Zhenghao Peng

Yan Wang, Wenjie Luo, Junjie Bai et al. · nvidia

End-to-end architectures trained via imitation learning have advanced autonomous driving by scaling model size and data, yet performance remains brittle in safety-critical long-tail scenarios where supervision is sparse and causal understanding is limited. To address this, we introduce Alpamayo-R1 (AR1), a vision-language-action model (VLA) that integrates Chain of Causation reasoning with trajectory planning to enhance decision-making in complex driving scenarios. Our approach features three key innovations: (1) the Chain of Causation (CoC) dataset, built through a hybrid auto-labeling and human-in-the-loop pipeline producing decision-grounded, causally linked reasoning traces aligned with driving behaviors; (2) a modular VLA architecture combining Cosmos-Reason, a Vision-Language Model pre-trained for Physical AI applications, with a diffusion-based trajectory decoder that generates dynamically feasible plans in real time; (3) a multi-stage training strategy using supervised fine-tuning to elicit reasoning and reinforcement learning (RL) to optimize reasoning quality via large reasoning model feedback and enforce reasoning-action consistency. Evaluation shows AR1 achieves up to a 12% improvement in planning accuracy on challenging cases compared to a trajectory-only baseline, with a 35% reduction in off-road rate and 25% reduction in close encounter rate in closed-loop simulation. RL post-training improves reasoning quality by 45% as measured by a large reasoning model critic and reasoning-action consistency by 37%. Model scaling from 0.5B to 7B parameters shows consistent improvements. On-vehicle road tests confirm real-time performance (99 ms latency) and successful urban deployment. By bridging interpretable reasoning with precise control, AR1 demonstrates a practical path towards Level 4 autonomous driving. We plan to release AR1 models and a subset of the CoC in a future update.

Mingxin Huang, Yuliang Liu, Zhenghao Peng et al.

End-to-end scene text spotting has attracted great attention in recent years due to the success of excavating the intrinsic synergy of the scene text detection and recognition. However, recent state-of-the-art methods usually incorporate detection and recognition simply by sharing the backbone, which does not directly take advantage of the feature interaction between the two tasks. In this paper, we propose a new end-to-end scene text spotting framework termed SwinTextSpotter. Using a transformer encoder with dynamic head as the detector, we unify the two tasks with a novel Recognition Conversion mechanism to explicitly guide text localization through recognition loss. The straightforward design results in a concise framework that requires neither additional rectification module nor character-level annotation for the arbitrarily-shaped text. Qualitative and quantitative experiments on multi-oriented datasets RoIC13 and ICDAR 2015, arbitrarily-shaped datasets Total-Text and CTW1500, and multi-lingual datasets ReCTS (Chinese) and VinText (Vietnamese) demonstrate SwinTextSpotter significantly outperforms existing methods. Code is available at https://github.com/mxin262/SwinTextSpotter.

Zhenghai Xue, Zhenghao Peng, Quanyi Li et al.

The Teacher-Student Framework (TSF) is a reinforcement learning setting where a teacher agent guards the training of a student agent by intervening and providing online demonstrations. Assuming optimal, the teacher policy has the perfect timing and capability to intervene in the learning process of the student agent, providing safety guarantee and exploration guidance. Nevertheless, in many real-world settings it is expensive or even impossible to obtain a well-performing teacher policy. In this work, we relax the assumption of a well-performing teacher and develop a new method that can incorporate arbitrary teacher policies with modest or inferior performance. We instantiate an Off-Policy Reinforcement Learning algorithm, termed Teacher-Student Shared Control (TS2C), which incorporates teacher intervention based on trajectory-based value estimation. Theoretical analysis validates that the proposed TS2C algorithm attains efficient exploration and substantial safety guarantee without being affected by the teacher's own performance. Experiments on various continuous control tasks show that our method can exploit teacher policies at different performance levels while maintaining a low training cost. Moreover, the student policy surpasses the imperfect teacher policy in terms of higher accumulated reward in held-out testing environments. Code is available at https://metadriverse.github.io/TS2C.

