Recently, methods such as Decision Transformer that reduce reinforcement learning to a prediction task and solve it via supervised learning (RvS) have become popular due to their simplicity, robustness to hyperparameters, and strong overall performance on offline RL tasks. However, simply conditioning a probabilistic model on a desired return and taking the predicted action can fail dramatically in stochastic environments since trajectories that result in a return may have only achieved that return due to luck. In this work, we describe the limitations of RvS approaches in stochastic environments and propose a solution. Rather than simply conditioning on the return of a single trajectory as is standard practice, our proposed method, ESPER, learns to cluster trajectories and conditions on average cluster returns, which are independent from environment stochasticity. Doing so allows ESPER to achieve strong alignment between target return and expected performance in real environments. We demonstrate this in several challenging stochastic offline-RL tasks including the challenging puzzle game 2048, and Connect Four playing against a stochastic opponent. In all tested domains, ESPER achieves significantly better alignment between the target return and achieved return than simply conditioning on returns. ESPER also achieves higher maximum performance than even the value-based baselines.

Existing offline reinforcement learning (RL) algorithms typically assume that training data is either: 1) generated by a known policy, or 2) of entirely unknown origin. We consider multi-demonstrator offline RL, a middle ground where we know which demonstrators generated each dataset, but make no assumptions about the underlying policies of the demonstrators. This is the most natural setting when collecting data from multiple human operators, yet remains unexplored. Since different demonstrators induce different data distributions, we show that this can be naturally framed as a domain generalization problem, with each demonstrator corresponding to a different domain. Specifically, we propose Domain-Invariant Model-based Offline RL (DIMORL), where we apply Risk Extrapolation (REx) (Krueger et al., 2020) to the process of learning dynamics and rewards models. Our results show that models trained with REx exhibit improved domain generalization performance when compared with the natural baseline of pooling all demonstrators' data. We observe that the resulting models frequently enable the learning of superior policies in the offline model-based RL setting, can improve the stability of the policy learning process, and potentially enable increased exploration.

Due to their decentralised and intermittent nature, market and tariff designs are challenged. In the Flemish region of Belgium the energy regulator (VREG) has recently announced a change of distribution fee design [28]. Previously, Flemish residential electricity distribution fees have been energy-based. The rise of residential photovoltaic (PV) installations and netmetering meant a reduction in income for the Distribution Network Operator (DNO). With the introduction of digital metering, the regulator takes the opportunity to introduce a capacity tariff, starting from 2023 [28]. The regulator motivates its decision by arguing that the DNO's main costs are capacity-based rather than energy-based [28]. As a result, a more cost-reflective price signal is provided to consumers. At the same time, by tying the distribution grid fees to power rather than energy consumption, the VREG encourages consumers, aiming to minizime cost, to reduce their peak power consumption. This causes an interesting application for Demand Response (DR), which we will consider in this work.

Value iteration (VI) is a foundational dynamic programming method, important for learning and planning in optimal control and reinforcement learning. VI proceeds in batches, where the update to the value of each state must be completed before the next batch of updates can begin. Completing a single batch is prohibitively expensive if the state space is large, rendering VI impractical for many applications. Asynchronous VI helps to address the large state space problem by updating one state at a time, in-place and in an arbitrary order. However, Asynchronous VI still requires a maximization over the entire action space, making it impractical for domains with large action space. To address this issue, we propose doubly-asynchronous value iteration (DAVI), a new algorithm that generalizes the idea of asynchrony from states to states and actions. More concretely, DAVI maximizes over a sampled subset of actions that can be of any user-defined size. This simple approach of using sampling to reduce computation maintains similarly appealing theoretical properties to VI without the need to wait for a full sweep through the entire action space in each update. In this paper, we show DAVI converges to the optimal value function with probability one, converges at a near-geometric rate with probability 1-delta, and returns a near-optimal policy in computation time that nearly matches a previously established bound for VI. We also empirically demonstrate DAVI's effectiveness in several experiments.

Constrained reinforcement learning (RL) is an area of RL whose objective is to find an optimal policy that maximizes expected cumulative return while satisfying a given constraint. Most of the previous constrained RL works consider expected cumulative sum cost as the constraint. However, optimization with this constraint cannot guarantee a target probability of outage event that the cumulative sum cost exceeds a given threshold. This paper proposes a framework, named Quantile Constrained RL (QCRL), to constrain the quantile of the distribution of the cumulative sum cost that is a necessary and sufficient condition to satisfy the outage constraint. This is the first work that tackles the issue of applying the policy gradient theorem to the quantile and provides theoretical results for approximating the gradient of the quantile. Based on the derived theoretical results and the technique of the Lagrange multiplier, we construct a constrained RL algorithm named Quantile Constrained Policy Optimization (QCPO). We use distributional RL with the Large Deviation Principle (LDP) to estimate quantiles and tail probability of the cumulative sum cost for the implementation of QCPO. The implemented algorithm satisfies the outage probability constraint after the training period.

