Lifeng Zhou

Zijian An, Augustus Sroka, Ran Yang et al.

Orientation-aware manipulation is essential in post-harvest agricultural processing, where produce must be grasped and placed in consistent configurations. This paper presents ROG-Grasp, a geometry-based robotic grasping and placement framework that estimates the produce orientation from root surface geometry using RGB-D perception. A YOLO-based root detector and point cloud plane fitting are used to infer the root normal, enabling stable grasp pose generation and orientation-constrained Cartesian motion planning. Experiments on tomatoes and onions demonstrate high success rates and stable execution time in both isolated and cluttered scenarios. Compared with vision-language-action (VLA) policies, the proposed method achieves more reliable and accurate grasp completion with faster execution. These results highlight the effectiveness of geometry-driven perception for practical orientation-controlled manipulation tasks. A video of our paper is available online https://youtu.be/Ir2UtGODdMo.

Zijian An, Ran Yang, Yiming Feng et al.

Vision-language-action (VLA) models have recently emerged as a promising paradigm for robotic control, enabling end-to-end policies that ground natural language instructions into visuomotor actions. However, current VLAs often struggle to satisfy precise task constraints, such as stopping based on numeric thresholds, since their observation-to-action mappings are implicitly shaped by training data and lack explicit mechanisms for condition monitoring. In this work, we propose CLAW (CLIP-Language-Action for Weight), a framework that decouples condition evaluation from action generation. CLAW leverages a fine-tuned CLIP model as a lightweight prompt generator, which continuously monitors the digital readout of a scale and produces discrete directives based on task-specific weight thresholds. These prompts are then consumed by $π_0$, a flow-based VLA policy, which integrates the prompts with multi-view camera observations to produce continuous robot actions. This design enables CLAW to combine symbolic weight reasoning with high-frequency visuomotor control. We validate CLAW on three experimental setups: single-object grasping and mixed-object tasks requiring dual-arm manipulation. Across all conditions, CLAW reliably executes weight-aware behaviors and outperforms both raw-$π_0$ and fine-tuned $π_0$ models. A video of our paper is available online https://youtu.be/MuMYj2QgReI.

Keith Truongcao, Christopher Nhu, Zijian An et al.

Vision-Language Action (VLA) models continue to face challenges such as slow inference speed and difficulty performing fine-grained motion adjustments, limiting their widespread adoption in industry. While the Real-Time Action Chunking (RTAC) algorithm has been proposed to address these bottlenecks, bridging the gap between the algorithm provided in pseudocode to a stable, real-world deployment on a low-cost robotic arm remains a challenge. In this work, we present a complete system-level implementation of RTAC tailored for a low-cost robotic manipulation system. We advance beyond the original high-level pseudocode by optimizing the threading implementation for the policy inference and control pipeline, reducing end-to-end latency and improving responsiveness without modifying the underlying policy. We evaluate this system on tasks involving the manipulation of agricultural produce, specifically garlic bulbs and walnuts. Experimental results demonstrate that our custom threading implementation significantly improves control stability and speed compared to the base implementation of RTAC.

Haonan Chang, Kowndinya Boyalakuntla, Shiyang Lu et al.

We present an Open-Vocabulary 3D Scene Graph (OVSG), a formal framework for grounding a variety of entities, such as object instances, agents, and regions, with free-form text-based queries. Unlike conventional semantic-based object localization approaches, our system facilitates context-aware entity localization, allowing for queries such as ``pick up a cup on a kitchen table" or ``navigate to a sofa on which someone is sitting". In contrast to existing research on 3D scene graphs, OVSG supports free-form text input and open-vocabulary querying. Through a series of comparative experiments using the ScanNet dataset and a self-collected dataset, we demonstrate that our proposed approach significantly surpasses the performance of previous semantic-based localization techniques. Moreover, we highlight the practical application of OVSG in real-world robot navigation and manipulation experiments.

Elijah S. Lee, Lifeng Zhou, Alejandro Ribeiro et al.

In this work, we study the problem of decentralized multi-agent perimeter defense that asks for computing actions for defenders with local perceptions and communications to maximize the capture of intruders. One major challenge for practical implementations is to make perimeter defense strategies scalable for large-scale problem instances. To this end, we leverage graph neural networks (GNNs) to develop an imitation learning framework that learns a mapping from defenders' local perceptions and their communication graph to their actions. The proposed GNN-based learning network is trained by imitating a centralized expert algorithm such that the learned actions are close to that generated by the expert algorithm. We demonstrate that our proposed network performs closer to the expert algorithm and is superior to other baseline algorithms by capturing more intruders. Our GNN-based network is trained at a small scale and can be generalized to large-scale cases. We run perimeter defense games in scenarios with different team sizes and configurations to demonstrate the performance of the learned network.

Zhongshun Zhang, Lifeng Zhou, Pratap Tokekar

We study the problem of designing false measurement data that is injected to corrupt and mislead the output of a Kalman filter. Unlike existing works that focus on detection and filtering algorithms for the observer, we study the problem from the attacker's point-of-view. In our model, the attacker can corrupt the measurements by injecting additive spoofing signals. The attacker seeks to create a separation between the estimate of the Kalman filter with and without spoofed signals. We present a number of results on how to inject spoofing signals while minimizing the magnitude of the injected signals. The resulting strategies are evaluated through simulations along with theoretical proofs. We also evaluate the spoofing strategy in the presence of a $χ^2$ spoof detector. Building on our main result, we present a strategy that is proven to successfully mislead a Kalman filter while ensuring it is not detected.

Elijah S. Lee, Lifeng Zhou, Alejandro Ribeiro et al.

We consider the problem of finding decentralized strategies for multi-agent perimeter defense games. In this work, we design a graph neural network-based learning framework to learn a mapping from defenders' local perceptions and the communication graph to defenders' actions such that the learned actions are close to that generated by a centralized expert algorithm. We demonstrate that our proposed networks stay closer to the expert policy and are superior to other baseline algorithms by capturing more intruders. Our GNN-based networks are trained at a small scale and can generalize to large scales. To validate our results, we run perimeter defense games in scenarios with different team sizes and initial configurations to evaluate the performance of the learned networks.

