Explicable Reward Design for Reinforcement Learning Agents

We study the design of explicable reward functions for a reinforcement learning agent while guaranteeing that an optimal policy induced by the function belongs to a set of target policies. By being explicable, we seek to capture two properties: (a) informativeness so that the rewards speed up the agent's convergence, and (b) sparseness as a proxy for ease of interpretability of the rewards. The key challenge is that higher informativeness typically requires dense rewards for many learning tasks, and existing techniques do not allow one to balance these two properties appropriately. In this paper, we investigate the problem from the perspective of discrete optimization and introduce a novel framework, EXPRD, to design explicable reward functions. EXPRD builds upon an informativeness criterion that captures the (sub-)optimality of target policies at different time horizons in terms of actions taken from any given starting state. We provide a mathematical analysis of EXPRD, and show its connections to existing reward design techniques, including potential-based reward shaping. Experimental results on two navigation tasks demonstrate the effectiveness of EXPRD in designing explicable reward functions. In addition, for any state s ∈ S, globally optimal actions Π * s ⊆ A under R are also myopically optimal under R PBRS since δ π 0 (s, a) = δ * ∞ (s, a) for all π ∈ Π * [3, 8] -this leads to a dramatic speed-up in the learning process. However, the potentialbased reward shaping produces dense reward function which is less interpretable (see Section 4). * and the ∞-step optimality gaps by δ * ∞ ; the quantities defined corresponding to R := R are denoted by a widehat, e.g., the optimal policy set by Π * .