Highway Reinforcement Learning

TL;DR

This paper introduces a novel off-policy reinforcement learning method with a highway gate that effectively utilizes distant future information, overcoming underestimation issues of traditional n-step methods and improving performance on delayed reward tasks.

Contribution

A new IS-free, multi-step off-policy RL algorithm with a highway gate that guarantees convergence to the optimal value function regardless of lookahead depth.

Findings

Outperforms existing multi-step off-policy algorithms on delayed reward tasks

Guarantees convergence to the optimal value function for any lookahead depth

Effectively utilizes distant future information through the highway gate mechanism

Abstract

Learning from multi-step off-policy data collected by a set of policies is a core problem of reinforcement learning (RL). Approaches based on importance sampling (IS) often suffer from large variances due to products of IS ratios. Typical IS-free methods, such as $n$-step Q-learning, look ahead for $n$ time steps along the trajectory of actions (where $n$ is called the lookahead depth) and utilize off-policy data directly without any additional adjustment. They work well for proper choices of $n$. We show, however, that such IS-free methods underestimate the optimal value function (VF), especially for large $n$, restricting their capacity to efficiently utilize information from distant future time steps. To overcome this problem, we introduce a novel, IS-free, multi-step off-policy method that avoids the underestimation issue and converges to the optimal VF. At its core lies a simple but non-trivial \emph{highway gate}, which controls the information flow from the distant future by comparing it to a threshold. The highway gate guarantees convergence to the optimal VF for arbitrary $n$ and arbitrary behavioral policies. It gives rise to a novel family of off-policy RL algorithms that safely learn even when $n$ is very large, facilitating rapid credit assignment from the far future to the past. On tasks with greatly delayed rewards, including video games where the reward is given only at the end of the game, our new methods outperform many existing multi-step off-policy algorithms.