Peihan Li, Zijian An, Shams Abrar et al.

The rapid advancement of Large Language Models (LLMs) has opened new possibilities in Multi-Robot Systems (MRS), enabling enhanced communication, task planning, and human-robot interaction. Unlike traditional single-robot and multi-agent systems, MRS poses unique challenges, including coordination, scalability, and real-world adaptability. This survey provides the first comprehensive exploration of LLM integration into MRS. It systematically categorizes their applications across high-level task allocation, mid-level motion planning, low-level action generation, and human intervention. We highlight key applications in diverse domains, such as household robotics, construction, formation control, target tracking, and robot games, showcasing the versatility and transformative potential of LLMs in MRS. Furthermore, we examine the challenges that limit adapting LLMs in MRS, including mathematical reasoning limitations, hallucination, latency issues, and the need for robust benchmarking systems. Finally, we outline opportunities for future research, emphasizing advancements in fine-tuning, reasoning techniques, and task-specific models. This survey aims to guide researchers in the intelligence and real-world deployment of MRS powered by LLMs. Based on the fast-evolving nature of research in the field, we keep updating the papers in the open-source GitHub repository.

Murat Isik, Matthew Oldland, Lifeng Zhou

Autoencoders are unsupervised neural networks that are used to process and compress input data and then reconstruct the data back to the original data size. This allows autoencoders to be used for different processing applications such as data compression, image classification, image noise reduction, and image coloring. Hardware-wise, re-configurable architectures like Field Programmable Gate Arrays (FPGAs) have been used for accelerating computations from several domains because of their unique combination of flexibility, performance, and power efficiency. In this paper, we look at the different autoencoders available and use the convolutional autoencoder in both FPGA and GPU-based implementations to process noisy static MNIST images. We compare the different results achieved with the FPGA and GPU-based implementations and then discuss the pros and cons of each implementation. The evaluation of the proposed design achieved 80%accuracy and our experimental results show that the proposed accelerator achieves a throughput of 21.12 Giga-Operations Per Second (GOP/s) with a 5.93 W on-chip power consumption at 100 MHz. The comparison results with off-the-shelf devices and recent state-of-the-art implementations illustrate that the proposed accelerator has obvious advantages in terms of energy efficiency and design flexibility. We also discuss future work that can be done with the use of our proposed accelerator.

Siji Chen, Yanshen Sun, Peihan Li et al.

Recently a line of researches has delved the use of graph neural networks (GNNs) for decentralized control in swarm robotics. However, it has been observed that relying solely on the states of immediate neighbors is insufficient to imitate a centralized control policy. To address this limitation, prior studies proposed incorporating $L$-hop delayed states into the computation. While this approach shows promise, it can lead to a lack of consensus among distant flock members and the formation of small clusters, consequently resulting in the failure of cohesive flocking behaviors. Instead, our approach leverages spatiotemporal GNN, named STGNN that encompasses both spatial and temporal expansions. The spatial expansion collects delayed states from distant neighbors, while the temporal expansion incorporates previous states from immediate neighbors. The broader and more comprehensive information gathered from both expansions results in more effective and accurate predictions. We develop an expert algorithm for controlling a swarm of robots and employ imitation learning to train our decentralized STGNN model based on the expert algorithm. We simulate the proposed STGNN approach in various settings, demonstrating its decentralized capacity to emulate the global expert algorithm. Further, we implemented our approach to achieve cohesive flocking, leader following and obstacle avoidance by a group of Crazyflie drones. The performance of STGNN underscores its potential as an effective and reliable approach for achieving cohesive flocking, leader following and obstacle avoidance tasks.

Aref Ghoreishee, Abhishek Mishra, Lifeng Zhou et al.

This paper develops an end-to-end fuzzy encoder-decoder architecture for enhancing vision-based multi-modal deep spiking Q-networks in autonomous driving. The method addresses two core limitations of spiking reinforcement learning: information loss stemming from the conversion of dense visual inputs into sparse spike trains, and the limited representational capacity of spike-based value functions, which often yields weakly discriminative Q-value estimates. The encoder introduces trainable fuzzy membership functions to generate expressive, population-based spike representations, and the decoder uses a lightweight neural decoder to reconstruct continuous Q-values from spiking outputs. Experiments on the HighwayEnv benchmark show that the proposed architecture substantially improves decision-making accuracy and closes the performance gap between spiking and non-spiking multi-modal Q-networks. The results highlight the potential of this framework for efficient and real-time autonomous driving with spiking neural networks.

Hanqing Liu, Lifeng Zhou, Huanqian Yan

Large language models have drawn significant attention to the challenge of safe alignment, especially regarding jailbreak attacks that circumvent security measures to produce harmful content. To address the limitations of existing methods like GCG, which perform well in single-model attacks but lack transferability, we propose several enhancements, including a scenario induction template, optimized suffix selection, and the integration of re-suffix attack mechanism to reduce inconsistent outputs. Our approach has shown superior performance in extensive experiments across various benchmarks, achieving nearly 100% success rates in both attack execution and transferability. Notably, our method has won the first place in the AISG-hosted Global Challenge for Safe and Secure LLMs. The code is released at https://github.com/HqingLiu/SI-GCG.

Amirhosein Chahe, Lifeng Zhou

This paper introduces a novel method for open-vocabulary 3D scene querying in autonomous driving by combining Language Embedded 3D Gaussians with Large Language Models (LLMs). We propose utilizing LLMs to generate both contextually canonical phrases and helping positive words for enhanced segmentation and scene interpretation. Our method leverages GPT-3.5 Turbo as an expert model to create a high-quality text dataset, which we then use to fine-tune smaller, more efficient LLMs for on-device deployment. Our comprehensive evaluation on the WayveScenes101 dataset demonstrates that LLM-guided segmentation significantly outperforms traditional approaches based on predefined canonical phrases. Notably, our fine-tuned smaller models achieve performance comparable to larger expert models while maintaining faster inference times. Through ablation studies, we discover that the effectiveness of helping positive words correlates with model scale, with larger models better equipped to leverage additional semantic information. This work represents a significant advancement towards more efficient, context-aware autonomous driving systems, effectively bridging 3D scene representation with high-level semantic querying while maintaining practical deployment considerations.

