Iterative combinatorial auctions are widely used in high stakes settings such as spectrum auctions. Such auctions can be hard to analyze, making it difficult for bidders to determine how to behave and for designers to optimize auction rules to ensure desirable outcomes such as high revenue or welfare. In this paper, we investigate whether multi-agent reinforcement learning (MARL) algorithms can be used to understand iterative combinatorial auctions, given that these algorithms have recently shown empirical success in several other domains. We find that MARL can indeed benefit auction analysis, but that deploying it effectively is nontrivial. We begin by describing modelling decisions that keep the resulting game tractable without sacrificing important features such as imperfect information or asymmetry between bidders. We also discuss how to navigate pitfalls of various MARL algorithms, how to overcome challenges in verifying convergence, and how to generate and interpret multiple equilibria. We illustrate the promise of our resulting approach by using it to evaluate a specific rule change to a clock auction, finding substantially different auction outcomes due to complex changes in bidders' behavior.
Large Language Models (LLMs) have demonstrated proficiency in utilizing various tools by coding, yet they face limitations in handling intricate logic and precise control. In embodied tasks, high-level planning is amenable to direct coding, while low-level actions often necessitate task-specific refinement, such as Reinforcement Learning (RL). To seamlessly integrate both modalities, we introduce a two-level hierarchical framework, RL-GPT, comprising a slow agent and a fast agent. The slow agent analyzes actions suitable for coding, while the fast agent executes coding tasks. This decomposition effectively focuses each agent on specific tasks, proving highly efficient within our pipeline. Our approach outperforms traditional RL methods and existing GPT agents, demonstrating superior efficiency. In the Minecraft game, it rapidly obtains diamonds within a single day on an RTX3090. Additionally, it achieves SOTA performance across all designated MineDojo tasks.
An agent employing reinforcement learning takes inputs (state variables) from an environment and performs actions that affect the environment in order to achieve some objective. Rewards (positive or negative) guide the agent toward improved future actions. This paper builds on prior clustering neural network research by constructing an agent with biologically plausible neo-Hebbian three-factor synaptic learning rules, with a reward signal as the third factor (in addition to pre- and post-synaptic spikes). The classic cart-pole problem (balancing an inverted pendulum) is used as a running example throughout the exposition. Simulation results demonstrate the efficacy of the approach, and the proposed method may eventually serve as a low-level component of a more general method.
Diverse studies in systems neuroscience begin with extended periods of curriculum training known as `shaping' procedures. These involve progressively studying component parts of more complex tasks, and can make the difference between learning a task quickly, slowly or not at all. Despite the importance of shaping to the acquisition of complex tasks, there is as yet no theory that can help guide the design of shaping procedures, or more fundamentally, provide insight into its key role in learning. Modern deep reinforcement learning systems might implicitly learn compositional primitives within their multilayer policy networks. Inspired by these models, we propose and analyse a model of deep policy gradient learning of simple compositional reinforcement learning tasks. Using the tools of statistical physics, we solve for exact learning dynamics and characterise different learning strategies including primitives pre-training, in which task primitives are studied individually before learning compositional tasks. We find a complex interplay between task complexity and the efficacy of shaping strategies. Overall, our theory provides an analytical understanding of the benefits of shaping in a class of compositional tasks and a quantitative account of how training protocols can disclose useful task primitives, ultimately yielding faster and more robust learning.
In the realm of reinforcement learning (RL), accounting for risk is crucial for making decisions under uncertainty, particularly in applications where safety and reliability are paramount. In this paper, we introduce a general framework on Risk-Sensitive Distributional Reinforcement Learning (RS-DisRL), with static Lipschitz Risk Measures (LRM) and general function approximation. Our framework covers a broad class of risk-sensitive RL, and facilitates analysis of the impact of estimation functions on the effectiveness of RSRL strategies and evaluation of their sample complexity. We design two innovative meta-algorithms: \texttt{RS-DisRL-M}, a model-based strategy for model-based function approximation, and \texttt{RS-DisRL-V}, a model-free approach for general value function approximation. With our novel estimation techniques via Least Squares Regression (LSR) and Maximum Likelihood Estimation (MLE) in distributional RL with augmented Markov Decision Process (MDP), we derive the first $\widetilde{\mathcal{O}}(\sqrt{K})$ dependency of the regret upper bound for RSRL with static LRM, marking a pioneering contribution towards statistically efficient algorithms in this domain.
