Metric State Space Reinforcement Learning for a Vision-Capable Mobile Robot

TL;DR

This paper introduces a novel metric-based reinforcement learning algorithm for vision-capable mobile robots that learns directly from real-world interaction without manual discretization or models, enabling efficient autonomous control.

Contribution

It extends McCallum's Nearest-Sequence Memory algorithm to general metrics over trajectories, allowing learning in continuous, partially observable environments without manual discretization.

Findings

Successfully applied on a real mobile robot

Learns efficiently with less experience

Handles partial observability and continuous perceptual spaces

Abstract

We address the problem of autonomously learning controllers for vision-capable mobile robots. We extend McCallum's (1995) Nearest-Sequence Memory algorithm to allow for general metrics over state-action trajectories. We demonstrate the feasibility of our approach by successfully running our algorithm on a real mobile robot. The algorithm is novel and unique in that it (a) explores the environment and learns directly on a mobile robot without using a hand-made computer model as an intermediate step, (b) does not require manual discretization of the sensor input space, (c) works in piecewise continuous perceptual spaces, and (d) copes with partial observability. Together this allows learning from much less experience compared to previous methods.