Qihang Zhang, Zhenghao Peng, Bolei Zhou

Deep visuomotor policy learning, which aims to map raw visual observation to action, achieves promising results in control tasks such as robotic manipulation and autonomous driving. However, it requires a huge number of online interactions with the training environment, which limits its real-world application. Compared to the popular unsupervised feature learning for visual recognition, feature pretraining for visuomotor control tasks is much less explored. In this work, we aim to pretrain policy representations for driving tasks by watching hours-long uncurated YouTube videos. Specifically, we train an inverse dynamic model with a small amount of labeled data and use it to predict action labels for all the YouTube video frames. A new contrastive policy pretraining method is then developed to learn action-conditioned features from the video frames with pseudo action labels. Experiments show that the resulting action-conditioned features obtain substantial improvements for the downstream reinforcement learning and imitation learning tasks, outperforming the weights pretrained from previous unsupervised learning methods and ImageNet pretrained weight. Code, model weights, and data are available at: https://metadriverse.github.io/ACO.

Quanyi Li, Zhenghao Peng, Haibin Wu et al.

Human-AI shared control allows human to interact and collaborate with AI to accomplish control tasks in complex environments. Previous Reinforcement Learning (RL) methods attempt the goal-conditioned design to achieve human-controllable policies at the cost of redesigning the reward function and training paradigm. Inspired by the neuroscience approach to investigate the motor cortex in primates, we develop a simple yet effective frequency-based approach called \textit{Policy Dissection} to align the intermediate representation of the learned neural controller with the kinematic attributes of the agent behavior. Without modifying the neural controller or retraining the model, the proposed approach can convert a given RL-trained policy into a human-interactive policy. We evaluate the proposed approach on the RL tasks of autonomous driving and locomotion. The experiments show that human-AI shared control achieved by Policy Dissection in driving task can substantially improve the performance and safety in unseen traffic scenes. With human in the loop, the locomotion robots also exhibit versatile controllable motion skills even though they are only trained to move forward. Our results suggest the promising direction of implementing human-AI shared autonomy through interpreting the learned representation of the autonomous agents. Demo video and code will be made available at https://metadriverse.github.io/policydissect.

Linrui Zhang, Zhenghao Peng, Quanyi Li et al.

Driving safety is a top priority for autonomous vehicles. Orthogonal to prior work handling accident-prone traffic events by algorithm designs at the policy level, we investigate a Closed-loop Adversarial Training (CAT) framework for safe end-to-end driving in this paper through the lens of environment augmentation. CAT aims to continuously improve the safety of driving agents by training the agent on safety-critical scenarios that are dynamically generated over time. A novel resampling technique is developed to turn log-replay real-world driving scenarios into safety-critical ones via probabilistic factorization, where the adversarial traffic generation is modeled as the multiplication of standard motion prediction sub-problems. Consequently, CAT can launch more efficient physical attacks compared to existing safety-critical scenario generation methods and yields a significantly less computational cost in the iterative learning pipeline. We incorporate CAT into the MetaDrive simulator and validate our approach on hundreds of driving scenarios imported from real-world driving datasets. Experimental results demonstrate that CAT can effectively generate adversarial scenarios countering the agent being trained. After training, the agent can achieve superior driving safety in both log-replay and safety-critical traffic scenarios on the held-out test set. Code and data are available at https://metadriverse.github.io/cat.

Zhenghao Peng, Wenjie Luo, Yiren Lu et al.

A major challenge in autonomous vehicle research is modeling agent behaviors, which has critical applications including constructing realistic and reliable simulations for off-board evaluation and forecasting traffic agents motion for onboard planning. While supervised learning has shown success in modeling agents across various domains, these models can suffer from distribution shift when deployed at test-time. In this work, we improve the reliability of agent behaviors by closed-loop fine-tuning of behavior models with reinforcement learning. Our method demonstrates improved overall performance, as well as improved targeted metrics such as collision rate, on the Waymo Open Sim Agents challenge. Additionally, we present a novel policy evaluation benchmark to directly assess the ability of simulated agents to measure the quality of autonomous vehicle planners and demonstrate the effectiveness of our approach on this new benchmark.