This work pioneers regret analysis of risk-sensitive reinforcement learning in partially observable environments with hindsight observation, addressing a gap in theoretical exploration. We introduce a novel formulation that integrates hindsight observations into a Partially Observable Markov Decision Process (POMDP) framework, where the goal is to optimize accumulated reward under the entropic risk measure. We develop the first provably efficient RL algorithm tailored for this setting. We also prove by rigorous analysis that our algorithm achieves polynomial regret $\tilde{O}\left(\frac{e^{|{\gamma}|H}-1}{|{\gamma}|H}H^2\sqrt{KHS^2OA}\right)$, which outperforms or matches existing upper bounds when the model degenerates to risk-neutral or fully observable settings. We adopt the method of change-of-measure and develop a novel analytical tool of beta vectors to streamline mathematical derivations. These techniques are of particular interest to the theoretical study of reinforcement learning.
The integration of Unmanned Aerial Vehicles (UAVs) into Search and Rescue (SAR) missions presents a promising avenue for enhancing operational efficiency and effectiveness. However, the success of these missions is not solely dependent on the technical capabilities of the drones but also on their acceptance and interaction with humans on the ground. This paper explores the effect of human-centric factor in UAV trajectory planning for SAR missions. We introduce a novel approach based on the reinforcement learning augmented with Analytic Hierarchy Process and novel similarity-based experience replay to optimize UAV trajectories, balancing operational objectives with human comfort and safety considerations. Additionally, through a comprehensive survey, we investigate the impact of gender cues and anthropomorphism in UAV design on public acceptance and trust, revealing significant implications for drone interaction strategies in SAR. Our contributions include (1) a reinforcement learning framework for UAV trajectory planning that dynamically integrates multi-objective considerations, (2) an analysis of human perceptions towards gendered and anthropomorphized drones in SAR contexts, and (3) the application of similarity-based experience replay for enhanced learning efficiency in complex SAR scenarios. The findings offer valuable insights into designing UAV systems that are not only technically proficient but also aligned with human-centric values.
Preference-based reinforcement learning (PbRL) aligns a robot behavior with human preferences via a reward function learned from binary feedback over agent behaviors. We show that dynamics-aware reward functions improve the sample efficiency of PbRL by an order of magnitude. In our experiments we iterate between: (1) learning a dynamics-aware state-action representation (z^{sa}) via a self-supervised temporal consistency task, and (2) bootstrapping the preference-based reward function from (z^{sa}), which results in faster policy learning and better final policy performance. For example, on quadruped-walk, walker-walk, and cheetah-run, with 50 preference labels we achieve the same performance as existing approaches with 500 preference labels, and we recover 83\% and 66\% of ground truth reward policy performance versus only 38\% and 21\%. The performance gains demonstrate the benefits of explicitly learning a dynamics-aware reward model. Repo: \texttt{https://github.com/apple/ml-reed}.
Low Earth Orbit (LEO) satellite-to-handheld connections herald a new era in satellite communications. Space-Division Multiple Access (SDMA) precoding is a method that mitigates interference among satellite beams, boosting spectral efficiency. While optimal SDMA precoding solutions have been proposed for ideal channel knowledge in various scenarios, addressing robust precoding with imperfect channel information has primarily been limited to simplified models. However, these models might not capture the complexity of LEO satellite applications. We use the Soft Actor-Critic (SAC) deep Reinforcement Learning (RL) method to learn robust precoding strategies without the need for explicit insights into the system conditions and imperfections. Our results show flexibility to adapt to arbitrary system configurations while performing strongly in terms of achievable rate and robustness to disruptive influences compared to analytical benchmark precoders.
Benchmarks play a crucial role in the development and analysis of reinforcement learning (RL) algorithms. We identify that existing benchmarks used for research into open-ended learning fall into one of two categories. Either they are too slow for meaningful research to be performed without enormous computational resources, like Crafter, NetHack and Minecraft, or they are not complex enough to pose a significant challenge, like Minigrid and Procgen. To remedy this, we first present Craftax-Classic: a ground-up rewrite of Crafter in JAX that runs up to 250x faster than the Python-native original. A run of PPO using 1 billion environment interactions finishes in under an hour using only a single GPU and averages 90% of the optimal reward. To provide a more compelling challenge we present the main Craftax benchmark, a significant extension of the Crafter mechanics with elements inspired from NetHack. Solving Craftax requires deep exploration, long term planning and memory, as well as continual adaptation to novel situations as more of the world is discovered. We show that existing methods including global and episodic exploration, as well as unsupervised environment design fail to make material progress on the benchmark. We believe that Craftax can for the first time allow researchers to experiment in a complex, open-ended environment with limited computational resources.
