Closed etiwal closed 3 years ago
@etiwal can you confirm the issue is fixed? I think the data structure was filled incorrectly with the new state given the previous input. Same holds for the cost computed with the new state and previous input.
In a first analysis, it does not seem to have a great effect but it is definitely a bug.
Hi Giuseppe
Not sure if this is an implementation issue or if it is just your convention and it makes all sense. From my perspective the time of your rollouts is shiftet by one. --> The 1st state x1 (which would be the dynamics stepped by the first input) is at t0_internal, but it should be at t0_internal+1*dt).
Cheers!