Using 2 frames of LIDAR to get acceleration information

[All-Ki]

Wouldnt using the previous and current 'frames' of LIDAR make the network able to extract speed information ?
That wouldnt be cheating per se since us humans also use our memory !

[yannbouteiller]

That is already what the environment does by default, it uses the 4 most recent screenshots so that the policy can infer its first order moments (speed/acceleration...) :)

[All-Ki]

Oh okay , i shouldve read the code better ^^