We study fair multi-objective reinforcement learning in which an agent must learn a policy that simultaneously achieves high reward on multiple dimensions of a vector-valued reward. Motivated by the fair resource allocation literature, we model this as an expected welfare maximization problem, for some non-linear fair welfare function of the vector of long-term cumulative rewards. One canonical example of such a function is the Nash Social Welfare, or geometric mean, the log transform of which is also known as the Proportional Fairness objective. We show that even approximately optimal optimization of the expected Nash Social Welfare is computationally intractable even in the tabular case. Nevertheless, we provide a novel adaptation of Q-learning that combines non-linear scalarized learning updates and non-stationary action selection to learn effective policies for optimizing nonlinear welfare functions. We show that our algorithm is provably convergent, and we demonstrate experimentally that our approach outperforms techniques based on linear scalarization, mixtures of optimal linear scalarizations, or stationary action selection for the Nash Social Welfare Objective.
相關內容
We revisit the estimation bias in policy gradients for the discounted episodic Markov decision process (MDP) from Deep Reinforcement Learning (DRL) perspective. The objective is formulated theoretically as the expected returns discounted over the time horizon. One of the major policy gradient biases is the state distribution shift: the state distribution used to estimate the gradients differs from the theoretical formulation in that it does not take into account the discount factor. Existing discussion of the influence of this bias was limited to the tabular and softmax cases in the literature. Therefore, in this paper, we extend it to the DRL setting where the policy is parameterized and demonstrate how this bias can lead to suboptimal policies theoretically. We then discuss why the empirically inaccurate implementations with shifted state distribution can still be effective. We show that, despite such state distribution shift, the policy gradient estimation bias can be reduced in the following three ways: 1) a small learning rate; 2) an adaptive-learning-rate-based optimizer; and 3) KL regularization. Specifically, we show that a smaller learning rate, or, an adaptive learning rate, such as that used by Adam and RSMProp optimizers, makes the policy optimization robust to the bias. We further draw connections between optimizers and the optimization regularization to show that both the KL and the reverse KL regularization can significantly rectify this bias. Moreover, we provide extensive experiments on continuous control tasks to support our analysis. Our paper sheds light on how successful PG algorithms optimize policies in the DRL setting, and contributes insights into the practical issues in DRL.
Communication in multi-agent reinforcement learning has been drawing attention recently for its significant role in cooperation. However, multi-agent systems may suffer from limitations on communication resources and thus need efficient communication techniques in real-world scenarios. According to the Shannon-Hartley theorem, messages to be transmitted reliably in worse channels require lower entropy. Therefore, we aim to reduce message entropy in multi-agent communication. A fundamental challenge is that the gradients of entropy are either 0 or infinity, disabling gradient-based methods. To handle it, we propose a pseudo gradient descent scheme, which reduces entropy by adjusting the distributions of messages wisely. We conduct experiments on two base communication frameworks with six environment settings and find that our scheme can reduce message entropy by up to 90% with nearly no loss of cooperation performance.
A practical challenge in reinforcement learning are combinatorial action spaces that make planning computationally demanding. For example, in cooperative multi-agent reinforcement learning, a potentially large number of agents jointly optimize a global reward function, which leads to a combinatorial blow-up in the action space by the number of agents. As a minimal requirement, we assume access to an argmax oracle that allows to efficiently compute the greedy policy for any Q-function in the model class. Building on recent work in planning with local access to a simulator and linear function approximation, we propose efficient algorithms for this setting that lead to polynomial compute and query complexity in all relevant problem parameters. For the special case where the feature decomposition is additive, we further improve the bounds and extend the results to the kernelized setting with an efficient algorithm.
In many applications of Reinforcement Learning (RL), it is critically important that the algorithm performs safely, such that instantaneous hard constraints are satisfied at each step, and unsafe states and actions are avoided. However, existing algorithms for ''safe'' RL are often designed under constraints that either require expected cumulative costs to be bounded or assume all states are safe. Thus, such algorithms could violate instantaneous hard constraints and traverse unsafe states (and actions) in practice. Therefore, in this paper, we develop the first near-optimal safe RL algorithm for episodic Markov Decision Processes with unsafe states and actions under instantaneous hard constraints and the linear mixture model. It not only achieves a regret $\tilde{O}(\frac{d H^3 \sqrt{dK}}{\Delta_c})$ that tightly matches the state-of-the-art regret in the setting with only unsafe actions and nearly matches that in the unconstrained setting, but is also safe at each step, where $d$ is the feature-mapping dimension, $K$ is the number of episodes, $H$ is the number of steps in each episode, and $\Delta_c$ is a safety-related parameter. We also provide a lower bound $\tilde{\Omega}(\max\{dH \sqrt{K}, \frac{H}{\Delta_c^2}\})$, which indicates that the dependency on $\Delta_c$ is necessary. Further, both our algorithm design and regret analysis involve several novel ideas, which may be of independent interest.