Lifeng Zhou, Yuke Li, Rui Deng et al.

The success of speech-image retrieval relies on establishing an effective alignment between speech and image. Existing methods often model cross-modal interaction through simple cosine similarity of the global feature of each modality, which fall short in capturing fine-grained details within modalities. To address this issue, we introduce an effective framework and a novel learning task named cross-modal denoising (CMD) to enhance cross-modal interaction to achieve finer-level cross-modal alignment. Specifically, CMD is a denoising task designed to reconstruct semantic features from noisy features within one modality by interacting features from another modality. Notably, CMD operates exclusively during model training and can be removed during inference without adding extra inference time. The experimental results demonstrate that our framework outperforms the state-of-the-art method by 2.0% in mean R@1 on the Flickr8k dataset and by 1.7% in mean R@1 on the SpokenCOCO dataset for the speech-image retrieval tasks, respectively. These experimental results validate the efficiency and effectiveness of our framework.

Aref Ghoreishee, Abhishek Mishra, Lifeng Zhou et al.

This work proposes an end-to-end multi-modal reinforcement learning framework for high-level decision-making in autonomous vehicles. The framework integrates heterogeneous sensory input, including camera images, LiDAR point clouds, and vehicle heading information, through a cross-attention transformer-based perception module. Although transformers have become the backbone of modern multi-modal architectures, their high computational cost limits their deployment in resource-constrained edge environments. To overcome this challenge, we propose a spiking temporal-aware transformer-like architecture that uses ternary spiking neurons for computationally efficient multi-modal fusion. Comprehensive evaluations across multiple tasks in the Highway Environment demonstrate the effectiveness and efficiency of the proposed approach for real-time autonomous decision-making.

Siwei Cai, Knut Peterson, Quan Tran et al.

Heterogeneous air-ground robot teams combine complementary sensing modalities, mobility characteristics, and spatial viewpoints that can significantly enhance perception in complex outdoor environments. However, progress in multi-robot collaborative perception has been constrained by the lack of real-world datasets featuring overlapping multi-modal observations from platforms operating in unstructured terrain. We present GA3T (Ground-Aerial Team for Terrain Traversal), a real-world multi-robot collaborative perception dataset collected using a Clearpath Husky UGV and an Autel EVO~II UAV across diverse unstructured environments, including forest trails, rocky paths, muddy terrain, snow piles, and grass-covered fields. The ground platform provides 3D LiDAR, stereo camera, IMU, and GPS data, while the aerial platform contributes RGB imagery, thermal/infrared observations, and GPS from a complementary overhead viewpoint, allowing for rich cross-modal and cross-view perception. The dataset is collected in 4 unique environments, with over 13,000 synchronized frames across approximately 29 minutes of operation, and includes both SAM~3-based zero-shot segmentation and over 8,000 manually labeled images. A unique aspect of the dataset is its early-spring collection period, during which sparse tree canopies allow the aerial robot to partially observe the ground robot and terrain through the trees, allowing for occlusion-aware collaborative perception. Unlike prior multi-robot datasets that focus on SLAM or simulated cooperative driving, GA3T is specifically designed to support research on cross-view perception, air-ground viewpoint fusion, traversability estimation, and collaborative scene understanding in real off-road environments.

Amirhosein Chahe, Lifeng Zhou

Navigating to a visually specified goal given natural language instructions remains a fundamental challenge in embodied AI. Existing approaches either rely on reactive policies that struggle with long-horizon planning, or employ world models that suffer from poor action initialization in high-dimensional spaces. We present PiJEPA, a two-stage framework that combines the strengths of learned navigation policies with latent world model planning for instruction-conditioned visual navigation. In the first stage, we finetune an Octo-based generalist policy, augmented with a frozen pretrained vision encoder (DINOv2 or V-JEPA-2), on the CAST navigation dataset to produce an informed action distribution conditioned on the current observation and language instruction. In the second stage, we use this policy-derived distribution to warm-start Model Predictive Path Integral (MPPI) planning over a separately trained JEPA world model, which predicts future latent states in the embedding space of the same frozen encoder. By initializing the MPPI sampling distribution from the policy prior rather than from an uninformed Gaussian, our planner converges faster to high-quality action sequences that reach the goal. We systematically study the effect of the vision encoder backbone, comparing DINOv2 and V-JEPA-2, across both the policy and world model components. Experiments on real-world navigation tasks demonstrate that PiJEPA significantly outperforms both standalone policy execution and uninformed world model planning, achieving improved goal-reaching accuracy and instruction-following fidelity.

Lifeng Zhou, Yuke Li

In this paper, we propose a novel framework for speech-image retrieval. We utilize speech-image contrastive (SIC) learning tasks to align speech and image representations at a coarse level and speech-image matching (SIM) learning tasks to further refine the fine-grained cross-modal alignment. SIC and SIM learning tasks are jointly trained in a unified manner. To optimize the learning process, we utilize an embedding queue that facilitates efficient sampling of high-quality and diverse negative representations during SIC learning. Additionally, it enhances the learning of SIM tasks by effectively mining hard negatives based on contrastive similarities calculated in SIC tasks. To further optimize learning under noisy supervision, we incorporate momentum distillation into the training process. Experimental results show that our framework outperforms the state-of-the-art method by more than 4% in R@1 on two benchmark datasets for the speech-image retrieval tasks. Moreover, as observed in zero-shot experiments, our framework demonstrates excellent generalization capabilities.

Zijian An, Hadi Khezam, Bill Cai et al.

We present VILAS, a fully low-cost, modular robotic manipulation platform designed to support end-to-end vision-language-action (VLA) policy learning and deployment on accessible hardware. The system integrates a Fairino FR5 collaborative arm, a Jodell RG52-50 electric gripper, and a dual-camera perception module, unified through a ZMQ-based communication architecture that seamlessly coordinates teleoperation, data collection, and policy deployment within a single framework. To enable safe manipulation of fragile objects without relying on explicit force sensing, we design a kirigami-based soft compliant gripper extension that induces predictable deformation under compressive loading, providing gentle and repeatable contact with delicate targets. We deploy and evaluate three state-of-the-art VLA models on the VILAS platform: pi_0, pi_0.5, and GR00T N1.6. All models are fine-tuned from publicly released pretrained checkpoints using an identical demonstration dataset collected via our teleoperation pipeline. Experiments on a grape grasping task validate the effectiveness of the proposed system, confirming that capable manipulation policies can be successfully trained and deployed on low-cost modular hardware. Our results further provide practical insights into the deployment characteristics of current VLA models in real-world settings.

