In contact-rich tasks, the hybrid, multi-modal nature of contact dynamics poses great challenges in model representation, planning, and control. Recent efforts have attempted to address these challenges via data-driven methods, learning dynamical models in combination with model predictive control. Those methods, while effective, rely solely on minimizing forward prediction errors to hope for better task performance with MPC controllers. This weak correlation can result in data inefficiency as well as limitations to overall performance. In response, we propose a novel strategy: using a policy gradient algorithm to find a simplified dynamics model that explicitly maximizes task performance. Specifically, we parameterize the stochastic policy as the perturbed output of the MPC controller, thus, the learned model representation can directly associate with the policy or task performance. We apply the proposed method to contact-rich tasks where a three-fingered robotic hand manipulates previously unknown objects. Our method significantly enhances task success rate by up to 15% in manipulating diverse objects compared to the existing method while sustaining data efficiency. Our method can solve some tasks with success rates of 70% or higher using under 30 minutes of data. All videos and codes are available at https://sites.google.com/view/lcs-rl.

Recently, model-based reinforcement learning algorithms have demonstrated remarkable efficacy in visual input environments. These approaches begin by constructing a parameterized simulation world model of the real environment through self-supervised learning. By leveraging the imagination of the world model, the agent's policy is enhanced without the constraints of sampling from the real environment. The performance of these algorithms heavily relies on the sequence modeling and generation capabilities of the world model. However, constructing a perfectly accurate model of a complex unknown environment is nearly impossible. Discrepancies between the model and reality may cause the agent to pursue virtual goals, resulting in subpar performance in the real environment. Introducing random noise into model-based reinforcement learning has been proven beneficial. In this work, we introduce Stochastic Transformer-based wORld Model (STORM), an efficient world model architecture that combines the strong sequence modeling and generation capabilities of Transformers with the stochastic nature of variational autoencoders. STORM achieves a mean human performance of $126.7\%$ on the Atari $100$k benchmark, setting a new record among state-of-the-art methods that do not employ lookahead search techniques. Moreover, training an agent with $1.85$ hours of real-time interaction experience on a single NVIDIA GeForce RTX 3090 graphics card requires only $4.3$ hours, showcasing improved efficiency compared to previous methodologies.

Pre-defined manipulation primitives are widely used for cloth manipulation. However, cloth properties such as its stiffness or density can highly impact the performance of these primitives. Although existing solutions have tackled the parameterisation of pick and place locations, the effect of factors such as the velocity or trajectory of quasi-static and dynamic manipulation primitives has been neglected. Choosing appropriate values for these parameters is crucial to cope with the range of materials present in house-hold cloth objects. To address this challenge, we introduce the Quasi-Dynamic Parameterisable (QDP) method, which optimises parameters such as the motion velocity in addition to the pick and place positions of quasi-static and dynamic manipulation primitives. In this work, we leverage the framework of Sequential Reinforcement Learning to decouple sequentially the parameters that compose the primitives. To evaluate the effectiveness of the method we focus on the task of cloth unfolding with a robotic arm in simulation and real-world experiments. Our results in simulation show that by deciding the optimal parameters for the primitives the performance can improve by 20% compared to sub-optimal ones. Real-world results demonstrate the advantage of modifying the velocity and height of manipulation primitives for cloths with different mass, stiffness, shape and size. Supplementary material, videos, and code, can be found at https://sites.google.com/view/qdp-srl.

Column generation (CG) is a vital method to solve large-scale problems by dynamically generating variables. It has extensive applications in common combinatorial optimization, such as vehicle routing and scheduling problems, where each iteration step requires solving an NP-hard constrained shortest path problem. Although some heuristic methods for acceleration already exist, they are not versatile enough to solve different problems. In this work, we propose a reinforcement learning-based hyper-heuristic framework, dubbed RLHH, to enhance the performance of CG. RLHH is a selection module embedded in CG to accelerate convergence and get better integer solutions. In each CG iteration, the RL agent selects a low-level heuristic to construct a reduced network only containing the edges with a greater chance of being part of the optimal solution. In addition, we specify RLHH to solve two typical combinatorial optimization problems: Vehicle Routing Problem with Time Windows (VRPTW) and Bus Driver Scheduling Problem (BDSP). The total cost can be reduced by up to 27.9\% in VRPTW and 15.4\% in BDSP compared to the best lower-level heuristic in our tested scenarios, within equivalent or even less computational time. The proposed RLHH is the first RL-based CG method that outperforms traditional approaches in terms of solution quality, which can promote the application of CG in combinatorial optimization.

Federated learning (FL) is emerging as a sought-after distributed machine learning architecture, offering the advantage of model training without direct exposure of raw data. With advancements in network infrastructure, FL has been seamlessly integrated into edge computing. However, the limited resources on edge devices introduce security vulnerabilities to FL in the context. While blockchain technology promises to bolster security, practical deployment on resource-constrained edge devices remains a challenge. Moreover, the exploration of FL with multiple aggregators in edge computing is still new in the literature. Addressing these gaps, we introduce the Blockchain-empowered Heterogeneous Multi-Aggregator Federated Learning Architecture (BMA-FL). We design a novel light-weight Byzantine consensus mechanism, namely PBCM, to enable secure and fast model aggregation and synchronization in BMA-FL. We also dive into the heterogeneity problem in BMA-FL that the aggregators are associated with varied number of connected trainers with Non-IID data distributions and diverse training speed. We proposed a multi-agent deep reinforcement learning algorithm to help aggregators decide the best training strategies. The experiments on real-word datasets demonstrate the efficiency of BMA-FL to achieve better models faster than baselines, showing the efficacy of PBCM and proposed deep reinforcement learning algorithm.

