The TORCS simulation provides a range of sensor and track data to the driver bots that can be captured.
Ideally we want to capture sensor data and let the model learn the track. The captured track data can be used as ground truth when evaluating the model's understanding of the track.
Video capture by TORCS is already having a significant performance hit since TORCS is mostly single threaded. Currently 12 frames per second of video capture is not too disruptive. We won't want the sensor & track capture to have only a small impact.
The TORCS simulation provides a range of sensor and track data to the driver bots that can be captured.
Ideally we want to capture sensor data and let the model learn the track. The captured track data can be used as ground truth when evaluating the model's understanding of the track.
Video capture by TORCS is already having a significant performance hit since TORCS is mostly single threaded. Currently 12 frames per second of video capture is not too disruptive. We won't want the sensor & track capture to have only a small impact.