Parking Planner for Diff-Drive Robot / Real-Life Issue

[cemtolgamunyas]

Hi everyone,
We are working on an autonomous truck for a specialized parking scenario. We have developed an algorithm that locates the nearest empty parking space inside the closest parking lot relative to the initial pose. There are some issues on implementing this algorithm to our specialized robot. We use a diff-drive robot and its steering ability is beyond the car-like robot. At first, this looks like an advantage but Autoware does not plan an optimal trajectory for parallel parking with the diff-drive car's vehicle info. Here is the planned trajectory with the real params of the vehicle.

So, I basically changed some parameters of the vehicle info in the freespace planner node to make the robot behave like the car-like vehicle. These are the changed parameters.

    vehicle_shape_.length = 3.0;
    vehicle_shape_.width = 0.235;
    vehicle_shape_.base_length = 2.00;
    vehicle_shape_.max_steering = 0.70;
    vehicle_shape_.base2back = 0.9;

With this modification, the planned trajectory to reach inside the parking space looks feasible.

However, in this case, our robot goes into reverse a bit much and that may cause danger in real-life scenarios in the environment that we work. So, the first question is

• Do we need to develop a specialized path planning algorithm just for planning the last trajectory to reach inside the parking space for our diff-drive robot? Do you suggest any modifications in the freespace_planner.param.yaml that outputs any differences in this case?

• On the other hand, in real-life scenario, our vehicle does not exceed the lanelets. It can locate the parking space and determines a goal pose but, it refuses to go beyond the lanelet to reach the goal pose. It just stops near the parking space and inside the lanelet. As you can see, in Rviz everything seems normal and works. What do you suggest for me to solve these problems?

Thanks in advance.

[liuXinGangChina]

hi @cemtolgamunyas
a screen cast plus rosbag may be better for us to help you

happy hacking

Xingang