Intersection Simulation and Path Estimation

Abstract

As autonomous vehicles begin to move towards full autonomy, the controllers and software within them are becoming incredibly more complex to deal with any plausible scenario. Automotive manufacturers must balance the need for safety with the customers' desire for performance and features. A robust set of tools is a necessity to develop vehicle control protocols and navigation strategies. Vehicle to everything communication protocols and path planning are two aspects of autonomous vehicles that need a large amount of development effort. The MathWorks has put a great amount of effort in developing a robust simulation tool for autonomous vehicles. However, it currently lacks a method to develop V2X communication and path routing. In this thesis, I developed an extension for the Mathworks Simulink autonomous driving toolbox to incorporate graph-based path planning and vehicle to vehicle communication. The navigation system models each road using standard civil engineering techniques, to calculate the intersection points and bounding areas for regions of interest. Based on these regions, a directed graph is created to aid in calculating the shortest path. The navigation system also provides a redundant method for path planning for poorly marked areas and intersections. The vehicle to vehicle communication system emulates the 802.11p protocol and deals with practical challenges such as latency to provide developers with a realistic environment in which to develop vehicle communication protocols. The final result is a simulation where multiple vehicles drive safely and efficiently throughout a city network, sending messages at regions of interest and follow computed paths to their desired destinations.