Qing-Shan Jia

Shuang Wu, Xiaoqiang Ren, Qing-Shan Jia et al.

We consider optimal sensor scheduling with unknown communication channel statistics. We formulate two types of scheduling problems with the communication rate being a soft or hard constraint, respectively. We first present some structural results on the optimal scheduling policy using dynamic programming and assuming the channel statistics is known. We prove that the Q-factor is monotonic and submodular, which leads to the threshold-like structures in both types of problems. Then we develop a stochastic approximation and parameter learning frameworks to deal with the two scheduling problems with unknown channel statistics. We utilize their structures to design specialized learning algorithms. We prove the convergence of these algorithms. Performance improvement compared with the standard Q-learning algorithm is shown through numerical examples.

Jianxiong Li, Xiao Hu, Haoran Xu et al. · tsinghua

Reward function is essential in reinforcement learning (RL), serving as the guiding signal to incentivize agents to solve given tasks, however, is also notoriously difficult to design. In many cases, only imperfect rewards are available, which inflicts substantial performance loss for RL agents. In this study, we propose a unified offline policy optimization approach, \textit{RGM (Reward Gap Minimization)}, which can smartly handle diverse types of imperfect rewards. RGM is formulated as a bi-level optimization problem: the upper layer optimizes a reward correction term that performs visitation distribution matching w.r.t. some expert data; the lower layer solves a pessimistic RL problem with the corrected rewards. By exploiting the duality of the lower layer, we derive a tractable algorithm that enables sampled-based learning without any online interactions. Comprehensive experiments demonstrate that RGM achieves superior performance to existing methods under diverse settings of imperfect rewards. Further, RGM can effectively correct wrong or inconsistent rewards against expert preference and retrieve useful information from biased rewards.

Yao Luan, Ni Mu, Hanfei Ge et al.

Unsupervised skill discovery (USD) aims to learn diverse behaviors without reward functions, but often results in task-irrelevant or hazardous behaviors due to uniform exploration. Guided skill discovery (GSD) addresses this issue by incorporating human intent to focus exploration on meaningful regions. However, existing GSD methods typically require training additional guidance models, and rely on pre-defined rules or expert demonstration, which can be ineffective under sparse, online-collected human feedback. To overcome this, we propose COLLIE, a GSD framework that leverages dense unsupervised data to construct a semantically coherent skill latent space. This latent space is well-structured, enabling reliable guidance with sparse online feedback. Moreover, its semantic coherence property enables training-free construction of guidance signals, eliminating the need for additional model training beyond skill learning. Theoretical analysis justifies the effectiveness of our training-free guidance signal, while experiments across diverse state-based and pixel-based tasks show that COLLIE learns diverse, human-aligned skills, avoids hazardous behaviors, and achieves superior downstream performance with minimal human feedback.

Qing-Shan Jia

The dynamics of many systems nowadays follow not only physical laws but also man-made rules. These systems are known as discrete event dynamic systems and their performances can be accurately evaluated only through simulations. Existing studies on simulation-based optimization (SBO) usually assume deterministic simulation time for each replication. However, in many applications such as evacuation, smoke detection, and territory exploration, the simulation time is stochastic due to the randomness in the system behavior. We consider the computing budget allocation for SBO's with stochastic simulation time in this paper, which has not been addressed in existing literatures to the author's best knowledge. We make the following major contribution. The relationship between simulation time and performance estimation accuracy is quantified. It is shown that when the asymptotic performance is of interest only the mean value of individual simulation time matters. Then based on the existing optimal computing budget allocation (OCBA) method for deterministic simulation time we develop OCBA for stochastic simulation time (OCBAS), and show that OCBAS is asymptotically optimal. Numerical experiments are used to discuss the impact of the variance of simulation time, the impact of correlated simulation time and performance estimation, and to demonstrate the performance of OCBAS on a smoke detection problem in wireless sensor network. The numerical results also show that OCBA for deterministic simulation time is robust even when the simulation time is stochastic.

Ni Mu, Yao Luan, Yiqin Yang et al.