Peihan Li, Joanna Gutierrez, Fabian Hernandez et al.

Swarm foraging algorithms, such as the central-place foraging algorithm (CPFA), typically rely on offline parameter optimization using genetic algorithms (GA) or reinforcement learning, yielding policies tightly coupled to a specific combination of team size, arena size, and resource distribution. When deployment conditions change, performance degrades, and retraining is computationally expensive. We propose LLM-Foraging, a decentralized swarm controller that augments the CPFA state machine with a large language model (LLM) tactical decision-maker at three structured decision points, namely post-deposit, central-zone arrival, and search starvation. Each robot runs its own LLM client and queries it using only locally observable state, while the existing CPFA motion and sensing stack executes the selected action. Because the LLM serves as a general decision policy rather than parameters fitted to a single configuration, the controller is training-free at deployment and transfers across configurations without re-optimization. We evaluate LLM-Foraging in Gazebo with TurtleBot3 robots across 36 configurations spanning team sizes of 4 to 10 robots, arena sizes from 6x6 to 10x10 meters, and three resource distributions (clustered, powerlaw, random). LLM-Foraging collects more resources than the GA-tuned CPFA baseline across the evaluated configurations and is more consistent, a property that the GA's single-configuration tuning does not transfer.

Yuyang Ji, Yixuan Shen, Kien Nguyen et al.

Video-based person recognition achieves robust identification by integrating face, body, and gait. However, current systems waste computational resources by processing all modalities with fixed heavyweight ensembles regardless of input complexity. To address these limitations, we propose IDSelect, a reinforcement learning-based cost-aware selector that chooses one pre-trained model per modality per-sequence to optimize the accuracy-efficiency trade-off. Our key insight is that an input-conditioned selector can discover complementary model choices that surpass fixed ensembles while using substantially fewer resources. IDSelect trains a lightweight agent end-to-end using actor-critic reinforcement learning with budget-aware optimization. The reward balances recognition accuracy with computational cost, while entropy regularization prevents premature convergence. At inference, the policy selects the most probable model per modality and fuses modality-specific similarities for the final score. Extensive experiments on challenging video-based datasets demonstrate IDSelect's superior efficiency: on CCVID, it achieves 95.9% Rank-1 accuracy with 92.4% less computation than strong baselines while improving accuracy by 1.8%; on MEVID, it reduces computation by 41.3% while maintaining competitive performance.

Peihan Li, Vishnu Menon, Bhavanaraj Gudiguntla et al.

Flocking is a behavior where multiple agents in a system attempt to stay close to each other while avoiding collision and maintaining a desired formation. This is observed in the natural world and has applications in robotics, including natural disaster search and rescue, wild animal tracking, and perimeter surveillance and patrol. Recently, large language models (LLMs) have displayed an impressive ability to solve various collaboration tasks as individual decision-makers. Solving multi-agent flocking with LLMs would demonstrate their usefulness in situations requiring spatial and decentralized decision-making. Yet, when LLM-powered agents are tasked with implementing multi-agent flocking, they fall short of the desired behavior. After extensive testing, we find that agents with LLMs as individual decision-makers typically opt to converge on the average of their initial positions or diverge from each other. After breaking the problem down, we discover that LLMs cannot understand maintaining a shape or keeping a distance in a meaningful way. Solving multi-agent flocking with LLMs would enhance their ability to understand collaborative spatial reasoning and lay a foundation for addressing more complex multi-agent tasks. This paper discusses the challenges LLMs face in multi-agent flocking and suggests areas for future improvement and research.

Xiaojun Jia, Yihao Huang, Yang Liu et al.

This paper introduces the Global Challenge for Safe and Secure Large Language Models (LLMs), a pioneering initiative organized by AI Singapore (AISG) and the CyberSG R&D Programme Office (CRPO) to foster the development of advanced defense mechanisms against automated jailbreaking attacks. With the increasing integration of LLMs in critical sectors such as healthcare, finance, and public administration, ensuring these models are resilient to adversarial attacks is vital for preventing misuse and upholding ethical standards. This competition focused on two distinct tracks designed to evaluate and enhance the robustness of LLM security frameworks. Track 1 tasked participants with developing automated methods to probe LLM vulnerabilities by eliciting undesirable responses, effectively testing the limits of existing safety protocols within LLMs. Participants were challenged to devise techniques that could bypass content safeguards across a diverse array of scenarios, from offensive language to misinformation and illegal activities. Through this process, Track 1 aimed to deepen the understanding of LLM vulnerabilities and provide insights for creating more resilient models.

Guodong Ma, Wenxuan Wang, Lifeng Zhou et al.

Recently, the Mixture of Expert (MoE) architecture, such as LR-MoE, is often used to alleviate the impact of language confusion on the multilingual ASR (MASR) task. However, it still faces language confusion issues, especially in mismatched domain scenarios. In this paper, we decouple language confusion in LR-MoE into confusion in self-attention and router. To alleviate the language confusion in self-attention, based on LR-MoE, we propose to apply attention-MoE architecture for MASR. In our new architecture, MoE is utilized not only on feed-forward network (FFN) but also on self-attention. In addition, to improve the robustness of the LID-based router on language confusion, we propose expert pruning and router augmentation methods. Combining the above, we get the boosted language-routing MoE (BLR-MoE) architecture. We verify the effectiveness of the proposed BLR-MoE in a 10,000-hour MASR dataset.

Muyu He, Yuxi Zheng, Yuchen Liu et al.

