Randomized Strategy for Walking in Streets for a Simple Robot

TL;DR

This paper introduces a randomized search strategy for a simple robot with minimal sensing to efficiently find a target in an unknown street, achieving near-optimal expected travel distance without scene memory.

Contribution

It presents a novel randomized search method based on a deterministic strategy, with proven bounds on expected travel distance, for robots with minimal sensing capabilities.

Findings

Expected travel distance is at most 5.33 times the shortest path.

The strategy does not require scene memory.

Proves optimality bounds for the randomized approach.

Abstract

We consider the problem of walking in an unknown street, for a robot that has a minimal sensing capability. The robot is equipped with a sensor that only detects the discontinuities in depth information (gaps) and can locate the target point as enters in its visibility region. First, we propose an online deterministic search strategy that generates an optimal search path for the simple robot to reach the target t, starting from s. In contrast with previously known research, the path is designed without memorizing any portion of the scene has seen so far. Then, we present a randomized search strategy, based on the deterministic strategy. We prove that the expected distance traveled by the robot is at most a 5.33 times longer than the shortest path to reach the target.