The aim of the project is to simulate how an autonomous vehicle would respond to various traffic scenarios that are unique to the University of Cape Town campus.
Due to its location on the base of the Devil's Peak Mountain, the campus is swathed with many steep and narrow roads as well as a myriad of smaller pathways and road networks. There are various vehicle types ranging from small sedans to large shuttle busses used for transporting students across the campus.
A 3D model of the campus was created with Blender which will be imported into MathLab's RoadRunner for road network and lane line generation then finally imported into Carla autonomous vehicle simulator where optimal ML algorithms will be applied to train an autonomous agent to navigate the roads of UCT. A web browser visualization of the agent carrying out the simulation will be available once the project is complete.
Scenarios to simulate:
- UCT Jammie Stairs Parking Lot autonomous valet parking
- M3 Highway lane change
- Jammie North & South Bus Stop navigation
- Drive across campus with pedestrians
Blender was used to create an accurate 3D model of the university campus. A rendering of the student parking lot by the Jammie stairs can be seen below
RoadRunner was used to create roads netowrks and street signs to be exported to the Carla simulator. A rendering of the student parking lot with a road network on the southern side can be seen below
A web browser visualization of the agent carrying out the simulation similar to the one in the image below will be available once the project is complete
Looking forward to sharing this project with everyone ❤️