This repository is a public version of the OSU F1/10 research made avaliable for review

This repository includes mapping, localization, planning, and control algorithms along with hardware level interface packages designed to work with the F1/10 platform. Used together, these tools enable fully autonomous navigation and racing for the vehicle.

This code has a large number of dependencies which are not fully listed here. This was originally developed on the F1/10 research platform using the Nvidia Jetson TX2 to run everything. Many configuration steps specific to the jetpack version were adopted.

Some of the larger external dependencies:

• properly configured ROS for use with C++,Python3,Python3.6
• specific dependencies for each node (various apt packages, pip installs)
• various build tools including python-dev and build essential
• (to run reachability) Hylaa (https://github.com/stanleybak/hylaa) and/or QuickZonoReach(https://github.com/NathanJewell/quickzonoreach/tree/cuda) code in the PATH
• same hardware, or properly configured launch files and ROS nodes for your specific hardware configuration

It is necessary to create the rasterized map using the cartographer nodes. For using the MPCC it is then also necessary to run some post-processing steps and generate the optimal racing lines etc for that specific map.

Once this code is properly configured, these steps, or equivalent (if you have had to create new launch files for you configuration) should start every node needed. I would reccomend running this code using screen/tmux etc since there are a number of simultaneous windows to monitor. Multiple terminal windows could also be used. Generally, I configured another computer with ROS_MASTER_URI and used that for visualization.

• run rviz on control pc
• ssh into the vehicle
• roslaunch nonlinear_mpc_casadi real_particle_filter.launch
• roslaunch nonlinear_mpc_casadi mpc_car1.launch
• roslaunch reachability quickzono.launch
• add rviz topics to visualization
• create nav waypoint in rviz
• use controller/deadman to start autonomous driving

By no means exhaustive ...

• The MPCC uses JIT compilation so takes a while to start
• Using the QZ_HYBRID reachability mode has bugs in the kernel, use <20 steps to avoid crashes in this mode
• If the visualization breaks or nav point cannot be set, close rviz and the mpcc node then restart rviz then the mpcc node

Be in touch if and when there are complications using this code.