README.md

Track Driving Simulator in MATLAB

• This simulator is intended for developing algorithms for autonomous cars.
• Can be easily tuned to implement reinforcement learning or inverse reinforcement learning algorithms.

Collecting Demonstrations (Run 'run_track01_manualcollect')

Check Collected Demonstrations (Run 'run_track02_checkcollected')

Train Controller with DMRL (Run 'run_track03_trainDMRL')

Test on More Complex Tracks (Run 'run_track04_testDMRL')

• For example, it was used for implementing an imitation learning algorithm named density matching reward learning.
• YouTube Link

Track Driving Simulation

• The blue car is controlled by the keyboard and the red cars autonomously maintain each lanes.
• YouTube Link

Main Features

1. Control the car with keyboard inputs.
2. Compose tracks with a block-building like manner.
3. Lane width, length, number of lane, and the number of segments can also be modified.
4. Obtainable features are

• Lane deviate distance / degree
• Distances from the closest right, center, and left cars
• Geodesic distance from the start lane
• Which lane and segment the current car is located at.

5. Traffic Lights are added.
6. Range finder sensors are added.

Contact

sungjoon.choi@cpslab.snu.ac.kr