In this project, we implemented an iLQR-based algorithm that allowed the self-driving trucks to complete adaptive cruise control and overtaking tasks. This was done by adapting the cost functions provided in previous labs and interpolating reference trajectories and velocities to ensure higher efficacy and robustness of the algorithm. These iLQR adaptations were tested by implementing them on the autonomous trucks in the MiniCity environment. The algorithm adaptations showed excellent performance on these tasks, as was seen during the in-class demonstration. This work is described here as well as discussions on shortcomings of the implementations and future experiments that could further boost the performance of autonomous navigation.
See full report for further details.
Clone the Git repository and setup the environment.
$ git clone https://github.com/dyang1234/ECE346_Group05_07_project.git
$ cd ECE346_Group05_07_project
$ ./setup_env.shTo set up the shell environment, source ONE of the setup files
$ source setup.sh # without ROS network communication$ source setup_client.sh 07 # set truck as client with host NX7$ source setup_host.sh # set truck as hostROS is a distributed computing environment. A running ROS system can comprise dozens, even hundreds of nodes, spread across multiple machines. Depending on how the system is configured, any node may need to communicate with any other node, at any time.
When running ROS over network, there should be exactly one host and can have multiple clients. The ROS Core is running on the host, and clients shared their messages with hosts and other clients.
To setup ROS network, we provided simple scripts. You only need to modify the host.sh and client.sh files to replace proper IP addresses of your host and client robots. Then for each new termainal, you opened you should source the .sh file before you source devel/setup.bash.
We provide perception, control, and base planning codes for you to start working on Final Projects. Both control and perception codes are same as you had in the lab and self-contained, so that you will not need to do any updates with submodules.
The new planning code still based on the iLQR that you had in labs. Instead of using a secondary PID controller to track planned trajectory, we use the planned iLQR policy directly for planning.
In order to run multiple robots over network using one ROS Core, you should setup the namespace in src/Planning/traj_planning_ros/launch/ilqr_planning.launch file. Otherwise, multiple robots will publish conflicted messages using the same name.
You should adjust this line in the src/Control/maestro_control_ros/rc_control/cfg/calibration.cfg
gen.add("steering_C", double_t, 0, "Calibrate the netural position of steering.", Choose a value between -1 and 1, -1, 1)