This paper introduces a Multi-Agent Deep Reinforcement Learning (MA-DRL) approach for routing in Low Earth Orbit Satellite Constellations (LSatCs). Each satellite is an independent decision-making agent with a partial knowledge of the environment, and supported by feedback received from the nearby agents. Building on our previous work that introduced a Q-routing solution, the contribution of this paper is to extend it to a deep learning framework able to quickly adapt to the network and traffic changes, and based on two phases: (1) An offline exploration learning phase that relies on a global Deep Neural Network (DNN) to learn the optimal paths at each possible position and congestion level; (2) An online exploitation phase with local, on-board, pre-trained DNNs. Results show that MA-DRL efficiently learns optimal routes offline that are then loaded for an efficient distributed routing online.
Online reinforcement learning (RL) algorithms offer great potential for personalizing treatment for participants in clinical trials. However, deploying an online, autonomous algorithm in the high-stakes healthcare setting makes quality control and data quality especially difficult to achieve. This paper proposes algorithm fidelity as a critical requirement for deploying online RL algorithms in clinical trials. It emphasizes the responsibility of the algorithm to (1) safeguard participants and (2) preserve the scientific utility of the data for post-trial analyses. We also present a framework for pre-deployment planning and real-time monitoring to help algorithm developers and clinical researchers ensure algorithm fidelity. To illustrate our framework's practical application, we present real-world examples from the Oralytics clinical trial. Since Spring 2023, this trial successfully deployed an autonomous, online RL algorithm to personalize behavioral interventions for participants at risk for dental disease.
Typical models of learning assume incremental estimation of continuously-varying decision variables like expected rewards. However, this class of models fails to capture more idiosyncratic, discrete heuristics and strategies that people and animals appear to exhibit. Despite recent advances in strategy discovery using tools like recurrent networks that generalize the classic models, the resulting strategies are often onerous to interpret, making connections to cognition difficult to establish. We use Bayesian program induction to discover strategies implemented by programs, letting the simplicity of strategies trade off against their effectiveness. Focusing on bandit tasks, we find strategies that are difficult or unexpected with classical incremental learning, like asymmetric learning from rewarded and unrewarded trials, adaptive horizon-dependent random exploration, and discrete state switching.
Preference-based Reinforcement Learning (PbRL) circumvents the need for reward engineering by harnessing human preferences as the reward signal. However, current PbRL methods excessively depend on high-quality feedback from domain experts, which results in a lack of robustness. In this paper, we present RIME, a robust PbRL algorithm for effective reward learning from noisy preferences. Our method utilizes a sample selection-based discriminator to dynamically filter out noise and ensure robust training. To counteract the cumulative error stemming from incorrect selection, we suggest a warm start for the reward model, which additionally bridges the performance gap during the transition from pre-training to online training in PbRL. Our experiments on robotic manipulation and locomotion tasks demonstrate that RIME significantly enhances the robustness of the state-of-the-art PbRL method. Code is available at https://github.com/CJReinforce/RIME_ICML2024.
Can we pre-train a generalist agent from a large amount of unlabeled offline trajectories such that it can be immediately adapted to any new downstream tasks in a zero-shot manner? In this work, we present a functional reward encoding (FRE) as a general, scalable solution to this zero-shot RL problem. Our main idea is to learn functional representations of any arbitrary tasks by encoding their state-reward samples using a transformer-based variational auto-encoder. This functional encoding not only enables the pre-training of an agent from a wide diversity of general unsupervised reward functions, but also provides a way to solve any new downstream tasks in a zero-shot manner, given a small number of reward-annotated samples. We empirically show that FRE agents trained on diverse random unsupervised reward functions can generalize to solve novel tasks in a range of simulated robotic benchmarks, often outperforming previous zero-shot RL and offline RL methods. Code for this project is provided at: https://github.com/kvfrans/fre
In reinforcement learning algorithms, the hyperparameters tuning method refers to choosing the optimal parameters that may increase the overall performance. Manual or random hyperparameter tuning methods can lead to different results in the reinforcement learning algorithms. In this paper, we propose a new method called QF-tuner for automatic hyperparameter tuning in the Q learning algorithm using the FOX optimization algorithm (FOX). Furthermore, a new objective function has been employed within FOX that prioritizes reward over learning error and time. QF tuner starts by running the FOX and tries to minimize the fitness value derived from observations at each iteration by executing the Q-learning algorithm. The proposed method has been evaluated using two control tasks from the OpenAI Gym: CartPole and FrozenLake. The empirical results indicate that the QF-tuner outperforms other optimization algorithms, such as particle swarm optimization (PSO), bees algorithm (BA), genetic algorithms (GA), and the random method. However, on the FrozenLake task, the QF-tuner increased rewards by 36% and reduced learning time by 26%, while on the CartPole task, it increased rewards by 57% and reduced learning time by 20%. Thus, the QF-tuner is an essential method for hyperparameter tuning in Q-learning algorithms, enabling more effective solutions to control task problems.