Preference-based reinforcement learning (PbRL) stands out by utilizing human preferences as a direct reward signal, eliminating the need for intricate reward engineering. However, despite its potential, traditional PbRL methods are often constrained by the indistinguishability of segments, which impedes the learning process. In this paper, we introduce Skill-Enhanced Preference Optimization Algorithm (S-EPOA), which addresses the segment indistinguishability issue by integrating skill mechanisms into the preference learning framework. Specifically, we first conduct the unsupervised pretraining to learn useful skills. Then, we propose a novel query selection mechanism to balance the information gain and distinguishability over the learned skill space. Experimental results on a range of tasks, including robotic manipulation and locomotion, demonstrate that S-EPOA significantly outperforms conventional PbRL methods in terms of both robustness and learning efficiency. The results highlight the effectiveness of skill-driven learning in overcoming the challenges posed by segment indistinguishability.

Ni Mu, Yao Luan, Qing-Shan Jia

Multi-objective reinforcement learning (MORL) is a structured approach for optimizing tasks with multiple objectives. However, it often relies on pre-defined reward functions, which can be hard to design for balancing conflicting goals and may lead to oversimplification. Preferences can serve as more flexible and intuitive decision-making guidance, eliminating the need for complicated reward design. This paper introduces preference-based MORL (Pb-MORL), which formalizes the integration of preferences into the MORL framework. We theoretically prove that preferences can derive policies across the entire Pareto frontier. To guide policy optimization using preferences, our method constructs a multi-objective reward model that aligns with the given preferences. We further provide theoretical proof to show that optimizing this reward model is equivalent to training the Pareto optimal policy. Extensive experiments in benchmark multi-objective tasks, a multi-energy management task, and an autonomous driving task on a multi-line highway show that our method performs competitively, surpassing the oracle method, which uses the ground truth reward function. This highlights its potential for practical applications in complex real-world systems.

Ni Mu, Hao Hu, Xiao Hu et al.

Preference-based reinforcement learning (PbRL) bypasses explicit reward engineering by inferring reward functions from human preference comparisons, enabling better alignment with human intentions. However, humans often struggle to label a clear preference between similar segments, reducing label efficiency and limiting PbRL's real-world applicability. To address this, we propose an offline PbRL method: Contrastive LeArning for ResolvIng Ambiguous Feedback (CLARIFY), which learns a trajectory embedding space that incorporates preference information, ensuring clearly distinguished segments are spaced apart, thus facilitating the selection of more unambiguous queries. Extensive experiments demonstrate that CLARIFY outperforms baselines in both non-ideal teachers and real human feedback settings. Our approach not only selects more distinguished queries but also learns meaningful trajectory embeddings.

Yao Luan, Ni Mu, Yiqin Yang et al.

Preference-based reinforcement learning (PbRL) bypasses complex reward engineering by learning rewards directly from human preferences, enabling better alignment with human intentions. However, its effectiveness in multi-stage tasks, where agents sequentially perform sub-tasks (e.g., navigation, grasping), is limited by stage misalignment: Comparing segments from mismatched stages, such as movement versus manipulation, results in uninformative feedback, thus hindering policy learning. In this paper, we validate the stage misalignment issue through theoretical analysis and empirical experiments. To address this issue, we propose STage-AlIgned Reward learning (STAIR), which first learns a stage approximation based on temporal distance, then prioritizes comparisons within the same stage. Temporal distance is learned via contrastive learning, which groups temporally close states into coherent stages, without predefined task knowledge, and adapts dynamically to policy changes. Extensive experiments demonstrate STAIR's superiority in multi-stage tasks and competitive performance in single-stage tasks. Furthermore, human studies show that stages approximated by STAIR are consistent with human cognition, confirming its effectiveness in mitigating stage misalignment.

Linxuan He, Qing-Shan Jia, Ang Li et al.