Brandon J. McMahan, Zhenghao Peng, Bolei Zhou et al.

The rapid development of artificial intelligence (AI) has unearthed the potential to assist humans in controlling advanced technologies. Shared autonomy (SA) facilitates control by combining inputs from a human pilot and an AI copilot. In prior SA studies, the copilot is constantly active in determining the action played at each time step. This limits human autonomy and may have deleterious effects on performance. In general, the amount of helpful copilot assistance can vary greatly depending on the task dynamics. We therefore hypothesize that human autonomy and SA performance improve through dynamic and selective copilot intervention. To address this, we develop a goal-agnostic intervention assistance (IA) that dynamically shares control by having the copilot intervene only when the expected value of the copilot's action exceeds that of the human's action across all possible goals. We implement IA with a diffusion copilot (termed IDA) trained on expert demonstrations with goal masking. We prove a lower bound on the performance of IA that depends on pilot and copilot performance. Experiments with simulated human pilots show that IDA achieves higher performance than pilot-only and traditional SA control in variants of the Reacher environment and Lunar Lander. We then demonstrate that IDA achieves better control in Lunar Lander with human-in-the-loop experiments. Human participants report greater autonomy with IDA and prefer IDA over pilot-only and traditional SA control. We attribute the success of IDA to preserving human autonomy while simultaneously offering assistance to prevent the human pilot from entering universally bad states.

Mingxin Huang, Dezhi Peng, Hongliang Li et al.

End-to-end scene text spotting, which aims to read the text in natural images, has garnered significant attention in recent years. However, recent state-of-the-art methods usually incorporate detection and recognition simply by sharing the backbone, which does not directly take advantage of the feature interaction between the two tasks. In this paper, we propose a new end-to-end scene text spotting framework termed SwinTextSpotter v2, which seeks to find a better synergy between text detection and recognition. Specifically, we enhance the relationship between two tasks using novel Recognition Conversion and Recognition Alignment modules. Recognition Conversion explicitly guides text localization through recognition loss, while Recognition Alignment dynamically extracts text features for recognition through the detection predictions. This simple yet effective design results in a concise framework that requires neither an additional rectification module nor character-level annotations for the arbitrarily-shaped text. Furthermore, the parameters of the detector are greatly reduced without performance degradation by introducing a Box Selection Schedule. Qualitative and quantitative experiments demonstrate that SwinTextSpotter v2 achieved state-of-the-art performance on various multilingual (English, Chinese, and Vietnamese) benchmarks. The code will be available at \href{https://github.com/mxin262/SwinTextSpotterv2}{SwinTextSpotter v2}.

Andy Wang, Xu Yan, Brandon McMahan et al.

Shared autonomy combines human user and AI copilot actions to control complex systems such as robotic arms. When a task is challenging, requires high dimensional control, or is subject to corruption, shared autonomy can significantly increase task performance by using a trained copilot to effectively correct user actions in a manner consistent with the user's goals. To significantly improve the performance of shared autonomy, we introduce Diffusion Sequence Copilots (DiSCo): a method of shared autonomy with diffusion policy that plans action sequences consistent with past user actions. DiSCo seeds and inpaints the diffusion process with user-provided actions with hyperparameters to balance conformity to expert actions, alignment with user intent, and perceived responsiveness. We demonstrate that DiSCo substantially improves task performance in simulated driving and robotic arm tasks. Project website: https://sites.google.com/view/disco-shared-autonomy/

