafael Rosales et al.

This paper introduces the Budding Ensemble Architecture (BEA), a novel reduced ensemble architecture for anchor-based object detection models. Object detection models are crucial in vision-based tasks, particularly in autonomous systems. They should provide precise bounding box detections while also calibrating their predicted confidence scores, leading to higher-quality uncertainty estimates. However, current models may make erroneous decisions due to false positives receiving high scores or true positives being discarded due to low scores. BEA aims to address these issues. The proposed loss functions in BEA improve the confidence score calibration and lower the uncertainty error, which results in a better distinction of true and false positives and, eventually, higher accuracy of the object detection models. Both Base-YOLOv3 and SSD models were enhanced using the BEA method and its proposed loss functions. The BEA on Base-YOLOv3 trained on the KITTI dataset results in a 6% and 3.7% increase in mAP and AP50, respectively. Utilizing a well-balanced uncertainty estimation threshold to discard samples in real-time even leads to a 9.6% higher AP50 than its base model. This is attributed to a 40% increase in the area under the AP50-based retention curve used to measure the quality of calibration of confidence scores. Furthermore, BEA-YOLOV3 trained on KITTI provides superior out-of-distribution detection on Citypersons, BDD100K, and COCO datasets compared to the ensembles and vanilla models of YOLOv3 and Gaussian-YOLOv3.

Wei Cao, Liguo Zhou, Yuhong Huang et al.

There are many artificial intelligence algorithms for autonomous driving, but directly installing these algorithms on vehicles is unrealistic and expensive. At the same time, many of these algorithms need an environment to train and optimize. Simulation is a valuable and meaningful solution with training and testing functions, and it can say that simulation is a critical link in the autonomous driving world. There are also many different applications or systems of simulation from companies or academies such as SVL and Carla. These simulators flaunt that they have the closest real-world simulation, but their environment objects, such as pedestrians and other vehicles around the agent-vehicle, are already fixed programmed. They can only move along the pre-setting trajectory, or random numbers determine their movements. What is the situation when all environmental objects are also installed by Artificial Intelligence, or their behaviors are like real people or natural reactions of other drivers? This problem is a blind spot for most of the simulation applications, or these applications cannot be easy to solve this problem. The Neurorobotics Platform from the TUM team of Prof. Alois Knoll has the idea about "Engines" and "Transceiver Functions" to solve the multi-agents problem. This report will start with a little research on the Neurorobotics Platform and analyze the potential and possibility of developing a new simulator to achieve the true real-world simulation goal. Then based on the NRP-Core Platform, this initial development aims to construct an initial demo experiment. The consist of this report starts with the basic knowledge of NRP-Core and its installation, then focus on the explanation of the necessary components for a simulation experiment, at last, about the details of constructions for the autonomous driving system, which is integrated object detection and autonomous control.

Ziou Hu, Xiangtong Yao, Yuan Meng et al.

Diffusion policies excel at visuomotor control but often fail catastrophically under severe out-of-distribution (OOD) disturbances, such as unexpected object displacements or visual corruptions. To address this vulnerability, we introduce the Dream Diffusion Policy (DDP), a framework that deeply integrates a diffusion world model into the policy's training objective via a shared 3D visual encoder. This co-optimization endows the policy with robust state-prediction capabilities. When encountering sudden OOD anomalies during inference, DDP detects the real-imagination discrepancy and actively abandons the corrupted visual stream. Instead, it relies on its internal "imagination" (autoregressively forecasted latent dynamics) to safely bypass the disruption, generating imagined trajectories before smoothly realigning with physical reality. Extensive evaluations demonstrate DDP's exceptional resilience. Notably, DDP achieves a 73.8% OOD success rate on MetaWorld (vs. 23.9% without predictive imagination) and an 83.3% success rate under severe real-world spatial shifts (vs. 3.3% without predictive imagination). Furthermore, as a stress test, DDP maintains a 76.7% real-world success rate even when relying entirely on open-loop imagination post-initialization.

Xinyi Li, Zijian Ma, Yinlong Liu et al.

