Cooperative Traffic Control where Autonomous Cars Meet Human Drivers

Abstract

Co-adaptive system is a close coupling between human and software system cooperating to achieve shared goals. This co-adaption requires adaptive actions to react to unpredictable circumstances. One of the challenges is to deal with uncertainties, and consequently, decision making under uncertainty, which may arise because of the change in the environment, the unpredictable resources, etc. Human behavior does contribute to large amounts of uncertainty. This paper presents an approach for using a simulator as a means of feedback to a human’s decision under uncertainty that can assist human in automated planning to generate cooperative and symbiotic strategy of human and the system to achieve given tasks. To validate the approach, this paper presents a customizable traffic simulator to measure the delays associated with passing vehicles through intersections. The simulator contains AI-based self-adaptive vehicles which can evaluate the quality of traffic at an intersection and change their driving behavior. The human operator from the outside of the system can manipulate the signaling time, the number of predicates per driving rule, number of rules per rule set, learning factor (adaption) etc. to overcome any unexpected traffic. This research proves that our simulator is more efficient than the individual human-operated and automated traffic system and makes a true cooperative traffic example.