Kana Kim
< Back to selected work
2024 Competition / Vehicle Control

Level 4 High-Speed Autonomous Driving

Two-hour driverless run in the rain on a high-speed track (Minister’s Award)

  • GitHub
  • Demo
Adaptive PID (jerk-aware)Pure PursuitFrenet PlanningLate FusionHD MapCAN (Avante CN7)ROS50Hz Control Loop
Level 4 High-Speed Autonomous Driving Overview

Overview

Implemented lane-specific driving strategies, curvature-based speed control, and a 50Hz Adaptive PID / Pure Pursuit controller to achieve a 2-hour driverless run at 100km/h (~129.2km) in the rain on the KIAPI high-speed circuit, finishing runner-up.

Level 4 High-Speed Autonomous Driving Background

Background

A competition to verify the highest level of autonomy. The four success factors are safety (emergency response), real-time (optimal path & speed), adaptability (algorithm generality), and durability (2-hour driving) — the target was a high-speed system meeting all four.

Level 4 High-Speed Autonomous Driving Approach 1

Approach 1

Vehicle & sensor configuration — LiDAR (OT-128), RTK GPS (MGI-2000), camera (FLIR), main PC (NUVO), and sub PC (Jetson Orin AGX) integrated on an Avante CN7.

Level 4 High-Speed Autonomous Driving Approach 2

Approach 2

A pipeline from Sensing (Camera·LiDAR·GPS·IMU) → Perception (detection/clustering for ACC and overtaking) → Decision (HD-map Frenet path & target speed) → Control (targets → CAN) → driving, with fail-safe on detected abnormalities.

Level 4 High-Speed Autonomous Driving Result

Result

Finished runner-up (2nd) at the 2024 University Autonomous Driving Competition, performing overtaking, speed adjustment, stopped-vehicle avoidance, and curve deceleration, and safely passing an accident zone.