velocity planning

[HuaweiAlgolux]

The run_out module has already planned velocity along the geometric path based on obstacles around the ego vehicle, then later in the motion planning part, there is another obstacle_avoidance_planner(kinematics is considered), and then the surround_obstacle_checker(keep ego vehicle stationary when ego vehicle has already stopped), and then obstacle_stop_planner, looks like they are just dealing with the same thing, Why not put everything into one node? What's the rationale behind this design?

[HuaweiAlgolux]

@maxime-clem @TakaHoribe

[takayuki5168]

@HuaweiAlgolux
First, obstacle_avoidance_planner updates the path shape, and does not update the velocity against obstacles.

• run_out module updates the velocity against obstacles which may run into the ego's trajectory.
• surround_obstacle_checker updates the velocity to keep ego vehicle stationary as you said.
• obstacle_stop_planner updates the velocity to obstacles inside the ego's trajectory. (There is also a slow-down function against obstacles besides the ego's trajectory, but it's a temporary function, so let me ignore it).

Like this, each of them has a different function.
As you said, we can put them into one node, but then the node will have too much code which makes it hard for us to read the code, add implementation, debug, ... (less maintainability)

To avoid it, we take an advantage of modularization by creating nodes that have a simple function and concatenating these nodes successively. Then we can add a lot of functions keeping the code not too complex.

[takayuki5168]

@HuaweiAlgolux Can I close the discussion?

[HuaweiAlgolux]

yes

[takayuki5168]

@HuaweiAlgolux It seems I cannot. Could you make this discussion answered please.

[maxime-clem]

This is a "Design" discussion so it cannot be "closed" or "answered".

[takayuki5168]

Ahh, I got it. Thanks for your information!