Point cloud registration is challenging in the presence of heavy outlier correspondences. This paper focuses on addressing the robust correspondence-based registration problem with gravity prior that often arises in practice. The gravity directions are typically obtained by inertial measurement units (IMUs) and can reduce the degree of freedom (DOF) of rotation from 3 to 1. We propose a novel transformation decoupling strategy by leveraging screw theory. This strategy decomposes the original 4-DOF problem into three sub-problems with 1-DOF, 2-DOF, and 1-DOF, respectively, thereby enhancing the computation efficiency. Specifically, the first 1-DOF represents the translation along the rotation axis and we propose an interval stabbing-based method to solve it. The second 2-DOF represents the pole which is an auxiliary variable in screw theory and we utilize a branch-and-bound method to solve it. The last 1-DOF represents the rotation angle and we propose a global voting method for its estimation. The proposed method sequentially solves three consensus maximization sub-problems, leading to efficient and deterministic registration. In particular, it can even handle the correspondence-free registration problem due to its significant robustness. Extensive experiments on both synthetic and real-world datasets demonstrate that our method is more efficient and robust than state-of-the-art methods, even when dealing with outlier rates exceeding 99%.

Xukun Li, Yu Sun, Lei Zhang et al.

Bimanual dexterous manipulation relies on integrating multimodal inputs to perform complex real-world tasks. To address the challenges of effectively combining these modalities, we propose DECO, a decoupled multimodal diffusion transformer that disentangles vision, proprioception, and tactile signals through specialized conditioning pathways, enabling structured and controllable integration of multimodal inputs, with a lightweight adapter for parameter-efficient injection of additional signals. Alongside DECO, we release DECO-50 dataset for bimanual dexterous manipulation with tactile sensing, consisting of 50 hours of data and over 5M frames, collected via teleoperation on real dual-arm robots. We train DECO on DECO-50 and conduct extensive real-world evaluation with over 2,000 robot rollouts. Experimental results show that DECO achieves the best performance across all tasks, with a 72.25% average success rate and a 21% improvement over the baseline. Moreover, the tactile adapter brings an additional 10.25% average success rate across all tasks and a 20% gain on complex contact-rich tasks while tuning less than 10% of the model parameters.

Sanqing Qu, Tianpei Zou, Lianghua He et al.

Universal Domain Adaptation (UniDA) targets knowledge transfer in the presence of both covariate and label shifts. Recently, Source-free Universal Domain Adaptation (SF-UniDA) has emerged to achieve UniDA without access to source data, which tends to be more practical due to data protection policies. The main challenge lies in determining whether covariate-shifted samples belong to target-private unknown categories. Existing methods tackle this either through hand-crafted thresholding or by developing time-consuming iterative clustering strategies. In this paper, we propose a new idea of LEArning Decomposition (LEAD), which decouples features into source-known and -unknown components to identify target-private data. Technically, LEAD initially leverages the orthogonal decomposition analysis for feature decomposition. Then, LEAD builds instance-level decision boundaries to adaptively identify target-private data. Extensive experiments across various UniDA scenarios have demonstrated the effectiveness and superiority of LEAD. Notably, in the OPDA scenario on VisDA dataset, LEAD outperforms GLC by 3.5% overall H-score and reduces 75% time to derive pseudo-labeling decision boundaries. Besides, LEAD is also appealing in that it is complementary to most existing methods. The code is available at https://github.com/ispc-lab/LEAD.

Yuping Wang, Shuo Xing, Cui Can et al.

Generative Artificial Intelligence (GenAI) constitutes a transformative technological wave that reconfigures industries through its unparalleled capabilities for content creation, reasoning, planning, and multimodal understanding. This revolutionary force offers the most promising path yet toward solving one of engineering's grandest challenges: achieving reliable, fully autonomous driving, particularly the pursuit of Level 5 autonomy. This survey delivers a comprehensive and critical synthesis of the emerging role of GenAI across the autonomous driving stack. We begin by distilling the principles and trade-offs of modern generative modeling, encompassing VAEs, GANs, Diffusion Models, and Large Language Models (LLMs). We then map their frontier applications in image, LiDAR, trajectory, occupancy, video generation as well as LLM-guided reasoning and decision making. We categorize practical applications, such as synthetic data workflows, end-to-end driving strategies, high-fidelity digital twin systems, smart transportation networks, and cross-domain transfer to embodied AI. We identify key obstacles and possibilities such as comprehensive generalization across rare cases, evaluation and safety checks, budget-limited implementation, regulatory compliance, ethical concerns, and environmental effects, while proposing research plans across theoretical assurances, trust metrics, transport integration, and socio-technical influence. By unifying these threads, the survey provides a forward-looking reference for researchers, engineers, and policymakers navigating the convergence of generative AI and advanced autonomous mobility. An actively maintained repository of cited works is available at https://github.com/taco-group/GenAI4AD.