Diffusion models have risen as a promising approach to data-driven planning, and have demonstrated impressive robotic control, reinforcement learning, and video planning performance. Given an effective planner, an important question to consider is replanning -- when given plans should be regenerated due to both action execution error and external environment changes. Direct plan execution, without replanning, is problematic as errors from individual actions rapidly accumulate and environments are partially observable and stochastic. Simultaneously, replanning at each timestep incurs a substantial computational cost, and may prevent successful task execution, as different generated plans prevent consistent progress to any particular goal. In this paper, we explore how we may effectively replan with diffusion models. We propose a principled approach to determine when to replan, based on the diffusion model's estimated likelihood of existing generated plans. We further present an approach to replan existing trajectories to ensure that new plans follow the same goal state as the original trajectory, which may efficiently bootstrap off previously generated plans. We illustrate how a combination of our proposed additions significantly improves the performance of diffusion planners leading to 38\% gains over past diffusion planning approaches on Maze2D, and further enables the handling of stochastic and long-horizon robotic control tasks. Videos can be found on the anonymous website: \url{https://vis-www.cs.umass.edu/replandiffuser/}.

The forthcoming 6G systems are expected to address a wide range of non-stationary tasks. This poses challenges to traditional medium access control (MAC) protocols that are static and predefined. In response, data-driven MAC protocols have recently emerged, offering ability to tailor their signaling messages for specific tasks. This article presents a novel categorization of these data-driven MAC protocols into three levels: Level 1 MAC. task-oriented neural protocols constructed using multi-agent deep reinforcement learning (MADRL); Level 2 MAC. neural network-oriented symbolic protocols developed by converting Level 1 MAC outputs into explicit symbols; and Level 3 MAC. language-oriented semantic protocols harnessing large language models (LLMs) and generative models. With this categorization, we aim to explore the opportunities and challenges of each level by delving into their foundational techniques. Drawing from information theory and associated principles as well as selected case studies, this study provides insights into the trajectory of data-driven MAC protocols and sheds light on future research directions.

Off-policy evaluation (OPE) is important for closing the gap between offline training and evaluation of reinforcement learning (RL), by estimating performance and/or rank of target (evaluation) policies using offline trajectories only. It can improve the safety and efficiency of data collection and policy testing procedures in situations where online deployments are expensive, such as healthcare. However, existing OPE methods fall short in estimating human feedback (HF) signals, as HF may be conditioned over multiple underlying factors and is only sparsely available; as opposed to the agent-defined environmental rewards (used in policy optimization), which are usually determined over parametric functions or distributions. Consequently, the nature of HF signals makes extrapolating accurate OPE estimations to be challenging. To resolve this, we introduce an OPE for HF (OPEHF) framework that revives existing OPE methods in order to accurately evaluate the HF signals. Specifically, we develop an immediate human reward (IHR) reconstruction approach, regularized by environmental knowledge distilled in a latent space that captures the underlying dynamics of state transitions as well as issuing HF signals. Our approach has been tested over two real-world experiments, adaptive in-vivo neurostimulation and intelligent tutoring, as well as in a simulation environment (visual Q&A). Results show that our approach significantly improves the performance toward estimating HF signals accurately, compared to directly applying (variants of) existing OPE methods.

Recent advances in reinforcement learning (RL) have shown much promise across a variety of applications. However, issues such as scalability, explainability, and Markovian assumptions limit its applicability in certain domains. We observe that many of these shortcomings emanate from the simulator as opposed to the RL training algorithms themselves. As such, we propose a semantic proxy for simulation based on a temporal extension to annotated logic. In comparison with two high-fidelity simulators, we show up to three orders of magnitude speed-up while preserving the quality of policy learned. In addition, we show the ability to model and leverage non-Markovian dynamics and instantaneous actions while providing an explainable trace describing the outcomes of the agent actions.

This work focuses on equilibrium selection in no-conflict multi-agent games, where we specifically study the problem of selecting a Pareto-optimal Nash equilibrium among several existing equilibria. It has been shown that many state-of-the-art multi-agent reinforcement learning (MARL) algorithms are prone to converging to Pareto-dominated equilibria due to the uncertainty each agent has about the policy of the other agents during training. To address sub-optimal equilibrium selection, we propose Pareto Actor-Critic (Pareto-AC), which is an actor-critic algorithm that utilises a simple property of no-conflict games (a superset of cooperative games): the Pareto-optimal equilibrium in a no-conflict game maximises the returns of all agents and, therefore, is the preferred outcome for all agents. We evaluate Pareto-AC in a diverse set of multi-agent games and show that it converges to higher episodic returns compared to seven state-of-the-art MARL algorithms and that it successfully converges to a Pareto-optimal equilibrium in a range of matrix games. Finally, we propose PACDCG, a graph neural network extension of Pareto-AC, which is shown to efficiently scale in games with a large number of agents.