The advancement of vision language models (VLMs) has empowered embodied agents to accomplish simple multimodal planning tasks, but not long-horizon ones requiring long sequences of actions. In text-only simulations, long-horizon planning has seen significant improvement brought by repositioning the role of LLMs. Instead of directly generating action sequences, LLMs translate the planning domain and problem into a formal planning language like the Planning Domain Definition Language (PDDL), which can call a formal solver to derive the plan in a verifiable manner. In multimodal environments, research on VLM-as-formalizer remains scarce, usually involving gross simplifications such as predefined object vocabulary or overly similar few-shot examples. In this work, we present a suite of five VLM-as-formalizer pipelines that tackle one-shot, open-vocabulary, and multimodal PDDL formalization. We evaluate those on an existing benchmark while presenting another two that for the first time account for planning with authentic, multi-view, and low-quality images. We conclude that VLM-as-formalizer greatly outperforms end-to-end plan generation. We reveal the bottleneck to be vision rather than language, as VLMs often fail to capture an exhaustive set of necessary object relations. While generating intermediate, textual representations such as captions or scene graphs partially compensate for the performance, their inconsistent gain leaves headroom for future research directions on multimodal planning formalization.

Zijian An, Lifeng Zhou

Game-theoretic resource allocation on graphs (GRAG) involves two players competing over multiple steps to control nodes of interest on a graph, a problem modeled as a multi-step Colonel Blotto Game (MCBG). Finding optimal strategies is challenging due to the dynamic action space and structural constraints imposed by the graph. To address this, we formulate the MCBG as a Markov Decision Process (MDP) and apply Reinforcement Learning (RL) methods, specifically Deep Q-Network (DQN) and Proximal Policy Optimization (PPO). To enforce graph constraints, we introduce an action-displacement adjacency matrix that dynamically generates valid action sets at each step. We evaluate RL performance across a variety of graph structures and initial resource distributions, comparing against random, greedy, and learned RL policies. Experimental results show that both DQN and PPO consistently outperform baseline strategies and converge to a balanced $50\%$ win rate when competing against the learned RL policy. Particularly, on asymmetric graphs, RL agents successfully exploit structural advantages and adapt their allocation strategies, even under disadvantageous initial resource distributions.

Amirhosein Chahe, Lifeng Zhou

Vision-language models (VLMs) show promise for autonomous driving but often lack transparent reasoning capabilities that are critical for safety. We investigate whether explicitly modeling reasoning during fine-tuning enhances VLM performance on driving decision tasks. Using GPT-4o, we generate structured reasoning chains for driving scenarios from the DriveLM benchmark with category-specific prompting strategies. We compare reasoning-based fine-tuning, answer-only fine-tuning, and baseline instruction-tuned models across multiple small VLM families (Llama 3.2, Llava 1.5, and Qwen 2.5VL). Our results demonstrate that reasoning-based fine-tuning consistently outperforms alternatives, with Llama3.2-11B-reason achieving the highest performance. Models fine-tuned with reasoning show substantial improvements in accuracy and text generation quality, suggesting explicit reasoning enhances internal representations for driving decisions. These findings highlight the importance of transparent decision processes in safety-critical domains and offer a promising direction for developing more interpretable autonomous driving systems.

Chaoqun Li, Huanqian Yan, Lifeng Zhou et al.

Adversarial attacks in the physical world pose a significant threat to the security of vision-based systems, such as facial recognition and autonomous driving. Existing adversarial patch methods primarily focus on improving attack performance, but they often produce patches that are easily detectable by humans and struggle to achieve environmental consistency, i.e., blending patches into the environment. This paper introduces a novel approach for generating adversarial patches, which addresses both the visual naturalness and environmental consistency of the patches. We propose Prompt-Guided Environmentally Consistent Adversarial Patch (PG-ECAP), a method that aligns the patch with the environment to ensure seamless integration into the environment. The approach leverages diffusion models to generate patches that are both environmental consistency and effective in evading detection. To further enhance the naturalness and consistency, we introduce two alignment losses: Prompt Alignment Loss and Latent Space Alignment Loss, ensuring that the generated patch maintains its adversarial properties while fitting naturally within its environment. Extensive experiments in both digital and physical domains demonstrate that PG-ECAP outperforms existing methods in attack success rate and environmental consistency.

Seungyeop Lee, Knut Peterson, Solmaz Arezoomandan et al.

A major obstacle to the development of effective monocular depth estimation algorithms is the difficulty in obtaining high-quality depth data that corresponds to collected RGB images. Collecting this data is time-consuming and costly, and even data collected by modern sensors has limited range or resolution, and is subject to inconsistencies and noise. To combat this, we propose a method of data generation in simulation using 3D synthetic environments and CycleGAN domain transfer. We compare this method of data generation to the popular NYUDepth V2 dataset by training a depth estimation model based on the DenseDepth structure using different training sets of real and simulated data. We evaluate the performance of the models on newly collected images and LiDAR depth data from a Husky robot to verify the generalizability of the approach and show that GAN-transformed data can serve as an effective alternative to real-world data, particularly in depth estimation.

Amirhosein Chahe, Chenan Wang, Abhishek Jeyapratap et al.

This paper introduces an attacking mechanism to challenge the resilience of autonomous driving systems. Specifically, we manipulate the decision-making processes of an autonomous vehicle by dynamically displaying adversarial patches on a screen mounted on another moving vehicle. These patches are optimized to deceive the object detection models into misclassifying targeted objects, e.g., traffic signs. Such manipulation has significant implications for critical multi-vehicle interactions such as intersection crossing and lane changing, which are vital for safe and efficient autonomous driving systems. Particularly, we make four major contributions. First, we introduce a novel adversarial attack approach where the patch is not co-located with its target, enabling more versatile and stealthy attacks. Moreover, our method utilizes dynamic patches displayed on a screen, allowing for adaptive changes and movement, enhancing the flexibility and performance of the attack. To do so, we design a Screen Image Transformation Network (SIT-Net), which simulates environmental effects on the displayed images, narrowing the gap between simulated and real-world scenarios. Further, we integrate a positional loss term into the adversarial training process to increase the success rate of the dynamic attack. Finally, we shift the focus from merely attacking perceptual systems to influencing the decision-making algorithms of self-driving systems. Our experiments demonstrate the first successful implementation of such dynamic adversarial attacks in real-world autonomous driving scenarios, paving the way for advancements in the field of robust and secure autonomous driving.

