Thanks for this! If I understood your code correctly, running evaluate_scene() for a timestep, where not necessarily all obstacles are actually present will still yield a non-nan value, correct? If yes, I think that’s a very sensible approach to go with!
Do you think this approach is more sensible, or would it be better to exclude obstacles that do not exist in the scenario?
Don’t really know what you mean by that. But what I would do is: if some obstacle is generally present somewhere in the scenario, but not at the particular timestep, that you’re currently evaluating, then simply evaluate that particular simply as if this obstacle was not existing, i.e. only calc criticality with all the other ones.
[…] accommodates scenarios where the ego vehicle might not exist in certain time steps
I’d probably simplify it by presume the ego is guaranteed to be present for all timesteps.
I presume you are using the P_MC measure in your evaluation?
I’m using TTC, WTTC, PET, ALongReq, ALatReq, LongJ, LatJ.
Yes, this is fixed in the latest commit in the develop branch.
thanks, now I only evaluate the obstacles presented in the given time step:
other_veh_ids = [
veh.obstacle_id
for veh in self.sce.obstacles
if veh.obstacle_id is not self.ego_vehicle.obstacle_id
and veh.state_at_time(self.time_step) is not None
]
Thanks! Unfortunately, I cannot see where the issue lies given the measures used, as they should intuitively be lightweight on memory. But I will keep track on it
