[Proposal] Move the motion planning modules after the scenario selector

[ismetatabay]

Motivation

With the current architecture, the motion planning modules don't work for parking scenarios and can be dangerous during reverse driving. This situation limits the functionality of other potential scenarios for future upgrades. For example, if we want to utilize the third scenario in scenario_selector, which is a pre-defined scenario that could have been recorded previously, the ego vehicle will hit any obstacles that occur along this defined trajectory while following this trajectory.

Proposed Approach

Changing the launch structure to apply motion planning modules after the scenario_selector package output will only modify the velocity points of the reverse driving trajectory.

Target Motion Planning Packages:

• obstacle_velocity_limiter
• obstacle_stop_planner or obstacle_cruise_planner

Current Architecture	Proposed Architecture

Discussion

Please let me know if you have any concerns or suggestions regarding this proposal, thanks!

@xmfcx, @zulfaqar-azmi-t4, @maxime-clem, @isamu-takagi, @satoshi-ota, @mehmetdogru

[maxime-clem]

We had some internal discussion at TIER IV and we agree with the proposal.
Do you already know how you will implement the change ? We could think of 2 possible ways:

• Move the scenario_selector to the motion_planning.launch.xml to run before the stop planner (obstacle_cruise_planner/obstacle_stop_planner).
• Move the stop planner to the scenario_planning.launch.xml to run after the scenario_selector.

We only have one concern with the stop planner running after the parking scenario. The replan feature of the freespace planner allows the parking trajectory to be updated when a collision is detected and we are not sure if this will work correctly when the the stop planner also stops the vehicle because of the obstacle.

[ismetatabay]

Hi @maxime-clem, sorry for the late reply. I was on leave. Actually, I am planning to update the launch structure with input topics, instead of changing node launching in launch files. What do you think about it?

We only have one concern with the stop planner running after the parking scenario. The replan feature of the freespace planner allows the parking trajectory to be updated when a collision is detected and we are not sure if this will work correctly when the the stop planner also stops the vehicle because of the obstacle.

II will investigate this situation. From my initial assessment, it doesn't seem to be a problem. If an object is detected on the path before replanning, the vehicle will stop, and if the replanned trajectory doesn't encounter any objects, the ego vehicle will continue its movement. However, before confirming anything, I need to conduct further investigation. Thank you!