Novel research by Kyle Wray, combined with the competitive and vibrant atmosphere of HackUMass
What it does
The algorithm plans an optimally smooth path in log space using harmonic functions on an NVIDIA GPU, and executes a basic controller to follow the given path using odometry data.
How I built it
We took the standard Kobuki turtlebot platform, mounted an Intel NUC, and established a wireless connection over which to share the path planning results. We wrote a ROS node to process the command stream from the planner and send velocity commands to the motor
Challenges I ran into
Accomplishments that I'm proud of
We're surprised that we were even able to get ROS and the GPU path planner code to link in less than 24 hours. There's a lot of moving parts under the hood, and it's kind of a miracle it worked. We're proud of that for sure.
What I learned
ROS in python, and writing servers in python. General networking.
What's next for Cute Turtlebot
We'd like to implement a smoother controller, like a PD or PID controller