Lifeng Zhou, Vijay Kumar

The problem of multi-robot target tracking asks for actively planning the joint motion of robots to track targets. In this paper, we focus on such target tracking problems in adversarial environments, where attacks or failures may deactivate robots' sensors and communications. In contrast to the previous works that consider no attacks or sensing attacks only, we formalize the first robust multi-robot tracking framework that accounts for any fixed numbers of worst-case sensing and communication attacks. To secure against such attacks, we design the first robust planning algorithm, named Robust Active Target Tracking (RATT), which approximates the communication attacks to equivalent sensing attacks and then optimizes against the approximated and original sensing attacks. We show that RATT provides provable suboptimality bounds on the tracking quality for any non-decreasing objective function. Our analysis utilizes the notations of curvature for set functions introduced in combinatorial optimization. In addition, RATT runs in polynomial time and terminates with the same running time as state-of-the-art algorithms for (non-robust) target tracking. Finally, we evaluate RATT with both qualitative and quantitative simulations across various scenarios. In the evaluations, RATT exhibits a tracking quality that is near-optimal and superior to varying non-robust heuristics. We also demonstrate RATT's superiority and robustness against varying attack models (e.g., worst-case and bounded rational attacks).

Lifeng Zhou, Vishnu D. Sharma, Qingbiao Li et al.

The problem of decentralized multi-robot target tracking asks for jointly selecting actions, e.g., motion primitives, for the robots to maximize target tracking performance with local communications. One major challenge for practical implementations is to make target tracking approaches scalable for large-scale problem instances. In this work, we propose a general-purpose learning architecture toward collaborative target tracking at scale, with decentralized communications. Particularly, our learning architecture leverages a graph neural network (GNN) to capture local interactions of the robots and learns decentralized decision-making for the robots. We train the learning model by imitating an expert solution and implement the resulting model for decentralized action selection involving local observations and communications only. We demonstrate the performance of our GNN-based learning approach in a scenario of active target tracking with large networks of robots. The simulation results show our approach nearly matches the tracking performance of the expert algorithm, and yet runs several orders faster with up to 100 robots. Moreover, it slightly outperforms a decentralized greedy algorithm but runs faster (especially with more than 20 robots). The results also exhibit our approach's generalization capability in previously unseen scenarios, e.g., larger environments and larger networks of robots.

Jun Liu, Lifeng Zhou, Pratap Tokekar et al.

In this letter, we consider a distributed submodular maximization problem for multi-robot systems when attacked by adversaries. One of the major challenges for multi-robot systems is to increase resilience against failures or attacks. This is particularly important for distributed systems under attack as there is no central point of command that can detect, mitigate, and recover from attacks. Instead, a distributed multi-robot system must coordinate effectively to overcome adversarial attacks. In this work, our distributed submodular action selection problem models a broad set of scenarios where each robot in a multi-robot system has multiple action selections that may fulfill a global objective, such as exploration or target tracking. To increase resilience in this context, we propose a fully distributed algorithm to guide each robot's action selection when the system is attacked. The proposed algorithm guarantees performance in a worst-case scenario where up to a portion of the robots malfunction due to attacks. Importantly, the proposed algorithm is also consistent, as it is shown to converge to the same solution as a centralized method. Finally, a distributed resilient multi-robot exploration problem is presented to confirm the performance of the proposed algorithm.

Siddharth Mayya, Ragesh K. Ramachandran, Lifeng Zhou et al.

We consider a scenario where a team of robots with heterogeneous sensors must track a set of hostile targets which induce sensory failures on the robots. In particular, the likelihood of failures depends on the proximity between the targets and the robots. We propose a control framework that implicitly addresses the competing objectives of performance maximization and sensor preservation (which impacts the future performance of the team). Our framework consists of a predictive component -- which accounts for the risk of being detected by the target, and a reactive component -- which maximizes the performance of the team regardless of the failures that have already occurred. Based on a measure of the abundance of sensors in the team, our framework can generate aggressive and risk-averse robot configurations to track the targets. Crucially, the heterogeneous sensing capabilities of the robots are explicitly considered in each step, allowing for a more expressive risk-performance trade-off. Simulated experiments with induced sensor failures demonstrate the efficacy of the proposed approach.

Lifeng Zhou, Pratap Tokekar

Deploying a team of robots that can carefully coordinate their actions can make the entire system robust to individual failures. In this report, we review recent algorithmic development in making multi-robot systems robust to environmental uncertainties, failures, and adversarial attacks. We find the following three trends in the recent research in the area of multi-robot coordination: (1) resilient coordination to either withstand failures and/or attack or recover from failures/attacks; (2) risk-aware coordination to manage the trade-off risk and reward, where the risk stems due to environmental uncertainty; (3) Graph Neural Networks based coordination to learn decentralized multi-robot coordination policies. These algorithms have been applied to tasks such as formation control, task assignment and scheduling, search and planning, and informative data collection. In order for multi-robot systems to become practical, we need coordination algorithms that can scale to large teams of robots dealing with dynamically changing, failure-prone, contested, and uncertain environments. There has been significant recent research on multi-robot coordination that has contributed resilient and risk-aware algorithms to deal with these issues and reduce the gap between theory and practice. Learning-based approaches have been seen to be promising, especially since they can learn who, when, and how to communicate for effective coordination. However, these algorithms have also been shown to be vulnerable to adversarial attacks, and as such developing learning-based coordination strategies that are resilient to such attacks and robust to uncertainties is an important open area of research.

Guangyao Shi, Ishat E Rabban, Lifeng Zhou et al.

Submodular maximization has been widely used in many multi-robot task planning problems including information gathering, exploration, and target tracking. However, the interplay between submodular maximization and communication is rarely explored in the multi-robot setting. In many cases, maximizing the submodular objective may drive the robots in a way so as to disconnect the communication network. Driven by such observations, in this paper, we consider the problem of maximizing submodular function with connectivity constraints. Specifically, we propose a problem called Communication-aware Submodular Maximization (CSM), in which communication maintenance and submodular maximization are jointly considered in the decision-making process. One heuristic algorithm that consists of two stages, i.e. \textit{topology generation} and \textit{deviation minimization} is proposed. We validate the formulation and algorithm through numerical simulation. We find that our algorithm on average suffers only slightly performance decrease compared to the pure greedy strategy.