Automated Market Makers (AMMs) have cemented themselves as an integral part of the decentralized finance (DeFi) space. AMMs are a type of exchange that allows users to trade assets without the need for a centralized exchange. They form the foundation for numerous decentralized exchanges (DEXs), which help facilitate the quick and efficient exchange of on-chain tokens. All present-day popular DEXs are static protocols, with fixed parameters controlling the fee and the curvature - they suffer from invariance and cannot adapt to quickly changing market conditions. This characteristic may cause traders to stay away during high slippage conditions brought about by intractable market movements. We propose an RL framework to optimize the fees collected on an AMM protocol. In particular, we develop a Q-Learning Agent for Market Making Protocols (QLAMMP) that learns the optimal fee rates and leverage coefficients for a given AMM protocol and maximizes the expected fee collected under a range of different market conditions. We show that QLAMMP is consistently able to outperform its static counterparts under all the simulated test conditions.

Deep reinforcement learning has achieved great success in various fields with its super decision-making ability. However, the policy learning process requires a large amount of training time, causing energy consumption. Inspired by the redundancy of neural networks, we propose a lightweight parallel training framework based on neural network compression, AcceRL, to accelerate the policy learning while ensuring policy quality. Specifically, AcceRL speeds up the experience collection by flexibly combining various neural network compression methods. Overall, the AcceRL consists of five components, namely Actor, Learner, Compressor, Corrector, and Monitor. The Actor uses the Compressor to compress the Learner's policy network to interact with the environment. And the generated experiences are transformed by the Corrector with Off-Policy methods, such as V-trace, Retrace and so on. Then the corrected experiences are feed to the Learner for policy learning. We believe this is the first general reinforcement learning framework that incorporates multiple neural network compression techniques. Extensive experiments conducted in gym show that the AcceRL reduces the time cost of the actor by about 2.0 X to 4.13 X compared to the traditional methods. Furthermore, the AcceRL reduces the whole training time by about 29.8% to 40.3% compared to the traditional methods while keeps the same policy quality.

Pessimism is of great importance in offline reinforcement learning (RL). One broad category of offline RL algorithms fulfills pessimism by explicit or implicit behavior regularization. However, most of them only consider policy divergence as behavior regularization, ignoring the effect of how the offline state distribution differs with that of the learning policy, which may lead to under-pessimism for some states and over-pessimism for others. Taking account of this problem, we propose a principled algorithmic framework for offline RL, called \emph{State-Aware Proximal Pessimism} (SA-PP). The key idea of SA-PP is leveraging discounted stationary state distribution ratios between the learning policy and the offline dataset to modulate the degree of behavior regularization in a state-wise manner, so that pessimism can be implemented in a more appropriate way. We first provide theoretical justifications on the superiority of SA-PP over previous algorithms, demonstrating that SA-PP produces a lower suboptimality upper bound in a broad range of settings. Furthermore, we propose a new algorithm named \emph{State-Aware Conservative Q-Learning} (SA-CQL), by building SA-PP upon representative CQL algorithm with the help of DualDICE for estimating discounted stationary state distribution ratios. Extensive experiments on standard offline RL benchmark show that SA-CQL outperforms the popular baselines on a large portion of benchmarks and attains the highest average return.

Reinforcement learning (RL) is a promising solution for autonomous vehicles to deal with complex and uncertain traffic environments. The RL training process is however expensive, unsafe, and time consuming. Algorithms are often developed first in simulation and then transferred to the real world, leading to a common sim2real challenge that performance decreases when the domain changes. In this paper, we propose a transfer learning process to minimize the gap by exploiting digital twin technology, relying on a systematic and simultaneous combination of virtual and real world data coming from vehicle dynamics and traffic scenarios. The model and testing environment are evolved from model, hardware to vehicle in the loop and proving ground testing stages, similar to standard development cycle in automotive industry. In particular, we also integrate other transfer learning techniques such as domain randomization and adaptation in each stage. The simulation and real data are gradually incorporated to accelerate and make the transfer learning process more robust. The proposed RL methodology is applied to develop a path following steering controller for an autonomous electric vehicle. After learning and deploying the real-time RL control policy on the vehicle, we obtained satisfactory and safe control performance already from the first deployment, demonstrating the advantages of the proposed digital twin based learning process.

Recent technological advancements in space, air and ground components have made possible a new network paradigm called "space-air-ground integrated network" (SAGIN). Unmanned aerial vehicles (UAVs) play a key role in SAGINs. However, due to UAVs' high dynamics and complexity, the real-world deployment of a SAGIN becomes a major barrier for realizing such SAGINs. Compared to the space and terrestrial components, UAVs are expected to meet performance requirements with high flexibility and dynamics using limited resources. Therefore, employing UAVs in various usage scenarios requires well-designed planning in algorithmic approaches. In this paper, we provide a comprehensive review of recent learning-based algorithmic approaches. We consider possible reward functions and discuss the state-of-the-art algorithms for optimizing the reward functions, including Q-learning, deep Q-learning, multi-armed bandit (MAB), particle swarm optimization (PSO) and satisfaction-based learning algorithms. Unlike other survey papers, we focus on the methodological perspective of the optimization problem, which can be applicable to various UAV-assisted missions on a SAGIN using these algorithms. We simulate users and environments according to real-world scenarios and compare the learning-based and PSO-based methods in terms of throughput, load, fairness, computation time, etc. We also implement and evaluate the 2-dimensional (2D) and 3-dimensional (3D) variations of these algorithms to reflect different deployment cases. Our simulation suggests that the $3$D satisfaction-based learning algorithm outperforms the other approaches for various metrics in most cases. We discuss some open challenges at the end and our findings aim to provide design guidelines for algorithm selections while optimizing the deployment of UAV-assisted SAGINs.