I am currently working with the HighD dataset, and I have used the Commonroad dataset converter to convert the dataset into different scenarios. Now, my objective is to find out the vehicles which are undergoing a lane change, and visualize the lane change. I have found out the ids of the vehicles undergoing lane changes from the metadata in the dataset, but the main issue is during the visualization part. Since I want to visualize the lane change, I am trying to visualize the scenario in which it happens. Now, the converter converts the dataset into multiple scenarios, and the same vehicles appear in multiple scenarios. My question is, how can I find out the exact scenario where the lane change occurs, so I don’t have to comb through different scenarios to get the desired visualization?
The raw recording of highD indeed contains numLaneChanges in the **_tracksMeta.csv file. But it does not contain the exact time frames when the lane changes happened. Also, the number of lane changes is also lost during the conversion into CommonRoad scenarios.
What you could do is to access the numLaneChanges in the tracks_meta_df in this function and log the obstacle id of a vehicle if its numLaneChanges > 0 and log the corresponding benchmark_id of this scenario. After the conversion, you can use this log to visualize the corresponding obstacles.
However, this does not prevent that a lane change maneuver might be cut into multiple scenarios. Even if you cut the raw recording into the time horizon of a lane-changing vehicle, you can not ensure that the other vehicles all finish their lane changes.