Haoyuan Cai, Zhenghao Peng, Bolei Zhou

Interactive Imitation Learning (IIL) allows agents to acquire desired behaviors through human interventions, but current methods impose high cognitive demands on human supervisors. We propose the Adaptive Intervention Mechanism (AIM), a novel robot-gated IIL algorithm that learns an adaptive criterion for requesting human demonstrations. AIM utilizes a proxy Q-function to mimic the human intervention rule and adjusts intervention requests based on the alignment between agent and human actions. By assigning high Q-values when the agent deviates from the expert and decreasing these values as the agent becomes proficient, the proxy Q-function enables the agent to assess the real-time alignment with the expert and request assistance when needed. Our expert-in-the-loop experiments reveal that AIM significantly reduces expert monitoring efforts in both continuous and discrete control tasks. Compared to the uncertainty-based baseline Thrifty-DAgger, our method achieves a 40% improvement in terms of human take-over cost and learning efficiency. Furthermore, AIM effectively identifies safety-critical states for expert assistance, thereby collecting higher-quality expert demonstrations and reducing overall expert data and environment interactions needed. Code and demo video are available at https://github.com/metadriverse/AIM.

Zhenghao Peng, Wenjie Mo, Chenda Duan et al.

Learning from active human involvement enables the human subject to actively intervene and demonstrate to the AI agent during training. The interaction and corrective feedback from human brings safety and AI alignment to the learning process. In this work, we propose a new reward-free active human involvement method called Proxy Value Propagation for policy optimization. Our key insight is that a proxy value function can be designed to express human intents, wherein state-action pairs in the human demonstration are labeled with high values, while those agents' actions that are intervened receive low values. Through the TD-learning framework, labeled values of demonstrated state-action pairs are further propagated to other unlabeled data generated from agents' exploration. The proxy value function thus induces a policy that faithfully emulates human behaviors. Human-in-the-loop experiments show the generality and efficiency of our method. With minimal modification to existing reinforcement learning algorithms, our method can learn to solve continuous and discrete control tasks with various human control devices, including the challenging task of driving in Grand Theft Auto V. Demo video and code are available at: https://metadriverse.github.io/pvp

Ziyang Xie, Zhizheng Liu, Zhenghao Peng et al.

Sim-to-real gap has long posed a significant challenge for robot learning in simulation, preventing the deployment of learned models in the real world. Previous work has primarily focused on domain randomization and system identification to mitigate this gap. However, these methods are often limited by the inherent constraints of the simulation and graphics engines. In this work, we propose Vid2Sim, a novel framework that effectively bridges the sim2real gap through a scalable and cost-efficient real2sim pipeline for neural 3D scene reconstruction and simulation. Given a monocular video as input, Vid2Sim can generate photorealistic and physically interactable 3D simulation environments to enable the reinforcement learning of visual navigation agents in complex urban environments. Extensive experiments demonstrate that Vid2Sim significantly improves the performance of urban navigation in the digital twins and real world by 31.2% and 68.3% in success rate compared with agents trained with prior simulation methods.

Weizhen Wang, Chenda Duan, Zhenghao Peng et al.

Vision Language Models (VLMs) demonstrate significant potential as embodied AI agents for various mobility applications. However, a standardized, closed-loop benchmark for evaluating their spatial reasoning and sequential decision-making capabilities is lacking. To address this, we present MetaVQA: a comprehensive benchmark designed to assess and enhance VLMs' understanding of spatial relationships and scene dynamics through Visual Question Answering (VQA) and closed-loop simulations. MetaVQA leverages Set-of-Mark prompting and top-down view ground-truth annotations from nuScenes and Waymo datasets to automatically generate extensive question-answer pairs based on diverse real-world traffic scenarios, ensuring object-centric and context-rich instructions. Our experiments show that fine-tuning VLMs with the MetaVQA dataset significantly improves their spatial reasoning and embodied scene comprehension in safety-critical simulations, evident not only in improved VQA accuracies but also in emerging safety-aware driving maneuvers. In addition, the learning demonstrates strong transferability from simulation to real-world observation. Code and data will be publicly available at https://metadriverse.github.io/metavqa .

