Actor-critic (AC) is a powerful method for learning an optimal policy in reinforcement learning, where the critic uses algorithms, e.g., temporal difference (TD) learning with function approximation, to evaluate the current policy and the actor updates the policy along an approximate gradient direction using information from the critic. This paper provides the \textit{tightest} non-asymptotic convergence bounds for both the AC and natural AC (NAC) algorithms. Specifically, existing studies show that AC converges to an $\epsilon+\varepsilon_{\text{critic}}$ neighborhood of stationary points with the best known sample complexity of $\mathcal{O}(\epsilon^{-2})$ (up to a log factor), and NAC converges to an $\epsilon+\varepsilon_{\text{critic}}+\sqrt{\varepsilon_{\text{actor}}}$ neighborhood of the global optimum with the best known sample complexity of $\mathcal{O}(\epsilon^{-3})$, where $\varepsilon_{\text{critic}}$ is the approximation error of the critic and $\varepsilon_{\text{actor}}$ is the approximation error induced by the insufficient expressive power of the parameterized policy class. This paper analyzes the convergence of both AC and NAC algorithms with compatible function approximation. Our analysis eliminates the term $\varepsilon_{\text{critic}}$ from the error bounds while still achieving the best known sample complexities. Moreover, we focus on the challenging single-loop setting with a single Markovian sample trajectory. Our major technical novelty lies in analyzing the stochastic bias due to policy-dependent and time-varying compatible function approximation in the critic, and handling the non-ergodicity of the MDP due to the single Markovian sample trajectory. Numerical results are also provided in the appendix.

Algorithmic trading refers to executing buy and sell orders for specific assets based on automatically identified trading opportunities. Strategies based on reinforcement learning (RL) have demonstrated remarkable capabilities in addressing algorithmic trading problems. However, the trading patterns differ among market conditions due to shifted distribution data. Ignoring multiple patterns in the data will undermine the performance of RL. In this paper, we propose MOT, which designs multiple actors with disentangled representation learning to model the different patterns of the market. Furthermore, we incorporate the Optimal Transport (OT) algorithm to allocate samples to the appropriate actor by introducing a regularization loss term. Additionally, we propose Pretrain Module to facilitate imitation learning by aligning the outputs of actors with expert strategy and better balance the exploration and exploitation of RL. Experimental results on real futures market data demonstrate that MOT exhibits excellent profit capabilities while balancing risks.

As a data-driven paradigm, offline reinforcement learning (RL) has been formulated as sequence modeling that conditions on the hindsight information including returns, goal or future trajectory. Although promising, this supervised paradigm overlooks the core objective of RL that maximizes the return. This overlook directly leads to the lack of trajectory stitching capability that affects the sequence model learning from sub-optimal data. In this work, we introduce the concept of max-return sequence modeling which integrates the goal of maximizing returns into existing sequence models. We propose Reinforced Transformer (Reinformer), indicating the sequence model is reinforced by the RL objective. Reinformer additionally incorporates the objective of maximizing returns in the training phase, aiming to predict the maximum future return within the distribution. During inference, this in-distribution maximum return will guide the selection of optimal actions. Empirically, Reinformer is competitive with classical RL methods on the D4RL benchmark and outperforms state-of-the-art sequence model particularly in trajectory stitching ability. Code is public at https://github.com/Dragon-Zhuang/Reinformer.

This paper proposes a novel Reinforcement Learning (RL) approach for sim-to-real policy transfer of Vertical Take-Off and Landing Unmanned Aerial Vehicle (VTOL-UAV). The proposed approach is designed for VTOL-UAV landing on offshore docking stations in maritime operations. VTOL-UAVs in maritime operations encounter limitations in their operational range, primarily stemming from constraints imposed by their battery capacity. The concept of autonomous landing on a charging platform presents an intriguing prospect for mitigating these limitations by facilitating battery charging and data transfer. However, current Deep Reinforcement Learning (DRL) methods exhibit drawbacks, including lengthy training times, and modest success rates. In this paper, we tackle these concerns comprehensively by decomposing the landing procedure into a sequence of more manageable but analogous tasks in terms of an approach phase and a landing phase. The proposed architecture utilizes a model-based control scheme for the approach phase, where the VTOL-UAV is approaching the offshore docking station. In the Landing phase, DRL agents were trained offline to learn the optimal policy to dock on the offshore station. The Joint North Sea Wave Project (JONSWAP) spectrum model has been employed to create a wave model for each episode, enhancing policy generalization for sim2real transfer. A set of DRL algorithms have been tested through numerical simulations including value-based agents and policy-based agents such as Deep \textit{Q} Networks (DQN) and Proximal Policy Optimization (PPO) respectively. The numerical experiments show that the PPO agent can learn complicated and efficient policies to land in uncertain environments, which in turn enhances the likelihood of successful sim-to-real transfer.

Generating diverse and realistic human motion that can physically interact with an environment remains a challenging research area in character animation. Meanwhile, diffusion-based methods, as proposed by the robotics community, have demonstrated the ability to capture highly diverse and multi-modal skills. However, naively training a diffusion policy often results in unstable motions for high-frequency, under-actuated control tasks like bipedal locomotion due to rapidly accumulating compounding errors, pushing the agent away from optimal training trajectories. The key idea lies in using RL policies not just for providing optimal trajectories but for providing corrective actions in sub-optimal states, giving the policy a chance to correct for errors caused by environmental stimulus, model errors, or numerical errors in simulation. Our method, Physics-Based Character Animation via Diffusion Policy (PDP), combines reinforcement learning (RL) and behavior cloning (BC) to create a robust diffusion policy for physics-based character animation. We demonstrate PDP on perturbation recovery, universal motion tracking, and physics-based text-to-motion synthesis.

