Optimal Alarms for Vehicular Collision Detection

TL;DR

This paper reviews and analyzes various real-time collision detection methods for intelligent vehicles, highlighting the effectiveness of Monte Carlo sampling and providing theoretical and empirical insights into their performance.

Contribution

It introduces a framework for optimal alarm selection in vehicle collision detection and compares multiple techniques including Monte Carlo, deterministic, and machine learning approaches.

Findings

Monte Carlo sampling is a robust collision detection method.

Deterministic approximations offer faster alternatives with acceptable accuracy.

Empirical results validate the effectiveness of the surveyed techniques.

Abstract

An important application of intelligent vehicles is advance detection of dangerous events such as collisions. This problem is framed as a problem of optimal alarm choice given predictive models for vehicle location and motion. Techniques for real-time collision detection are surveyed and grouped into three classes: random Monte Carlo sampling, faster deterministic approximations, and machine learning models trained by simulation. Theoretical guarantees on the performance of these collision detection techniques are provided where possible, and empirical analysis is provided for two example scenarios. Results validate Monte Carlo sampling as a robust solution despite its simplicity.