Zhenghao Peng, Yuxin Liu, Bolei Zhou

Realistic and interactive traffic simulation is essential for training and evaluating autonomous driving systems. However, most existing data-driven simulation methods rely on static initialization or log-replay data, limiting their ability to model dynamic, long-horizon scenarios with evolving agent populations. We propose InfGen, a scenario generation framework that outputs agent states and trajectories in an autoregressive manner. InfGen represents the entire scene as a sequence of tokens, including traffic light signals, agent states, and motion vectors, and uses a transformer model to simulate traffic over time. This design enables InfGen to continuously insert new agents into traffic, supporting infinite scene generation. Experiments demonstrate that InfGen produces realistic, diverse, and adaptive traffic behaviors. Furthermore, reinforcement learning policies trained in InfGen-generated scenarios achieve superior robustness and generalization, validating its utility as a high-fidelity simulation environment for autonomous driving. More information is available at https://metadriverse.github.io/infgen/.

Zhenghao Peng, Zhizheng Liu, Bolei Zhou

Mobile robots are essential in applications such as autonomous delivery and hospitality services. Applying learning-based methods to address mobile robot tasks has gained popularity due to its robustness and generalizability. Traditional methods such as Imitation Learning (IL) and Reinforcement Learning (RL) offer adaptability but require large datasets, carefully crafted reward functions, and face sim-to-real gaps, making them challenging for efficient and safe real-world deployment. We propose an online human-in-the-loop learning method PVP4Real that combines IL and RL to address these issues. PVP4Real enables efficient real-time policy learning from online human intervention and demonstration, without reward or any pretraining, significantly improving data efficiency and training safety. We validate our method by training two different robots -- a legged quadruped, and a wheeled delivery robot -- in two mobile robot tasks, one of which even uses raw RGBD image as observation. The training finishes within 15 minutes. Our experiments show the promising future of human-in-the-loop learning in addressing the data efficiency issue in real-world robotic tasks. More information is available at: https://metadriverse.github.io/pvp4real/

Haoyuan Cai, Zhenghao Peng, Bolei Zhou

Learning from human involvement aims to incorporate the human subject to monitor and correct agent behavior errors. Although most interactive imitation learning methods focus on correcting the agent's action at the current state, they do not adjust its actions in future states, which may be potentially more hazardous. To address this, we introduce Predictive Preference Learning from Human Interventions (PPL), which leverages the implicit preference signals contained in human interventions to inform predictions of future rollouts. The key idea of PPL is to bootstrap each human intervention into L future time steps, called the preference horizon, with the assumption that the agent follows the same action and the human makes the same intervention in the preference horizon. By applying preference optimization on these future states, expert corrections are propagated into the safety-critical regions where the agent is expected to explore, significantly improving learning efficiency and reducing human demonstrations needed. We evaluate our approach with experiments on both autonomous driving and robotic manipulation benchmarks and demonstrate its efficiency and generality. Our theoretical analysis further shows that selecting an appropriate preference horizon L balances coverage of risky states with label correctness, thereby bounding the algorithmic optimality gap. Demo and code are available at: https://metadriverse.github.io/ppl

Yuxin Liu, Zhenghao Peng, Xuanhao Cui et al.

Scenario-based testing is essential for validating the performance of autonomous driving (AD) systems. However, such testing is limited by the scarcity of long-tailed, safety-critical scenarios in existing datasets collected in the real world. To tackle the data issue, we propose the Adv-BMT framework, which augments real-world scenarios with diverse and realistic adversarial traffic interactions. The core component of Adv-BMT is a bidirectional motion transformer (BMT) model to perform inverse traffic motion predictions, which takes agent information in the last time step of the scenario as input, and reconstructs the traffic in the inverse of chronological order until the initial time step. The Adv-BMT framework is a two-staged pipeline: it first conducts adversarial initializations and then inverse motion predictions. Different from previous work, we do not need any collision data for pretraining, and are able to generate realistic and diverse collision interactions. Our experimental results validate the quality of generated collision scenarios by Adv-BMT: training in our augmented dataset would reduce episode collision rates by 20%. Demo and code are available at: https://metadriverse.github.io/adv-bmt/.

