Quadrotor Robust Adaptive Control Research
This is a repository undergoing development for use in my research in robust and adaptive control methods. I will be implementing adaptive control barrier function, robust adaptive tube MPC, and L1 adaptive control for quadrotor trajectory tracking. The package also includes convenient lightweight simulation tools, and I have future plans to incorporate ROS functionalities as well. The project can be found here.
Below are two scenarios involving the same nonlinear MPC with modeling error: the true mass and moments of inertia of the drone are 8 times greater than those of the MPC's model. The first scenario demonstrates the MPC attempting to converge on a position setpoint of x = 5, y = 5, and z = 5 meters. Due to the modeling error, the MPC produces unstable tracking and steady-state error in the z-direction.
The second scenario demonstrates the same MPC and modeling error but in a cascaded control architecture.
The MPC serves as the outer reference controller for an inner L1 controller with piecewise constant adaptation.
With no information of the discrepancies between the MPC's model and the true plant dynamics,
the L1-augmented MPC is able to arrive at the same position setpoint more efficiently and with less error.
At the end of the simulated 10 seconds, the final position is visible as the first 3 states in the terminal output.
The L1-augmented MPC, still in the process of adapting, can be seen converging on the setpoint with a position of
~ x = 4.9, y = 4.9, z = 4.7.
For information on implementation of L1 with piecewise constant adaptation, check out these papers:
Performance, Precision, and Payloads: Adaptive Nonlinear MPC for Quadrotors
L1-Adaptive MPPI Architecture for Robust and Agile Control of Multirotors
L1 adaptive controller for multi-input multi-output systems in the presence of nonlinear unmatched uncertainties
L1 adaptive controller for MIMO systems with unmatched uncertainties using modified piecewise constant adaptation law