Marco Oliva, Soubarna Banik, Josip Josifovski et al.

State-of-the-art reinforcement learning algorithms predominantly learn a policy from either a numerical state vector or images. Both approaches generally do not take structural knowledge of the task into account, which is especially prevalent in robotic applications and can benefit learning if exploited. This work introduces a neural network architecture that combines relational inductive bias and visual feedback to learn an efficient position control policy for robotic manipulation. We derive a graph representation that models the physical structure of the manipulator and combines the robot's internal state with a low-dimensional description of the visual scene generated by an image encoding network. On this basis, a graph neural network trained with reinforcement learning predicts joint velocities to control the robot. We further introduce an asymmetric approach of training the image encoder separately from the policy using supervised learning. Experimental results demonstrate that, for a 2-DoF planar robot in a geometrically simplistic 2D environment, a learned representation of the visual scene can replace access to the explicit coordinates of the reaching target without compromising on the quality and sample efficiency of the policy. We further show the ability of the model to improve sample efficiency for a 6-DoF robot arm in a visually realistic 3D environment.

Mert Keser, Gesina Schwalbe, Niki Amini-Naieni et al.

Ontological commitment, i.e., used concepts, relations, and assumptions, are a corner stone of qualitative reasoning (QR) models. The state-of-the-art for processing raw inputs, though, are deep neural networks (DNNs), nowadays often based off from multimodal foundation models. These automatically learn rich representations of concepts and respective reasoning. Unfortunately, the learned qualitative knowledge is opaque, preventing easy inspection, validation, or adaptation against available QR models. So far, it is possible to associate pre-defined concepts with latent representations of DNNs, but extractable relations are mostly limited to semantic similarity. As a next step towards QR for validation and verification of DNNs: Concretely, we propose a method that extracts the learned superclass hierarchy from a multimodal DNN for a given set of leaf concepts. Under the hood we (1) obtain leaf concept embeddings using the DNN's textual input modality; (2) apply hierarchical clustering to them, using that DNNs encode semantic similarities via vector distances; and (3) label the such-obtained parent concepts using search in available ontologies from QR. An initial evaluation study shows that meaningful ontological class hierarchies can be extracted from state-of-the-art foundation models. Furthermore, we demonstrate how to validate and verify a DNN's learned representations against given ontologies. Lastly, we discuss potential future applications in the context of QR.

Dingyi Zhou, Mu He, Zhuowei Fang et al.

We introduce AffordanceGrasp-R1, a reasoning-driven affordance segmentation framework for robotic grasping that combines a chain-of-thought (CoT) cold-start strategy with reinforcement learning to enhance deduction and spatial grounding. In addition, we redesign the grasping pipeline to be more context-aware by generating grasp candidates from the global scene point cloud and subsequently filtering them using instruction-conditioned affordance masks. Extensive experiments demonstrate that AffordanceGrasp-R1 consistently outperforms state-of-the-art (SOTA) methods on benchmark datasets, and real-world robotic grasping evaluations further validate its robustness and generalization under complex language-conditioned manipulation scenarios.

Liguo Zhou, Yinglei Song, Yichao Gao et al.

Conducting real road testing for autonomous driving algorithms can be expensive and sometimes impractical, particularly for small startups and research institutes. Thus, simulation becomes an important method for evaluating these algorithms. However, the availability of free and open-source simulators is limited, and the installation and configuration process can be daunting for beginners and interdisciplinary researchers. We introduce an autonomous driving simulator with photorealistic scenes, meanwhile keeping a user-friendly workflow. The simulator is able to communicate with external algorithms through ROS2 or Socket.IO, making it compatible with existing software stacks. Furthermore, we implement a highly accurate vehicle dynamics model within the simulator to enhance the realism of the vehicle's physical effects. The simulator is able to serve various functions, including generating synthetic data and driving with machine learning-based algorithms. Moreover, we prioritize simplicity in the deployment process, ensuring that beginners find it approachable and user-friendly.

Jiong Liu, Yingjie Xu, Xingcheng Zhou et al.

Semantic segmentation across arbitrary sensor modalities faces significant challenges due to diverse sensor characteristics, and the traditional configurations for this task result in redundant development efforts. We address these challenges by introducing a universal arbitrary-modal semantic segmentation framework that unifies segmentation across multiple modalities. Our approach features three key innovations: (1) the Modality-aware CLIP (MA-CLIP), which provides modality-specific scene understanding guidance through LoRA fine-tuning; (2) Modality-aligned Embeddings for capturing fine-grained features; and (3) the Domain-specific Refinement Module (DSRM) for dynamic feature adjustment. Evaluated on five diverse datasets with different complementary modalities (event, thermal, depth, polarization, and light field), our model surpasses specialized multi-modal methods and achieves state-of-the-art performance with a mIoU of 65.03%. The codes will be released upon acceptance.

