UCT CS Research Document Archive

Evolving Collective Driving Behaviors

Huang, Allen and Geoff Nitschke (2017) Evolving Collective Driving Behaviors. In Proceedings International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2017), pages 1573-1574, Sao Paulo, Brazil.

Full text available as:
PDF - Requires Adobe Acrobat Reader or other PDF viewer.

Abstract

Recently there has been increased research interest in developing autonomous, adaptive control systems of self-driving vehicles. However, there has been little work on synthesizing collective behaviours for autonomous vehicles that must safely interact and coordinate so as traffic throughput on any given road network is maximized. This work uses neuro-evolution to automate car controller design, testing various vehicle sensor configurations and collective driving behaviours resulting from car interactions on roads without constraints of traffic lights, stop signals at intersections or lanes that vehicles must adhere to and thus simulates potential future scenarios where vehicles must drive autonomously without special road infrastructure constraints. Results indicate that neuro-evolution is an effective method for automatically synthesizing collective driving behaviours that are behaviourally robust across a range of vehicle sensor configurations and generalize to different task environments.

EPrint Type:Conference Poster
Subjects:I Computing Methodologies: I.2 ARTIFICIAL INTELLIGENCE
ID Code:1186
Deposited By:Nitschke, Geoff
Deposited On:23 November 2017