Visualizing Model Predictive Control
- Matthew Brown
A lot of work in my lab deals with planning for autonomous cars. Until now, we have run experiments, collected data, and then analyzed data; we haven't been able to see what our controllers are planning do to during a test. It would be very useful to see a visualization of the vehicle's planned states in real time. My plan is to send data from algorithms planning steering, gas, and brake inputs to a program running site view-able in a browser. This site would show actual states and inputs up until the current point in time and planned inputs in the future. A major challenge of this project is that these controllers often run at a high rate, so the visualization needs to handle new data very quickly.