In this post, I'll introduce a reinforcement learning (RL) algorithm based on an "alternative" paradigm: divide and conquer We can do Reinforcement Learning (RL) based on divide and conquer, instead of temporal difference (TD) learning. There are two classes of algorithms in RL: on-policy RL and off-policy RL. On-policy RL means we can use fresh data collected by the current policy. In other words, we have to throw away old data each time we update the policy. Algorithms like PPO and GRPO (and policy gradient methods in general) belong to this category.

We draw an analogy between static friction in classical mechanics and extrapolation error in off-policy RL, and use it to formulate a constraint that prevents the policy from drifting toward unsupported actions. In this study, we present Frictional Q-learning, a deep reinforcement learning algorithm for continuous control, which extends batch-constrained reinforcement learning. Our algorithm constrains the agent's action space to encourage behavior similar to that in the replay buffer, while maintaining a distance from the manifold of the orthonormal action space. The constraint preserves the simplicity of batch-constrained, and provides an intuitive physical interpretation of extrapolation error. Empirically, we further demonstrate that our algorithm is robustly trained and achieves competitive performance across standard continuous control benchmarks.

Conventional off-policy reinforcement learning (RL) focuses on maximizing the expected return of scalar rewards. Distributional RL (DRL), in contrast, studies the distribution of returns with the distributional Bellman operator in a Euclidean space, leading to highly flexible choices for utility. This paper establishes robust theoretical foundations for DRL. We prove the contraction property of the Bellman operator even when the reward space is an infinite-dimensional separable Banach space. Furthermore, we demonstrate that the behavior of high- or infinite-dimensional returns can be effectively approximated using a lower-dimensional Euclidean space. Leveraging these theoretical insights, we propose a novel DRL algorithm that tackles problems which have been previously intractable using conventional reinforcement learning approaches.

We propose policy gradient algorithms for solving a risk-sensitive reinforcement learning (RL) problem in on-policy as well as off-policy settings. We consider episodic Markov decision processes, and model the risk using the broad class of smooth risk measures of the cumulative discounted reward. We propose two template policy gradient algorithms that optimize a smooth risk measure in on-policy and off-policy RL settings, respectively. We derive non-asymptotic bounds that quantify the rate of convergence of our proposed algorithms to a stationary point of the smooth risk measure. As special cases, we establish that our algorithms apply to optimization of mean-variance and distortion risk measures, respectively.

Multi-task reinforcement learning (RL) aims to simultaneously learn policies for solving many tasks. Several prior works have found that relabeling past experience with different reward functions can improve sample efficiency. Relabeling methods typically ask: if, in hindsight, we assume that our experience was optimal for some task, for what task was it optimal? In this paper, we show that hindsight relabeling is inverse RL, an observation that suggests that we can use inverse RL in tandem for RL algorithms to efficiently solve many tasks. We use this idea to generalize goal-relabeling techniques from prior work to arbitrary classes of tasks. Our experiments confirm that relabeling data using inverse RL accelerates learning in general multi-task settings, including goal-reaching, domains with discrete sets of rewards, and those with linear reward functions.