A Modular Agent-Based Communications Framework for Testing Prototype Autonomous Vehicles in Virtual Environments

Abstract

The testing of autonomous vehicles in the automobile industry faces reduced effectiveness owing to their lack of vehicle-to-vehicle communications and component modularity. This paper presents an agent model that is modular and focuses on building a communications framework to solve the drawbacks that autonomous vehicles would otherwise encounter. The agent structures its sensors, actuators, communications mechanism and decision-making system according to a layered architecture using an abstraction layer to connect the modularized components. The communications mechanism of each vehicle shares its internal data to surrounding vehicles to complement the data that they gather through their sensors. The effectiveness of the model was successfully tested and visualized in an urban environment simulation constructed with Unity3D with test scenarios specifically built to challenge autonomous vehicles that would have limited success without the model hereby presented.