Autonomous driving, how Python drives Formula Racing cars
summary
Team builds autonomous Formula Student cars that run real laps at over 150 km/h and validate Python pipelines on hardware. Perception uses YOLO and LiDAR for cone detection, ROS for prototyping and SLAM for lap memorization; waypoint selection uses triangulation that maximizes minimum angle, then fits a smooth centerline with SciPy bspline/splprep to compute curvature and per-point speeds. Local planner handles immediate targets; global planner exploits full-map straights. Performance hotspots get Numba JIT and sometimes C++ offload (Numba JIT).
AI-written summary · about
chapters (13)
- 0:00Introduction to talk
- 0:25Two perspectives on autonomous driving
- 1:09Breaking down autonomous driving tasks
- 3:39Technologies and Python's role
- 4:44Perception models in autonomous racing
- 7:35Interpreting data for trajectory planning
- 8:41Local trajectory planning demo
- 10:03Global trajectory planning explained
- 13:20Central line path planning
- 15:32Triangulation for waypoints
- 17:00Smooth path with spline fitting
- 19:30Optimizing performance and resources
- 20:58Q&A session begins
related talks
49mBuilding a self driving RC car
35mDingoCar: Perception module Small scale, self-driving car
23mRacing Robocars
35mGetting started with a self-driving RC car
21mAPP4MC RaceCar: A Practical ADAS Demonstrator for Evaluating and Verifying Timing Behavior
45m(Deep) Learning to Fly
26mSelf Driving cars with Opencv
40mTeaching a Car to Drive Itself by Imitation and Imagination