Modern Control Theory | Modern Control For Robotics

CMU 24677 - Project Demos

In this class, we were tasked with developing our own controllers to enable a Tesla to complete a path in a Webot simulation environment. I develoepd 5 different controls (PID, FSF, LQR, ...) for the Tesla to complete the path as fast as possible while meeting the accuracy requirements. Through the projects, I was able to deepen my understanding of different control systems and their applications in autonomous vehicles.

Unoptimized Simulation

Simulation of an unoptimized run to compare against runs with custom controllers.

Watch the unoptimized simulation on YouTube

P1 : PID Control

Requirements:

Runtime: < 400s
Maximum Distance from Path:10m
Maximum Avg Distance from Path: 5m

Watch the PID control simulation on YouTube

P2 : Full State Feedback Control

Requirements:

Runtime: < 350s
Maximum Distance from Path: 9m
Maximum Avg Distance from Path: 4.5m

Watch the FSF control simulation on YouTube

P3 : LQR

Requirements:

Runtime: < 250s
Maximum Distance from Path: 7.0m Maximum Avg Distance from Path: 3.5m

Watch the LQR control simulation on YouTube

P5 : MRAC (Model Reference Adaptive Controller)

Developed a MRAC, LQR control for a drone that recovers from 65% thrust loss in one of the motors

Watch the MRAC drone control simulation on YouTube