Adaptive Robot Navigation and Control

I've developed a navigation and control system for autonomous vehicles that combines the Pure Pursuit controller with a novel O(1) motion profiling algorithm for increased reliability and accuracy. Real-time feedback from an IMU and wheel velocities ensures path faithfulness and accuracy. The system provides options for path replanning consideration based on confidence/probability metrics of completing a path successfully. Initial paths are generated from arc-length parameterized Bezier curves for consistency along with smooth and uniformly sampled routes. I've implemented automated PID tuning within this system, so feedback controllers receive variable gains that are generated from regression on a dataset of previously executed paths and most successful gains.

Live Videos


Below are videos of the system deployed on a VEX robotics testbed, with paths generated to optimize scoring points.


Simulation



Video of my standalone Pure Pursuit and motion profiling algorithm simulated in MATLAB.