Active Voltage Control (AVC) on the Power Distribution Networks (PDNs) aims to stabilize the voltage levels to ensure efficient and reliable operation of power systems. With the increasing integration of distributed energy resources, recent efforts have explored employing multi-agent reinforcement learning (MARL) techniques to realize effective AVC. Existing methods mainly focus on the acquisition of short-term AVC strategies, i.e., only learning AVC within the short-term training trajectories of a singular diurnal cycle. However, due to the dynamic nature of load demands and renewable energy, the operation states of real-world PDNs may exhibit significant distribution shifts across varying timescales (e.g., daily and seasonal changes). This can render those short-term strategies suboptimal or even obsolete when performing continuous AVC over extended periods. In this paper, we propose a novel temporal prototype-aware learning method, abbreviated as TPA, to learn time-adaptive AVC under short-term training trajectories. At the heart of TPA are two complementary components, namely multi-scale dynamic encoder and temporal prototype-aware policy, that can be readily incorporated into various MARL methods. The former component integrates a stacked transformer network to learn underlying temporal dependencies at different timescales of the PDNs, while the latter implements a learnable prototype matching mechanism to construct a dedicated AVC policy that can dynamically adapt to the evolving operation states. Experimental results on the AVC benchmark with different PDN sizes demonstrate that the proposed TPA surpasses the state-of-the-art counterparts not only in terms of control performance but also by offering model transferability. Our code is available at https://github.com/Canyizl/TPA-for-AVC.

Payload transport over flat terrain via multi-wheel robot carriers is well-understood, highly effective, and configurable. In this paper, our goal is to provide similar effectiveness and configurability for transport over rough terrain that is more suitable for legs rather than wheels. For this purpose, we consider multi-biped robot carriers, where wheels are replaced by multiple bipedal robots attached to the carrier. Our main contribution is to design a decentralized controller for such systems that can be effectively applied to varying numbers and configurations of rigidly attached bipedal robots without retraining. We present a reinforcement learning approach for training the controller in simulation that supports transfer to the real world. Our experiments in simulation provide quantitative metrics showing the effectiveness of the approach over a wide variety of simulated transport scenarios. In addition, we demonstrate the controller in the real-world for systems composed of two and three Cassie robots. To our knowledge, this is the first example of a scalable multi-biped payload transport system.

We consider the problem of privacy protection in Reinforcement Learning (RL) algorithms that operate over population processes, a practical but understudied setting that includes, for example, the control of epidemics in large populations of dynamically interacting individuals. In this setting, the RL algorithm interacts with the population over $T$ time steps by receiving population-level statistics as state and performing actions which can affect the entire population at each time step. An individual's data can be collected across multiple interactions and their privacy must be protected at all times. We clarify the Bayesian semantics of Differential Privacy (DP) in the presence of correlated data in population processes through a Pufferfish Privacy analysis. We then give a meta algorithm that can take any RL algorithm as input and make it differentially private. This is achieved by taking an approach that uses DP mechanisms to privatize the state and reward signal at each time step before the RL algorithm receives them as input. Our main theoretical result shows that the value-function approximation error when applying standard RL algorithms directly to the privatized states shrinks quickly as the population size and privacy budget increase. This highlights that reasonable privacy-utility trade-offs are possible for differentially private RL algorithms in population processes. Our theoretical findings are validated by experiments performed on a simulated epidemic control problem over large population sizes.

In standard reinforcement learning settings, agents typically assume immediate feedback about the effects of their actions after taking them. However, in practice, this assumption may not hold true due to physical constraints and can significantly impact the performance of learning algorithms. In this paper, we address observation delays in partially observable environments. We propose leveraging world models, which have shown success in integrating past observations and learning dynamics, to handle observation delays. By reducing delayed POMDPs to delayed MDPs with world models, our methods can effectively handle partial observability, where existing approaches achieve sub-optimal performance or degrade quickly as observability decreases. Experiments suggest that one of our methods can outperform a naive model-based approach by up to 250%. Moreover, we evaluate our methods on visual delayed environments, for the first time showcasing delay-aware reinforcement learning continuous control with visual observations.