Rishab Balasubramanian, Lifeng Zhou, Pratap Tokekar et al.

We introduce a risk-aware multi-objective Traveling Salesperson Problem (TSP) variant, where the robot tour cost and tour reward have to be optimized simultaneously. The robot obtains reward along the edges in the graph. We study the case where the rewards and the costs exhibit diminishing marginal gains, i.e., are submodular. Unlike prior work, we focus on the scenario where the costs and the rewards are uncertain and seek to maximize the Conditional-Value-at-Risk (CVaR) metric of the submodular function. We propose a risk-aware greedy algorithm (RAGA) to find a bounded-approximation algorithm. The approximation algorithm runs in polynomial time and is within a constant factor of the optimal and an additive term that depends on the optimal solution. We use the submodular function's curvature to improve approximation results further and verify the algorithm's performance through simulations.

Guangyao Shi, Lifeng Zhou, Pratap Tokekar

The multiple-path orienteering problem asks for paths for a team of robots that maximize the total reward collected while satisfying budget constraints on the path length. This problem models many multi-robot routing tasks such as exploring unknown environments and information gathering for environmental monitoring. In this paper, we focus on how to make the robot team robust to failures when operating in adversarial environments. We introduce the Robust Multiple-path Orienteering Problem (RMOP) where we seek worst-case guarantees against an adversary that is capable of attacking at most $α$ robots. We consider two versions of this problem: RMOP offline and RMOP online. In the offline version, there is no communication or replanning when robots execute their plans and our main contribution is a general approximation scheme with a bounded approximation guarantee that depends on $α$ and the approximation factor for single robot orienteering. In particular, we show that the algorithm yields a (i) constant-factor approximation when the cost function is modular; (ii) $\log$ factor approximation when the cost function is submodular; and (iii) constant-factor approximation when the cost function is submodular but the robots are allowed to exceed their path budgets by a bounded amount. In the online version, RMOP is modeled as a two-player sequential game and solved adaptively in a receding horizon fashion based on Monte Carlo Tree Search (MCTS). In addition to theoretical analysis, we perform simulation studies for ocean monitoring and tunnel information-gathering applications to demonstrate the efficacy of our approach.

Vishnu D. Sharma, Maymoonah Toubeh, Lifeng Zhou et al.

We propose a risk-aware framework for multi-robot, multi-demand assignment and planning in unknown environments. Our motivation is disaster response and search-and-rescue scenarios where ground vehicles must reach demand locations as soon as possible. We consider a setting where the terrain information is available only in the form of an aerial, georeferenced image. Deep learning techniques can be used for semantic segmentation of the aerial image to create a cost map for safe ground robot navigation. Such segmentation may still be noisy. Hence, we present a joint planning and perception framework that accounts for the risk introduced due to noisy perception. Our contributions are two-fold: (i) we show how to use Bayesian deep learning techniques to extract risk at the perception level; and (ii) use a risk-theoretical measure, CVaR, for risk-aware planning and assignment. The pipeline is theoretically established, then empirically analyzed through two datasets. We find that accounting for risk at both levels produces quantifiably safer paths and assignments.

Lifeng Zhou, Pratap Tokekar

We study the problem of incorporating risk while making combinatorial decisions under uncertainty. We formulate a discrete submodular maximization problem for selecting a set using Conditional-Value-at-Risk (CVaR), a risk metric commonly used in financial analysis. While CVaR has recently been used in optimization of linear cost functions in robotics, we take the first step towards extending this to discrete submodular optimization and provide several positive results. Specifically, we propose the Sequential Greedy Algorithm that provides an approximation guarantee on finding the maxima of the CVaR cost function under a matroidal constraint. The approximation guarantee shows that the solution produced by our algorithm is within a constant factor of the optimal and an additive term that depends on the optimal. Our analysis uses the curvature of the submodular set function, and proves that the algorithm runs in polynomial time. This formulates a number of combinatorial optimization problems that appear in robotics. We use two such problems, vehicle assignment under uncertainty for mobility-on-demand and sensor selection with failures for environmental monitoring, as case studies to demonstrate the efficacy of our formulation. In particular, for the mobility-on-demand study, we propose an online triggering assignment algorithm that triggers a new assignment only can potentially lead to reducing the waiting time at demand locations. We verify the performance of the Sequential Greedy Algorithm and the online triggering assignment algorithm through simulations.

Lifeng Zhou, Vasileios Tzoumas, George J. Pappas et al.

In this paper, we design algorithms to protect swarm-robotics applications against sensor denial-of-service (DoS) attacks on robots. We focus on applications requiring the robots to jointly select actions, e.g., which trajectory to follow, among a set of available ones. Such applications are central in large-scale robotic applications, such as multi-robot motion planning for target tracking. But the current attack-robust algorithms are centralized. In this paper, we propose a general-purpose distributed algorithm towards robust optimization at scale, with local communications only. We name it Distributed Robust Maximization (DRM). DRM proposes a divide-and-conquer approach that distributively partitions the problem among cliques of robots. Then, the cliques optimize in parallel, independently of each other. We prove DRM achieves a close-to-optimal performance. We demonstrate DRM's performance in both Gazebo and MATLAB simulations, in scenarios of active target tracking with swarms of robots. In the simulations, DRM achieves computational speed-ups, being 1-2 orders faster than the centralized algorithms; yet, it nearly matches the tracking performance of the centralized counterparts. Since, DRM overestimates the number of attacks in each clique, in this paper we also introduce an Improved Distributed Robust Maximization (IDRM) algorithm. IDRM infers the number of attacks in each clique less conservatively than DRM by leveraging 3-hop neighboring communications. We verify IDRM improves DRM's performance in simulations.

Lifeng Zhou, Vasileios Tzoumas, George J. Pappas et al.

