Lifetime policy reuse and the importance of task capacity

TL;DR

This paper introduces Lifetime Policy Reuse, a model-agnostic algorithm that optimizes a fixed set of policies for lifelong reinforcement learning, and proposes task capacity as a measure of how many tasks a policy can handle, improving transfer learning.

Contribution

The paper presents a novel lifetime policy reuse algorithm and introduces task capacity as a new metric, enhancing lifelong reinforcement learning efficiency and scalability.

Findings

Lifetime Policy Reuse outperforms existing methods in 18-task Pacman domain.

Task capacity effectively predicts the maximum number of tasks a policy can solve.

Pre-selection based on task capacity improves performance in large task sets.

Abstract

A long-standing challenge in artificial intelligence is lifelong reinforcement learning, where learners are given many tasks in sequence and must transfer knowledge between tasks while avoiding catastrophic forgetting. Policy reuse and other multi-policy reinforcement learning techniques can learn multiple tasks but may generate many policies. This paper presents two novel contributions, namely 1) Lifetime Policy Reuse, a model-agnostic policy reuse algorithm that avoids generating many policies by optimising a fixed number of near-optimal policies through a combination of policy optimisation and adaptive policy selection; and 2) the task capacity, a measure for the maximal number of tasks that a policy can accurately solve. Comparing two state-of-the-art base-learners, the results demonstrate the importance of Lifetime Policy Reuse and task capacity based pre-selection on an 18-task partially observable Pacman domain and a Cartpole domain of up to 125 tasks.