Yunsong Zhou, Michael Simon, Zhenghao Peng et al.

Controllable synthetic data generation can substantially lower the annotation cost of training data. Prior works use diffusion models to generate driving images conditioned on the 3D object layout. However, those models are trained on small-scale datasets like nuScenes, which lack appearance and layout diversity. Moreover, overfitting often happens, where the trained models can only generate images based on the layout data from the validation set of the same dataset. In this work, we introduce a simulator-conditioned scene generation framework called SimGen that can learn to generate diverse driving scenes by mixing data from the simulator and the real world. It uses a novel cascade diffusion pipeline to address challenging sim-to-real gaps and multi-condition conflicts. A driving video dataset DIVA is collected to enhance the generative diversity of SimGen, which contains over 147.5 hours of real-world driving videos from 73 locations worldwide and simulated driving data from the MetaDrive simulator. SimGen achieves superior generation quality and diversity while preserving controllability based on the text prompt and the layout pulled from a simulator. We further demonstrate the improvements brought by SimGen for synthetic data augmentation on the BEV detection and segmentation task and showcase its capability in safety-critical data generation.

Quanyi Li, Zhenghao Peng, Bolei Zhou

Human intervention is an effective way to inject human knowledge into the training loop of reinforcement learning, which can bring fast learning and ensured training safety. Given the very limited budget of human intervention, it remains challenging to design when and how human expert interacts with the learning agent in the training. In this work, we develop a novel human-in-the-loop learning method called Human-AI Copilot Optimization (HACO).To allow the agent's sufficient exploration in the risky environments while ensuring the training safety, the human expert can take over the control and demonstrate how to avoid probably dangerous situations or trivial behaviors. The proposed HACO then effectively utilizes the data both from the trial-and-error exploration and human's partial demonstration to train a high-performing agent. HACO extracts proxy state-action values from partial human demonstration and optimizes the agent to improve the proxy values meanwhile reduce the human interventions. The experiments show that HACO achieves a substantially high sample efficiency in the safe driving benchmark. HACO can train agents to drive in unseen traffic scenarios with a handful of human intervention budget and achieve high safety and generalizability, outperforming both reinforcement learning and imitation learning baselines with a large margin. Code and demo videos are available at: https://decisionforce.github.io/HACO/.

Zhenghao Peng, Quanyi Li, Ka Ming Hui et al.

Self-Driven Particles (SDP) describe a category of multi-agent systems common in everyday life, such as flocking birds and traffic flows. In a SDP system, each agent pursues its own goal and constantly changes its cooperative or competitive behaviors with its nearby agents. Manually designing the controllers for such SDP system is time-consuming, while the resulting emergent behaviors are often not realistic nor generalizable. Thus the realistic simulation of SDP systems remains challenging. Reinforcement learning provides an appealing alternative for automating the development of the controller for SDP. However, previous multi-agent reinforcement learning (MARL) methods define the agents to be teammates or enemies before hand, which fail to capture the essence of SDP where the role of each agent varies to be cooperative or competitive even within one episode. To simulate SDP with MARL, a key challenge is to coordinate agents' behaviors while still maximizing individual objectives. Taking traffic simulation as the testing bed, in this work we develop a novel MARL method called Coordinated Policy Optimization (CoPO), which incorporates social psychology principle to learn neural controller for SDP. Experiments show that the proposed method can achieve superior performance compared to MARL baselines in various metrics. Noticeably the trained vehicles exhibit complex and diverse social behaviors that improve performance and safety of the population as a whole. Demo video and source code are available at: https://decisionforce.github.io/CoPO/

Zhenghao Peng, Quanyi Li, Chunxiao Liu et al.

