Date of Defense

4-21-2026

Date of Graduation

5-2026

Department

Computer Science

First Advisor

Wuwei Shen

Second Advisor

Steven Carr

Abstract

This capstone sought out a method to make a safer self-driving vehicle by using the Carla Simulator.  The Carla simulator has a Python API which can be used to create and modify actors in the environment.  Once the car was able to drive on a route without difficulty, the exploration began to use sensors to increase safety.  An obstacle detector, collision sensor, and the world’s map were used to detect when obstacles were entering the path.  By utilizing a MAPE (Monitor, Analyzer, Planner, Executor) structure, the vehicle reacted directly to the environment to decide it’s next controls.  The monitor portion of the structure takes in all of the information from the environment and sends it to the next stage in the structure.  The analyzer then interprets the information given to decide what parts the vehicle will need to react to or utilize in control determination.  The planner takes the information from the analyzer and uses it to decide the brake, throttle, and steering angle the vehicle should implement.  The executor then directly applies those controls to the vehicle.  This process occurs every cycle or tick of the program.  To increase safety, a Control Barrier Function was implemented.  This decides the safest throttle and brake for the car inside of the analyzer and planner step.  It uses the distance between the car and an obstacle as well as the car’s given speed to calculate the best throttle for the car.  This implementation was then tested against several scenarios such as red traffic lights, pedestrians walking on a crosswalk, and a parked car in the road.  Once the program was fully created, testing was completed using unit testing, metamorphic testing, and mutation testing.  Unit testing and metamorphic testing techniques were chosen to ensure that all of the code created was doing exactly what it should be.  Mutation testing was also completed to ensure that the unit testing completed was accurate and able to tell us our program was accurate.   The final result of this capstone process was a program which shows a safer self-driving vehicle which can be seen through the Carla simulator.  This program has been fully tested to ensure that it is running with as few errors as possible.

Comments

Co-authored with:

Arin Briody

Calypso Harden

Abby Wheaton

Access Setting

Honors Thesis-Restricted

Download

Share

COinS