Panagiotis Petropoulakis, Ludwig Gräf, Mohammadhossein Malmir et al.

Choosing an appropriate representation of the environment for the underlying decision-making process of the reinforcement learning agent is not always straightforward. The state representation should be inclusive enough to allow the agent to informatively decide on its actions and disentangled enough to simplify policy training and the corresponding sim2real transfer. Given this outlook, this work examines the effect of various representations in incentivizing the agent to solve a specific robotic task: antipodal and planar object grasping. A continuum of state representations is defined, starting from hand-crafted numerical states to encoded image-based representations, with decreasing levels of induced task-specific knowledge. The effects of each representation on the ability of the agent to solve the task in simulation and the transferability of the learned policy to the real robot are examined and compared against a model-based approach with complete system knowledge. The results show that reinforcement learning agents using numerical states can perform on par with non-learning baselines. Furthermore, we find that agents using image-based representations from pre-trained environment embedding vectors perform better than end-to-end trained agents, and hypothesize that separation of representation learning from reinforcement learning can benefit sim2real transfer. Finally, we conclude that incentivizing the state representation with task-specific knowledge facilitates faster convergence for agent training and increases success rates in sim2real robot control.

Xingcheng Zhou, Hao Guo, Rui Song et al.

Safety-critical traffic reasoning requires contrastive consistency: models must detect true hazards when an accident occurs, and reliably reject plausible-but-false hypotheses under near-identical counterfactual scenes. We present CCTVBench, a Contrastive Consistency Traffic VideoQA Benchmark built on paired real accident videos and world-model-generated counterfactual counterparts, together with minimally different, mutually exclusive hypothesis questions. CCTVBench enforces a single structured decision pattern over each video question quadruple and provides actionable diagnostics that decompose failures into positive omission, positive swap, negative hallucination, and mutual-exclusivity violation, while separating video versus question consistency. Experiments across open-source and proprietary video LLMs reveal a large and persistent gap between standard per-instance QA metrics and quadruple-level contrastive consistency, with unreliable none-of-the-above rejection as a key bottleneck. Finally, we introduce C-TCD, a contrastive decoding approach leveraging a semantically exclusive counterpart video as the contrast input at inference time, improving both instance-level QA and contrastive consistency.

Mert Keser, Youssef Shoeb, Alois Knoll

Deep Neural Networks (DNNs) have become central for the perception functions of autonomous vehicles, substantially enhancing their ability to understand and interpret the environment. However, these systems exhibit inherent limitations such as brittleness, opacity, and unpredictable behavior in out-of-distribution scenarios. The European Union (EU) Artificial Intelligence (AI) Act, as a pioneering legislative framework, aims to address these challenges by establishing stringent norms and standards for AI systems, including those used in autonomous driving (AD), which are categorized as high-risk AI. In this work, we explore how the newly available generative AI models can potentially support addressing upcoming regulatory requirements in AD perception, particularly with respect to safety. This short review paper summarizes the requirements arising from the EU AI Act regarding DNN-based perception systems and systematically categorizes existing generative AI applications in AD. While generative AI models show promise in addressing some of the EU AI Acts requirements, such as transparency and robustness, this review examines their potential benefits and discusses how developers could leverage these methods to enhance compliance with the Act. The paper also highlights areas where further research is needed to ensure reliable and safe integration of these technologies.

Xingcheng Zhou, Mingyu Liu, Walter Zimmer et al.

We present Scene-Graph Based Multi-Modal Traffic Agent (SGTA), a modular framework for traffic video understanding that combines structured scene graphs with multi-modal reasoning. It constructs a traffic scene graph from roadside videos using detection, tracking, and lane extraction, followed by tool-based reasoning over both symbolic graph queries and visual inputs. SGTA adopts ReAct to process interleaved reasoning traces from large language models with tool invocations, enabling interpretable decision-making for complex video questions. Experiments on selected TUMTraffic VideoQA dataset sample demonstrate that SGTA achieves competitive accuracy across multiple question types while providing transparent reasoning steps. These results highlight the potential of integrating structured scene representations with multi-modal agents for traffic video understanding.