The problem of target tracking with multiple robots consists of actively planning the motion of the robots to track the targets. A major challenge for practical deployments is to make the robots resilient to failures. In particular, robots may be attacked in adversarial scenarios, or their sensors may fail or get occluded. In this paper, we introduce planning algorithms for multi-target tracking that are resilient to such failures. In general, resilient target tracking is computationally hard. Contrary to the case where there are no failures, no scalable approximation algorithms are known for resilient target tracking when the targets are indistinguishable, or unknown in number, or with unknown motion model. In this paper we provide the first such algorithm, that also has the following properties: First, it achieves maximal resiliency, since the algorithm is valid for any number of failures. Second, it is scalable, as our algorithm terminates with the same running time as state-of-the-art algorithms for (non-resilient) target tracking. Third, it provides provable approximation bounds on the tracking performance, since our algorithm guarantees a solution that is guaranteed to be close to the optimal. We quantify our algorithm's approximation performance using a novel notion of curvature for monotone set functions subject to matroid constraints. Finally, we demonstrate the efficacy of our algorithm through MATLAB and Gazebo simulations, and a sensitivity analysis; we focus on scenarios that involve a known number of distinguishable targets.

Zhongshun Zhang, Yoonchang Sung, Lifeng Zhou et al.

We introduce and study the problem of planning a trajectory for an agent to carry out a scouting mission while avoiding being detected by an adversarial guard. This introduces an adversarial version of classical visibility-based planning problems such as the Watchman Route Problem. The agent receives a positive reward for increasing its visibility and a negative penalty when it is detected by the guard. The objective is to find a finite-horizon path for the agent that balances the trade-off maximizing visibility and minimizing detectability. We model this problem as a sequential two-player zero-sum discrete game. A minimax tree search can give the optimal policy for the agent but requires an exponential-time computation and space. We propose several pruning techniques to reduce the computational cost while still preserving optimality guarantees. Simulation results show that the proposed strategy prunes approximately three orders of magnitude nodes as compared to the brute-force strategy.

Lifeng Zhou, Pratap Tokekar

We study the problem of incorporating risk while making combinatorial decisions under uncertainty. We formulate a discrete submodular maximization problem for selecting a set using Conditional-Value-at-Risk (CVaR), a risk metric commonly used in financial analysis. While CVaR has recently been used in optimization of linear cost functions in robotics, we take the first stages towards extending this to discrete submodular optimization and provide several positive results. Specifically, we propose the Sequential Greedy Algorithm that provides an approximation guarantee on finding the maxima of the CVaR cost function under a matroidal constraint. The approximation guarantee shows that the solution produced by our algorithm is within a constant factor of the optimal and an additive term that depends on the optimal. Our analysis uses the curvature of the submodular set function, and proves that the algorithm runs in polynomial time. This formulates a number of combinatorial optimization problems that appear in robotics. We use two such problems, vehicle assignment under uncertainty for mobility-on-demand and sensor selection with failures for environmental monitoring, as case studies to demonstrate the efficacy of our formulation.

Zhongshun Zhang, Lifeng Zhou, Pratap Tokekar

We study the problem of designing false measurement data that is injected to corrupt and mislead the output of a Kalman filter. Unlike existing works that focus on detection and filtering algorithms for the observer, we study the problem from the attacker's point-of-view. In our model, the attacker can corrupt the measurements by injecting additive spoofing signals. The attacker seeks to create a separation between the estimate of the Kalman filter with and without spoofed signals. We present a number of results on how to inject spoofing signals while minimizing the magnitude of the injected signals. The resulting strategies are evaluated through simulations along with theoretical proofs. We also evaluate the spoofing strategy in the presence of a $χ^2$ spoof detector. The results show that the proposed strategy can successfully mislead a Kalman filter while ensuring it is not detected.

Lifeng Zhou, Pratap Tokekar

We study two sensor assignment problems for multi-target tracking with the goal of improving the observability of the underlying estimator. We consider various measures of the observability matrix as the assignment value function. We first study the general version where the sensors must form teams to track individual targets. If the value function is monotonically increasing and submodular then a greedy algorithm yields a 1/2-approximation. We then study a restricted version where exactly two sensors must be assigned to each target. We present a 1/3-approximation algorithm for this problem which holds for arbitrary value functions (not necessarily submodular or monotone). In addition to approximation algorithms, we also present various properties of observability measures. We show that the inverse of the condition number of the observability matrix is neither monotone nor submodular, but present other measures which are. Specifically, we show that the trace and rank of the symmetric observability matrix are monotone and submodular and the log determinant of the symmetric observability matrix is monotone and submodular when the matrix is non-singular. If the target's motion model is not known, the inverse cannot be computed exactly. Instead, we present a lower bound for distance sensors. In addition to theoretical results, we evaluate our results empirically through simulations.

Lifeng Zhou, Pratap Tokekar

We study two sensor assignment problems for multi-target tracking with the goal of improving the observability of the underlying estimator. In the restricted version of the problem, we focus on assigning unique pairs of sensors to each target. We present a 1/3-approximation algorithm for this problem. We use the inverse of the condition number as the value function. If the target's motion model is not known, the inverse cannot be computed exactly. Instead, we present a lower bound for range-only sensing. In the general version, the sensors must form teams to track individual targets. We do not force any specific constraints on the size of each team, instead assume that the value function is monotonically increasing and is submodular. A greedy algorithm that yields a 1/2-approximation. However, we show that the inverse of the condition number is neither monotone nor submodular. Instead, we present other measures that are monotone and submodular. In addition to theoretical results, we evaluate our results empirically through simulations.

Lifeng Zhou, Pratap Tokekar

We study the problem of reducing the amount of communication in decentralized target tracking. We focus on the scenario where a team of robots are allowed to move on the boundary of the environment. Their goal is to seek a formation so as to best track a target moving in the interior of the environment. The robots are capable of measuring distances to the target. Decentralized control strategies have been proposed in the past that guarantee that the robots asymptotically converge to the optimal formation. However, existing methods require that the robots exchange information with their neighbors at all time steps. Instead, we focus on decentralized strategies to reduce the amount of communication among robots. We propose a self-triggered communication strategy that decides when a particular robot should seek up-to-date information from its neighbors and when it is safe to operate with possibly outdated information. We prove that this strategy converges asymptotically to a desired formation when the target is stationary. For the case of a mobile target, we propose an extension whereby each robot decides its optimal partner to share its measurements with using observability as a criterion. We evaluate all the approaches (constant communication and self-triggered communication with centralized and decentralized sensor fusion) through simulations.