When learning common skills like driving, beginners usually have domain experts standing by to ensure the safety of the learning process. We formulate such learning scheme under the Expert-in-the-loop Reinforcement Learning where a guardian is introduced to safeguard the exploration of the learning agent. While allowing the sufficient exploration in the uncertain environment, the guardian intervenes under dangerous situations and demonstrates the correct actions to avoid potential accidents. Thus ERL enables both exploration and expert's partial demonstration as two training sources. Following such a setting, we develop a novel Expert Guided Policy Optimization (EGPO) method which integrates the guardian in the loop of reinforcement learning. The guardian is composed of an expert policy to generate demonstration and a switch function to decide when to intervene. Particularly, a constrained optimization technique is used to tackle the trivial solution that the agent deliberately behaves dangerously to deceive the expert into taking over. Offline RL technique is further used to learn from the partial demonstration generated by the expert. Safe driving experiments show that our method achieves superior training and test-time safety, outperforms baselines with a substantial margin in sample efficiency, and preserves the generalizabiliy to unseen environments in test-time. Demo video and source code are available at: https://decisionforce.github.io/EGPO/

Quanyi Li, Zhenghao Peng, Lan Feng et al.

Driving safely requires multiple capabilities from human and intelligent agents, such as the generalizability to unseen environments, the safety awareness of the surrounding traffic, and the decision-making in complex multi-agent settings. Despite the great success of Reinforcement Learning (RL), most of the RL research works investigate each capability separately due to the lack of integrated environments. In this work, we develop a new driving simulation platform called MetaDrive to support the research of generalizable reinforcement learning algorithms for machine autonomy. MetaDrive is highly compositional, which can generate an infinite number of diverse driving scenarios from both the procedural generation and the real data importing. Based on MetaDrive, we construct a variety of RL tasks and baselines in both single-agent and multi-agent settings, including benchmarking generalizability across unseen scenes, safe exploration, and learning multi-agent traffic. The generalization experiments conducted on both procedurally generated scenarios and real-world scenarios show that increasing the diversity and the size of the training set leads to the improvement of the RL agent's generalizability. We further evaluate various safe reinforcement learning and multi-agent reinforcement learning algorithms in MetaDrive environments and provide the benchmarks. Source code, documentation, and demo video are available at \url{ https://metadriverse.github.io/metadrive}.

Hao Sun, Ziping Xu, Meng Fang et al.

Safe exploration is crucial for the real-world application of reinforcement learning (RL). Previous works consider the safe exploration problem as Constrained Markov Decision Process (CMDP), where the policies are being optimized under constraints. However, when encountering any potential dangers, human tends to stop immediately and rarely learns to behave safely in danger. Motivated by human learning, we introduce a new approach to address safe RL problems under the framework of Early Terminated MDP (ET-MDP). We first define the ET-MDP as an unconstrained MDP with the same optimal value function as its corresponding CMDP. An off-policy algorithm based on context models is then proposed to solve the ET-MDP, which thereby solves the corresponding CMDP with better asymptotic performance and improved learning efficiency. Experiments on various CMDP tasks show a substantial improvement over previous methods that directly solve CMDP.

Quanyi Li, Zhenghao Peng, Qihang Zhang et al.

Over the past few years there is a growing interest in the learning-based self driving system. To ensure safety, such systems are first developed and validated in simulators before being deployed in the real world. However, most of the existing driving simulators only contain a fixed set of scenes and a limited number of configurable settings. That might easily cause the overfitting issue for the learning-based driving systems as well as the lack of their generalization ability to unseen scenarios. To better evaluate and improve the generalization of end-to-end driving, we introduce an open-ended and highly configurable driving simulator called PGDrive, following a key feature of procedural generation. Diverse road networks are first generated by the proposed generation algorithm via sampling from elementary road blocks. Then they are turned into interactive training environments where traffic flows of nearby vehicles with realistic kinematics are rendered. We validate that training with the increasing number of procedurally generated scenes significantly improves the generalization of the agent across scenarios of different traffic densities and road networks. Many applications such as multi-agent traffic simulation and safe driving benchmark can be further built upon the simulator. To facilitate the joint research effort of end-to-end driving, we release the simulator and pretrained models at https://decisionforce.github.io/pgdrive

