Closed TrackerSB closed 5 years ago
The performance improvement was only due to missing synchronization which causes #12.
Since popping/putting from/into the queue blocks the processes a queue per process may decrease blocking calls.
This aspect is not dealt with anymore.
The generation of state predictions does not fully utilize the CPU even allowing multiple threads. The generation needs to be speeded up to generate 50 time steps having a better resolution in the allowed steering angle.