In the challenge announcement it is stated that the past observation horizon is 2.0s. Also, in the default_simulation.yaml config, the simulation history buffer duration is set to 2.0 seconds. Thereby I would assume, that a planner can use up to 2.0s of past observations in simulation. However, calculating features from simulation using a past_trajectory_sampling with a time_horizon of 2.0s is not possible, as the actually available history is slightly below 2.0s (depending on the scenario between 1.99s and 1.90s). In consequence, simulation will fail as the past trajectory exceeds the available history.
Steps To Reproduce
Steps to reproduce the behavior:
Run a simulation setting past_trajectory_sampling.time_horizon=2.0
patk-motional
commented
1 year ago
Describe the bug
In the challenge announcement it is stated that the past observation horizon is 2.0s. Also, in the
default_simulation.yamlconfig, the simulation history buffer duration is set to 2.0 seconds. Thereby I would assume, that a planner can use up to 2.0s of past observations in simulation. However, calculating features from simulation using apast_trajectory_samplingwith atime_horizonof 2.0s is not possible, as the actually available history is slightly below 2.0s (depending on the scenario between 1.99s and 1.90s). In consequence, simulation will fail as the past trajectory exceeds the available history.Steps To Reproduce
Steps to reproduce the behavior:
past_trajectory_sampling.time_horizon=2.0