Adapting Large Language Models (LLMs) for agent tasks is critical in developing language agents. Direct Preference Optimization (DPO) is a promising technique for this adaptation with the alleviation of compounding errors, offering a means to directly optimize Reinforcement Learning (RL) objectives. However, applying DPO to multi-turn tasks presents challenges due to the inability to cancel the partition function. Overcoming this obstacle involves making the partition function independent of the current state and addressing length disparities between preferred and dis-preferred trajectories. In this light, we replace the policy constraint with the state-action occupancy measure constraint in the RL objective and add length normalization to the Bradley-Terry model, yielding a novel loss function named DMPO for multi-turn agent tasks with theoretical explanations. Extensive experiments on three multi-turn agent task datasets confirm the effectiveness and superiority of the DMPO loss.

Recent work has shown that deep reinforcement learning agents have difficulty in effectively using their network parameters. We leverage prior insights into the advantages of sparse training techniques and demonstrate that gradual magnitude pruning enables value-based agents to maximize parameter effectiveness. This results in networks that yield dramatic performance improvements over traditional networks, using only a small fraction of the full network parameters.

We study the finite-time convergence of TD learning with linear function approximation under Markovian sampling. Existing proofs for this setting either assume a projection step in the algorithm to simplify the analysis, or require a fairly intricate argument to ensure stability of the iterates. We ask: \textit{Is it possible to retain the simplicity of a projection-based analysis without actually performing a projection step in the algorithm?} Our main contribution is to show this is possible via a novel two-step argument. In the first step, we use induction to prove that under a standard choice of a constant step-size $\alpha$, the iterates generated by TD learning remain uniformly bounded in expectation. In the second step, we establish a recursion that mimics the steady-state dynamics of TD learning up to a bounded perturbation on the order of $O(\alpha^2)$ that captures the effect of Markovian sampling. Combining these pieces leads to an overall approach that considerably simplifies existing proofs. We conjecture that our inductive proof technique will find applications in the analyses of more complex stochastic approximation algorithms, and conclude by providing some examples of such applications.

An agent's ability to leverage past experience is critical for efficiently solving new tasks. Prior work has focused on using value function estimates to obtain zero-shot approximations for solutions to a new task. In soft Q-learning, we show how any value function estimate can also be used to derive double-sided bounds on the optimal value function. The derived bounds lead to new approaches for boosting training performance which we validate experimentally. Notably, we find that the proposed framework suggests an alternative method for updating the Q-function, leading to boosted performance.

Cyber-physical systems (CPS) with reinforcement learning (RL)-based controllers are increasingly being deployed in complex physical environments such as autonomous vehicles, the Internet-of-Things(IoT), and smart cities. An important property of a CPS is tolerance; i.e., its ability to function safely under possible disturbances and uncertainties in the actual operation. In this paper, we introduce a new, expressive notion of tolerance that describes how well a controller is capable of satisfying a desired system requirement, specified using Signal Temporal Logic (STL), under possible deviations in the system. Based on this definition, we propose a novel analysis problem, called the tolerance falsification problem, which involves finding small deviations that result in a violation of the given requirement. We present a novel, two-layer simulation-based analysis framework and a novel search heuristic for finding small tolerance violations. To evaluate our approach, we construct a set of benchmark problems where system parameters can be configured to represent different types of uncertainties and disturbancesin the system. Our evaluation shows that our falsification approach and heuristic can effectively find small tolerance violations.

Temporal distances lie at the heart of many algorithms for planning, control, and reinforcement learning that involve reaching goals, allowing one to estimate the transit time between two states. However, prior attempts to define such temporal distances in stochastic settings have been stymied by an important limitation: these prior approaches do not satisfy the triangle inequality. This is not merely a definitional concern, but translates to an inability to generalize and find shortest paths. In this paper, we build on prior work in contrastive learning and quasimetrics to show how successor features learned by contrastive learning (after a change of variables) form a temporal distance that does satisfy the triangle inequality, even in stochastic settings. Importantly, this temporal distance is computationally efficient to estimate, even in high-dimensional and stochastic settings. Experiments in controlled settings and benchmark suites demonstrate that an RL algorithm based on these new temporal distances exhibits combinatorial generalization (i.e., "stitching") and can sometimes learn more quickly than prior methods, including those based on quasimetrics.