Delay-aware-Robust-Tube-MPC

The codes for :-

1. Delay aware robust control for safe autonomous driving
2. Delay aware robust control for safe autonomous driving and racing

NOTE : There are some notation errors in our IEEE IV conference paper (1). The original non-linear model is linearized around the current state at every time step into a linear system. However, in the MPC's prediction horizon, we assume time-invariant dynamics. An updated version can be found here : Updated arxiv version

Folder structure

• Controller A/ and Controller B/ : All the code files for Controller A and B implementation of Delay aware robust control for safe autonomous driving which includes delay aware robust tube MPC formulation for global frame of reference with simulations in ROS Gazebo

• Controller racing A/ and Controller racing B/ : All the code files for Controller A and B implementation of Delay aware robust control for safe autonomous driving and racing which includes delay aware robust tube MPC formulation for frenet frame of reference with simulations on race-like environments on Carla

• global_racetrajectory_optimization/ : Adapted from Link. Contains code files for generating global racing line reference for 2nd work.

• launch/ : Contains launch files to be used for ROS gazebo simulations

• models/ : Contains additional models used in Gazebo simulation

• worlds/ : Contains world files for gazebo simulation

Requirements

• Carla (>=0.9.7)
• ROS (relevant version depending on OS)
• Gazebo (Required only for 1st work)
• Python (>3.8)
• pytope
• pygame
• scipy
• casadi

Steps to run 2nd work simulations

1. Set up carla(>=0.9.7, tested on 0.9.8) under 'Controller racing A' directory
2. Install additional map Town07 following the instructions from :-

Plan A experiments

1. Change directory to 'Controller racing A'

cd Controller\ racing\ A

2. (optional) Invariant set has already been pre-calculated, but if needed to recalculate for different parameters, change parameters in header file of inv_set_calc.py and run:-

python inv_set_calc.py

3. In each new terminal from now on, source ros at the beginning. In a new terminal run :-

roscore

4. In a new terminal, launch carla, make sure Town07 map is installed.
5. Set SCENE='standalone' for experiment 2, 'one_vehicle' for experiment 3, 'one_vehicle_turn' for experiment 4 in carla_utils.py
6. In a new terminal, for experiment 2, set hyperparameters in header of mpc_robust_frenet_without_obstacles.py as required and run :-

python mpc_robust_frenet_without_obstacles.py 

For experiment 3 and 4, change hyperparameters in header of mpc_robust_frenet.py and run :-

python mpc_robust_frenet.py 

7. In a new terminal, launch :-

python pre_compensator.py

The vehicle should be moving at this point. The experimental result files and plots should be saved at the end of the experiment under 'outputs_' folder with suffix 'with_comp' and 'without_comp' respectively based on if delay compensation is enabled or not in header of mpc_robust_frenet_without_obstacles.py or mpc_robust_frenet.py accordingly

Plan B experiments

1. Change directory to 'Controller racing B'

cd Controller\ racing\ B

2. In each new terminal from now on, source ros at the beginning. In a new terminal run :-

roscore

3. In a new terminal, launch carla, make sure Town07 map is installed.
4. Set SCENE='standalone' for experiment 1, 'one_vehicle' for experiment 3, 'one_vehicle_turn' for experiment 4 in carla_utils.py
5. In a new terminal, for experiment 2, set hyperparameters in header of mpc_robust_frenet_without_obstacles.py as required and run :-

python mpc_robust_frenet_without_obstacles.py 

For experiment 3 and 4, change hyperparameters in header of mpc_robust_frenet.py and run :-

python mpc_robust_frenet.py 

6. In a new terminal, launch :-

python pre_compensator.py