World models empower model-based agents to interactively explore, reason, and plan within imagined environments for real-world decision-making. However, the high demand for interactivity poses challenges in harnessing recent advancements in video generative models for developing world models at scale. This work introduces Interactive VideoGPT (iVideoGPT), a scalable autoregressive transformer framework that integrates multimodal signals--visual observations, actions, and rewards--into a sequence of tokens, facilitating an interactive experience of agents via next-token prediction. iVideoGPT features a novel compressive tokenization technique that efficiently discretizes high-dimensional visual observations. Leveraging its scalable architecture, we are able to pre-train iVideoGPT on millions of human and robotic manipulation trajectories, establishing a versatile foundation that is adaptable to serve as interactive world models for a wide range of downstream tasks. These include action-conditioned video prediction, visual planning, and model-based reinforcement learning, where iVideoGPT achieves competitive performance compared with state-of-the-art methods. Our work advances the development of interactive general world models, bridging the gap between generative video models and practical model-based reinforcement learning applications.

Bike-sharing systems play a crucial role in easing traffic congestion and promoting healthier lifestyles. However, ensuring their reliability and user acceptance requires effective strategies for rebalancing bikes. This study introduces a novel approach to address the real-time rebalancing problem with a fleet of vehicles. It employs a dual policy reinforcement learning algorithm that decouples inventory and routing decisions, enhancing realism and efficiency compared to previous methods where both decisions were made simultaneously. We first formulate the inventory and routing subproblems as a multi-agent Markov Decision Process within a continuous time framework. Subsequently, we propose a DQN-based dual policy framework to jointly estimate the value functions, minimizing the lost demand. To facilitate learning, a comprehensive simulator is applied to operate under a first-arrive-first-serve rule, which enables the computation of immediate rewards across diverse demand scenarios. We conduct extensive experiments on various datasets generated from historical real-world data, affected by both temporal and weather factors. Our proposed algorithm demonstrates significant performance improvements over previous baseline methods. It offers valuable practical insights for operators and further explores the incorporation of reinforcement learning into real-world dynamic programming problems, paving the way for more intelligent and robust urban mobility solutions.

Deep reinforcement learning (RL) algorithms typically parameterize the policy as a deep network that outputs either a deterministic action or a stochastic one modeled as a Gaussian distribution, hence restricting learning to a single behavioral mode. Meanwhile, diffusion models emerged as a powerful framework for multimodal learning. However, the use of diffusion policies in online RL is hindered by the intractability of policy likelihood approximation, as well as the greedy objective of RL methods that can easily skew the policy to a single mode. This paper presents Deep Diffusion Policy Gradient (DDiffPG), a novel actor-critic algorithm that learns from scratch multimodal policies parameterized as diffusion models while discovering and maintaining versatile behaviors. DDiffPG explores and discovers multiple modes through off-the-shelf unsupervised clustering combined with novelty-based intrinsic motivation. DDiffPG forms a multimodal training batch and utilizes mode-specific Q-learning to mitigate the inherent greediness of the RL objective, ensuring the improvement of the diffusion policy across all modes. Our approach further allows the policy to be conditioned on mode-specific embeddings to explicitly control the learned modes. Empirical studies validate DDiffPG's capability to master multimodal behaviors in complex, high-dimensional continuous control tasks with sparse rewards, also showcasing proof-of-concept dynamic online replanning when navigating mazes with unseen obstacles.

We observe that current state-of-the-art (SOTA) methods suffer from the performance imbalance issue when performing multi-task reinforcement learning (MTRL) tasks. While these methods may achieve impressive performance on average, they perform extremely poorly on a few tasks. To address this, we propose a new and effective method called STARS, which consists of two novel strategies: a shared-unique feature extractor and task-aware prioritized sampling. First, the shared-unique feature extractor learns both shared and task-specific features to enable better synergy of knowledge between different tasks. Second, the task-aware sampling strategy is combined with the prioritized experience replay for efficient learning on tasks with poor performance. The effectiveness and stability of our STARS are verified through experiments on the mainstream Meta-World benchmark. From the results, our STARS statistically outperforms current SOTA methods and alleviates the performance imbalance issue. Besides, we visualize the learned features to support our claims and enhance the interpretability of STARS.

In this study, we enhance the Maximum Entropy Deep Inverse Reinforcement Learning (MEDIRL) framework, targeting its application in human robot interaction (HRI) for modeling pedestrian behavior in crowded environments. Our work is grounded in the pioneering research by Fahad, Chen, and Guo, and aims to elevate MEDIRL's efficacy in real world HRI settings. We replicated the original MEDIRL model and conducted detailed ablation studies, focusing on key model components like learning rates, state dimensions, and network layers. Our findings reveal the effectiveness of a two dimensional state representation over three dimensional approach, significantly improving model accuracy for pedestrian behavior prediction in HRI scenarios. These results not only demonstrate MEDIRL's enhanced performance but also offer valuable insights for future HRI system development, emphasizing the importance of model customization to specific environmental contexts. Our research contributes to advancing the field of socially intelligent navigation systems, promoting more intuitive and safer human robot interactions.