Embodied AI systems, comprising AI models and physical plants, are increasingly prevalent across various applications. Due to the rarity of system failures, ensuring their safety in complex operating environments remains a major challenge, which severely hinders their large-scale deployment in safety-critical domains, such as autonomous vehicles, medical devices, and robotics. While achieving provable deterministic safety--verifying system safety across all possible scenarios--remains theoretically ideal, the rarity and complexity of corner cases make this approach impractical for scalable embodied AI systems. Instead, empirical safety evaluation is employed as an alternative, but the absence of provable guarantees imposes significant limitations. To address these issues, we argue for a paradigm shift to provable probabilistic safety that integrates provable guarantees with progressive achievement toward a probabilistic safety boundary on overall system performance. The new paradigm better leverages statistical methods to enhance feasibility and scalability, and a well-defined probabilistic safety boundary enables embodied AI systems to be deployed at scale. In this Perspective, we outline a roadmap for provable probabilistic safety, along with corresponding challenges and potential solutions. By bridging the gap between theoretical safety assurance and practical deployment, this Perspective offers a pathway toward safer, large-scale adoption of embodied AI systems in safety-critical applications.

Xiao Hu, Jianxiong Li, Xianyuan Zhan et al.

Preference-based reinforcement learning (PbRL) provides a natural way to align RL agents' behavior with human desired outcomes, but is often restrained by costly human feedback. To improve feedback efficiency, most existing PbRL methods focus on selecting queries to maximally improve the overall quality of the reward model, but counter-intuitively, we find that this may not necessarily lead to improved performance. To unravel this mystery, we identify a long-neglected issue in the query selection schemes of existing PbRL studies: Query-Policy Misalignment. We show that the seemingly informative queries selected to improve the overall quality of reward model actually may not align with RL agents' interests, thus offering little help on policy learning and eventually resulting in poor feedback efficiency. We show that this issue can be effectively addressed via near on-policy query and a specially designed hybrid experience replay, which together enforce the bidirectional query-policy alignment. Simple yet elegant, our method can be easily incorporated into existing approaches by changing only a few lines of code. We showcase in comprehensive experiments that our method achieves substantial gains in both human feedback and RL sample efficiency, demonstrating the importance of addressing query-policy misalignment in PbRL tasks.

Kuo Li, Qing-Shan Jia

We discuss the problem of decentralized multi-agent reinforcement learning (MARL) in this work. In our setting, the global state, action, and reward are assumed to be fully observable, while the local policy is protected as privacy by each agent, and thus cannot be shared with others. There is a communication graph, among which the agents can exchange information with their neighbors. The agents make individual decisions and cooperate to reach a higher accumulated reward. Towards this end, we first propose a decentralized actor-critic (AC) setting. Then, the policy evaluation and policy improvement algorithms are designed for discrete and continuous state-action-space Markov Decision Process (MDP) respectively. Furthermore, convergence analysis is given under the discrete-space case, which guarantees that the policy will be reinforced by alternating between the processes of policy evaluation and policy improvement. In order to validate the effectiveness of algorithms, we design experiments and compare them with previous algorithms, e.g., Q-learning \cite{watkins1992q} and MADDPG \cite{lowe2017multi}. The results show that our algorithms perform better from the aspects of both learning speed and final performance. Moreover, the algorithms can be executed in an off-policy manner, which greatly improves the data efficiency compared with on-policy algorithms.

Kuo Li, Qing-Shan Jia, Jiaqi Yan

We focus on a simulation-based optimization problem of choosing the best design from the feasible space. Although the simulation model can be queried with finite samples, its internal processing rule cannot be utilized in the optimization process. We formulate the sampling process as a policy searching problem and give a solution from the perspective of Reinforcement Learning (RL). Concretely, Actor-Critic (AC) framework is applied, where the Actor serves as a surrogate model to predict the performance on unknown designs, whereas the actor encodes the sampling policy to be optimized. We design the updating rule and propose two algorithms for the cases where the feasible spaces are continuous and discrete respectively. Some experiments are designed to validate the effectiveness of proposed algorithms, including two toy examples, which intuitively explain the algorithms, and two more complex tasks, i.e., adversarial attack task and RL task, which validate the effectiveness in large-scale problems. The results show that the proposed algorithms can successfully deal with these problems. Especially note that in the RL task, our methods give a new perspective to robot control by treating the task as a simulation model and solving it by optimizing the policy generating process, while existing works commonly optimize the policy itself directly.