Evaluating CUDA Applications under EDF Scheduling for use in Autonomous Vehicles

Advances in Advanced Driver Assistance Systems (ADAS) and autonomous vehicles provide safety-critical control functions such as pedestrian detection, collision warnings, and lane tracking through an array of sensors mounted throughout a vehicle. To support the constant stream of sensor data, real-time systems are needed to periodically process the sensor data to ensure correct functionality of assistive and autonomous control functionality. In this paper, I discuss the use of CUDA, a programmable GPU platform for parallelizing data processing, within the context of ADAS and autonomous vehicles. I also discuss the development of a platform on which to run CUDA applications under Earliest Deadline First (EDF) scheduling, a real-time scheduler, and I evaluate the feasibility of running CUDA applications under EDF scheduling to process sensor data.Bachelor of Scienc