Traditional model predictive control trajectory tracking - wavy trajectory (function package has been updated)

More than a month later , Finally, I have time to upgrade this feature pack framework .

above Traditional model predictive control trajectory tracking —— Circular trajectory The function packs in are updated as follows ：

mpc_simulation_ROS_and_experiment, Please refer to README.md.

Updated features

1. When sending the path, the planned speed and angular speed are also sent out

2. stay mpc_client.cpp Intermediate speed 、 Angular velocity is transmitted to mpc_server.py To participate in the optimization calculation as the expected value

3. A wavy path is updated to verify the tracking algorithm

4. The procedure and operation manual of the real vehicle experiment have been updated （ see path_track/readme.txt）

What needs to be explained

1. Only the car is considered x Directional velocity and angular velocity , Of course y The speed of the direction is already path_track/path_point.srv Reserved interface , You can join in y Try the direction and speed , There seems to be some problems .

2. Due to the design of prediction model , The state quantity is pose error , The control quantity is speed increment , therefore For the given path, the speed information of the path point must be required , This path can be designed in send_point.launch Set in , You can also change it directly cpp file .

3. Simulated or real car running mpc when , The distance between the position and posture of the trolley and the starting point of the path should not exceed 1m, Otherwise, the simulation effect will become worse , This may be related to some parameters in the controller , Such as prediction time domain .

4. Can't run the whole circle , It may be related to the attitude angle error , When transiting from the front semicircle to the back semicircle, the angle error does not give a suitable calculation method , I haven't dealt with it yet .

Simulation results

Wavy curve , Red is the car track , Green is the planning track .

The tail of the last point didn't close because there was a three-stage deceleration in the optimizer , If necessary, you can comment out the deceleration judgment program .

The specific tracking situation is as follows ：

mpc_track_wave_trajectory

  experimental result

Track straight line trajectory

The specific tracking situation is as follows ：

mpc_track_line_experiment