Many sequential decision-making problems need optimization of different objectives which possibly conflict with each other. The conventional way to deal with a multi-task problem is to establish a scalar objective function based on a linear combination of different objectives. However, for the case of having conflicting objectives with different scales, this method needs a trial-and-error approach to properly find proper weights for the combination. As such, in most cases, this approach cannot guarantee an optimal Pareto solution. In this paper, we develop a single-agent scale-independent multi-objective reinforcement learning on the basis of the Advantage Actor-Critic (A2C) algorithm. A convergence analysis is then done for the devised multi-objective algorithm providing a convergence-in-mean guarantee. We then perform some experiments over a multi-task problem to evaluate the performance of the proposed algorithm. Simulation results show the superiority of developed multi-objective A2C approach against the single-objective algorithm.
In offline reinforcement learning (RL), one detrimental issue to policy learning is the error accumulation of deep Q function in out-of-distribution (OOD) areas. Unfortunately, existing offline RL methods are often over-conservative, inevitably hurting generalization performance outside data distribution. In our study, one interesting observation is that deep Q functions approximate well inside the convex hull of training data. Inspired by this, we propose a new method, DOGE (Distance-sensitive Offline RL with better GEneralization). DOGE marries dataset geometry with deep function approximators in offline RL, and enables exploitation in generalizable OOD areas rather than strictly constraining policy within data distribution. Specifically, DOGE trains a state-conditioned distance function that can be readily plugged into standard actor-critic methods as a policy constraint. Simple yet elegant, our algorithm enjoys better generalization compared to state-of-the-art methods on D4RL benchmarks. Theoretical analysis demonstrates the superiority of our approach to existing methods that are solely based on data distribution or support constraints.
Goal-conditioned reinforcement learning (GCRL) refers to learning general-purpose skills which aim to reach diverse goals. In particular, offline GCRL only requires purely pre-collected datasets to perform training tasks without additional interactions with the environment. Although offline GCRL has become increasingly prevalent and many previous works have demonstrated its empirical success, the theoretical understanding of efficient offline GCRL algorithms is not well established, especially when the state space is huge and the offline dataset only covers the policy we aim to learn. In this paper, we propose a novel provably efficient algorithm (the sample complexity is $\tilde{O}({\rm poly}(1/\epsilon))$ where $\epsilon$ is the desired suboptimality of the learned policy) with general function approximation. Our algorithm only requires nearly minimal assumptions of the dataset (single-policy concentrability) and the function class (realizability). Moreover, our algorithm consists of two uninterleaved optimization steps, which we refer to as $V$-learning and policy learning, and is computationally stable since it does not involve minimax optimization. To the best of our knowledge, this is the first algorithm with general function approximation and single-policy concentrability that is both statistically efficient and computationally stable.
The development of autonomous agents which can interact with other agents to accomplish a given task is a core area of research in artificial intelligence and machine learning. Towards this goal, the Autonomous Agents Research Group develops novel machine learning algorithms for autonomous systems control, with a specific focus on deep reinforcement learning and multi-agent reinforcement learning. Research problems include scalable learning of coordinated agent policies and inter-agent communication; reasoning about the behaviours, goals, and composition of other agents from limited observations; and sample-efficient learning based on intrinsic motivation, curriculum learning, causal inference, and representation learning. This article provides a broad overview of the ongoing research portfolio of the group and discusses open problems for future directions.
The Q-learning algorithm is known to be affected by the maximization bias, i.e. the systematic overestimation of action values, an important issue that has recently received renewed attention. Double Q-learning has been proposed as an efficient algorithm to mitigate this bias. However, this comes at the price of an underestimation of action values, in addition to increased memory requirements and a slower convergence. In this paper, we introduce a new way to address the maximization bias in the form of a "self-correcting algorithm" for approximating the maximum of an expected value. Our method balances the overestimation of the single estimator used in conventional Q-learning and the underestimation of the double estimator used in Double Q-learning. Applying this strategy to Q-learning results in Self-correcting Q-learning. We show theoretically that this new algorithm enjoys the same convergence guarantees as Q-learning while being more accurate. Empirically, it performs better than Double Q-learning in domains with rewards of high variance, and it even attains faster convergence than Q-learning in domains with rewards of zero or low variance. These advantages transfer to a Deep Q Network implementation that we call Self-correcting DQN and which outperforms regular DQN and Double DQN on several tasks in the Atari 2600 domain.
This paper presents a new multi-objective deep reinforcement learning (MODRL) framework based on deep Q-networks. We propose the use of linear and non-linear methods to develop the MODRL framework that includes both single-policy and multi-policy strategies. The experimental results on two benchmark problems including the two-objective deep sea treasure environment and the three-objective mountain car problem indicate that the proposed framework is able to converge to the optimal Pareto solutions effectively. The proposed framework is generic, which allows implementation of different deep reinforcement learning algorithms in different complex environments. This therefore overcomes many difficulties involved with standard multi-objective reinforcement learning (MORL) methods existing in the current literature. The framework creates a platform as a testbed environment to develop methods for solving various problems associated with the current MORL. Details of the framework implementation can be referred to //www.deakin.edu.au/~thanhthi/drl.htm.
小貼士
登錄享
相關主題
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191