Risk-Aware Model Predictive Control for Unmanned Ground Vehicles

The website for this project can be found at: https://frankgon1627.github.io/risk_mpc.github.io/

This is the base image used for negative obstacle detection at KR, for the Jackals to run ROS2-Jazzy on Ubuntu-24.04

TL;DR:

• It installs the dependencies required by many user packages (see risk-mpc/Dockerfile).
• It copies the ws main dcist workspace and builds it.
• It provides helper scripts build.bash, run.bash and join.bash

Architectures

Three architectures are provided:

• kumarrobotics/risk-mpc-bare - x86_64 CPU: To run in CPU-only platforms, such as the Intel NUC for the high-altitude quads. Based on ubuntu Docker images.
• kumarrobotics/risk-mpc-nvda - x86_64 CUDA: To run in x86_64 GPU-accelerated platforms, such as the computers on the Jackals. Based on nvidia/cuda Docker images.
• kumarrobotics/risk-mpc-nvda - arm64 CUDA: To run in Nvidia Jetson platforms.

How to build and use?

git clone https://github.com/KumarRobotics/risk_mpc.git
cd risk_mpc && git submodule update --init --recursive
./build.bash risk-mpc x86_64_nvda
./run.bash risk-mpc-nvda:latest

Notes on how to start up the Jackal

To launch the base hardware (no sensors), use the following commands:

ros2 launch /etc/clearpath/platform/launch/platform-service.launch.py
ros2 run safety_controller safety_controller

To ensure the jackal can respond to autonomous commands, ensure the controller is set to "Auto" mode and that command velocities are published as TwistStamped messages to the topic: /j100_0000/autonomous/cmd_vel.

If the jackal does not move with teleop, make sure to check the following things:

• The controller is in manual mode
• run ros2 service list. You'll find a service that looks like /platform/mcu/configure. Note that it may have a prefix, such as /jackal. If this is the case, the namespace needs to be udpated. Run the following command: ros2 service call /platform/mcu/configure clearpath_platform_msgs/srv/ConfigureMcu "{domain_id: 0, robot_namespace: 'j100_0000'}"

Notes on how to start the zed and ouster

To launch the camera and LIDAR, use the following commands:

ros2 launch ouster_ros sensor.composite.launch.py viz:=false
ros2 launch zed_wrapper zed_camera.launch.py camera_model:=zed2i

Notes on how to get odometry estimates

To receive odometry, use the following commands:

ros2 launch direct_lidar_inertial_odometry dlio.launch.py rviz:=false

Notes on running the planning framework

To get obstacle detection

ros2 launch groundgrid ground_grid.launch.py
ros2 launch obstacle_detection obstacle_detection.launch.py

To get the trajectory planning and control framework

ros2 launch planners planners.launch.py
ros2 run mpc_controller mpc_llc

Notes on launching the GPS for validation

ros2 launch ublox_gps ublox_gps_node-composed-launch.py