In recent years, deep reinforcement learning has emerged as a technique to solve closed-loop flow control problems. Employing simulation-based environments in reinforcement learning enables a priori end-to-end optimization of the control system, provides a virtual testbed for safety-critical control applications, and allows to gain a deep understanding of the control mechanisms. While reinforcement learning has been applied successfully in a number of rather simple flow control benchmarks, a major bottleneck toward real-world applications is the high computational cost and turnaround time of flow simulations. In this contribution, we demonstrate the benefits of model-based reinforcement learning for flow control applications. Specifically, we optimize the policy by alternating between trajectories sampled from flow simulations and trajectories sampled from an ensemble of environment models. The model-based learning reduces the overall training time by up to $85\%$ for the fluidic pinball test case. Even larger savings are expected for more demanding flow simulations.
In the rapidly evolving domain of electrical power systems, the Volt-VAR optimization (VVO) is increasingly critical, especially with the burgeoning integration of renewable energy sources. Traditional approaches to learning-based VVO in expansive and dynamically changing power systems are often hindered by computational complexities. To address this challenge, our research presents a novel framework that harnesses the potential of Deep Reinforcement Learning (DRL), specifically utilizing the Importance Weighted Actor-Learner Architecture (IMPALA) algorithm, executed on the RAY platform. This framework, built upon RLlib-an industry-standard in Reinforcement Learning-ingeniously capitalizes on the distributed computing capabilities and advanced hyperparameter tuning offered by RAY. This design significantly expedites the exploration and exploitation phases in the VVO solution space. Our empirical results demonstrate that our approach not only surpasses existing DRL methods in achieving superior reward outcomes but also manifests a remarkable tenfold reduction in computational requirements. The integration of our DRL agent with the RAY platform facilitates the creation of RLlib-IMPALA, a novel framework that efficiently uses RAY's resources to improve system adaptability and control. RLlib-IMPALA leverages RAY's toolkit to enhance analytical capabilities and significantly speeds up training to become more than 10 times faster than other state-of-the-art DRL methods.
This paper analyzes multi-step TD-learning algorithms within the `deadly triad' scenario, characterized by linear function approximation, off-policy learning, and bootstrapping. In particular, we prove that n-step TD-learning algorithms converge to a solution as the sampling horizon n increases sufficiently. The paper is divided into two parts. In the first part, we comprehensively examine the fundamental properties of their model-based deterministic counterparts, including projected value iteration, gradient descent algorithms, and the control theoretic approach, which can be viewed as prototype deterministic algorithms whose analysis plays a pivotal role in understanding and developing their model-free reinforcement learning counterparts. In particular, we prove that these algorithms converge to meaningful solutions when n is sufficiently large. Based on these findings, two n-step TD-learning algorithms are proposed and analyzed, which can be seen as the model-free reinforcement learning counterparts of the gradient and control theoretic algorithms.
Diffusion models excel at modeling complex data distributions, including those of images, proteins, and small molecules. However, in many cases, our goal is to model parts of the distribution that maximize certain properties: for example, we may want to generate images with high aesthetic quality, or molecules with high bioactivity. It is natural to frame this as a reinforcement learning (RL) problem, in which the objective is to fine-tune a diffusion model to maximize a reward function that corresponds to some property. Even with access to online queries of the ground-truth reward function, efficiently discovering high-reward samples can be challenging: they might have a low probability in the initial distribution, and there might be many infeasible samples that do not even have a well-defined reward (e.g., unnatural images or physically impossible molecules). In this work, we propose a novel reinforcement learning procedure that efficiently explores on the manifold of feasible samples. We present a theoretical analysis providing a regret guarantee, as well as empirical validation across three domains: images, biological sequences, and molecules.