Implementation Details
Vehicle
The majority of the work done on the vehicle was completed last year by Team Lo-Co, which can be seen here.
Our “implementation details” are a series of trial-and-error haps and mishaps. The Lo-Co is a complicated system with many subsystems of its own that took some time to crack.
The first step was learning how to communicate with the vehicle. This required using the special Team_Loco_2.4 router which communicates with the onboard Jetson TX1. With each of these powered in, you can ssh into the tx1 from your local machine.
Next, we had to learn how to make the vehicle obey our commands. Teleoperation was easy, but our attempts to run the car with the open_loop launch file led to some ugly crashes early on.
It took some digging in, but we learned over time how to work with the .pgm map files that allow the vehicle to localize itself using the adaptive monte carlo localizer. This allowed us to utilize the iterative linear quadratic regulator control scheme that the Loco team implemented.
After a little tuning for the yaw control, we are ready to run our vehicle to the waypoint as specified in our fall validation experiment:
What remains is teaching this vehicle to avoid pedestrians!