Zhenghao Peng, Hao Sun, Bolei Zhou

Conventional Reinforcement Learning (RL) algorithms usually have one single agent learning to solve the task independently. As a result, the agent can only explore a limited part of the state-action space while the learned behavior is highly correlated to the agent's previous experience, making the training prone to a local minimum. In this work, we empower RL with the capability of teamwork and propose a novel non-local policy optimization framework called Diversity-regularized Collaborative Exploration (DiCE). DiCE utilizes a group of heterogeneous agents to explore the environment simultaneously and share the collected experiences. A regularization mechanism is further designed to maintain the diversity of the team and modulate the exploration. We implement the framework in both on-policy and off-policy settings and the experimental results show that DiCE can achieve substantial improvement over the baselines in the MuJoCo locomotion tasks.

Hao Sun, Zhenghao Peng, Bo Dai et al.

In problem-solving, we humans can come up with multiple novel solutions to the same problem. However, reinforcement learning algorithms can only produce a set of monotonous policies that maximize the cumulative reward but lack diversity and novelty. In this work, we address the problem of generating novel policies in reinforcement learning tasks. Instead of following the multi-objective framework used in existing methods, we propose to rethink the problem under a novel perspective of constrained optimization. We first introduce a new metric to evaluate the difference between policies and then design two practical novel policy generation methods following the new perspective. The two proposed methods, namely the Constrained Task Novel Bisector (CTNB) and the Interior Policy Differentiation (IPD), are derived from the feasible direction method and the interior point method commonly known in the constrained optimization literature. Experimental comparisons on the MuJoCo control suite show our methods can achieve substantial improvement over previous novelty-seeking methods in terms of both the novelty of policies and their performances in the primal task.

Zhenghao Peng, Xuyang Chen, Chengwen Xu et al.

Neural network based approximate computing is a universal architecture promising to gain tremendous energy-efficiency for many error resilient applications. To guarantee the approximation quality, existing works deploy two neural networks (NNs), e.g., an approximator and a predictor. The approximator provides the approximate results, while the predictor predicts whether the input data is safe to approximate with the given quality requirement. However, it is non-trivial and time-consuming to make these two neural network coordinate---they have different optimization objectives---by training them separately. This paper proposes a novel neural network structure---AXNet---to fuse two NNs to a holistic end-to-end trainable NN. Leveraging the philosophy of multi-task learning, AXNet can tremendously improve the invocation (proportion of safe-to-approximate samples) and reduce the approximation error. The training effort also decrease significantly. Experiment results show 50.7% more invocation and substantial cuts of training time when compared to existing neural network based approximate computing framework.

Zhuoran Song, Ru Wang, Dongyu Ru et al.

The training phases of Deep neural network~(DNN) consumes enormous processing time and energy. Compression techniques utilizing the sparsity of DNNs can effectively accelerate the inference phase of DNNs. However, it can be hardly used in the training phase because the training phase involves dense matrix-multiplication using General Purpose Computation on Graphics Processors (GPGPU), which endorse regular and structural data layout. In this paper, we propose the Approximate Random Dropout that replaces the conventional random dropout of neurons and synapses with a regular and predefined patterns to eliminate the unnecessary computation and data access. To compensate the potential performance loss we develop a SGD-based Search Algorithm to produce the distribution of dropout patterns. We prove our approach is statistically equivalent to the previous dropout method. Experiments results on MLP and LSTM using well-known benchmarks show that the proposed Approximate Random Dropout can reduce the training time by $20\%$-$77\%$ ($19\%$-$60\%$) when dropout rate is $0.3$-$0.7$ on MLP (LSTM) with marginal accuracy drop.