In this paper, we address the problem of stable and robust control of vehicles with lateral error dynamics for the application of
lane keeping. With lane departure being one of the primary reasons for half of the fatalities in road accidents, developing
controllers that are stable, adaptive as well as robust is a necessity.
Traditional linear feedback controllers achieve satisfactory tracking performance, however, they exhibit unstable behavior when
uncertainties are induced into the system. In the context of lane keeping, any disturbance or uncertainty introduced to
the steering-angle input can be catastrophic for the vehicle. Therefore, controllers must be developed to actively tackle such
uncertainties.
In this regard, we introduce a Neural L1 Adaptive controller which learns the uncertainties in the lateral error dynamics
of a front-steered Ackermann vehicle and guarantees stability and robustness. Our contributions are threefold: i) We extend the theoretical
results for guaranteed stability and robustness of conventional L1 Adaptive controllers to Neural L1 Adaptive controller; ii)We implement a
Neural L1 Adaptive controller for the lane keeping application which learns uncertainties in the dynamics;
iii) We provide simulation results using Physics-based
simulation on PyBullet as well as experimental results using the F1TENTH platform to demonstrate superior reference trajectory
tracking performance of Neural L1 Adaptive controller compared to state-of-the-art controllers, with externally induced sensor noise and control
signal disturbance. Neural L1 Adaptive controller is able to complete traversing whole arbitrary trajectories while keeping the system stable and robust
in the presence of sensor noise, external disturbances and physical obstacles, whereas other controllers exhibit unstable behavior.
A lateral lane keeping system (LKS) for a front-steered Ackermann vehicle, depicting the effect of uncertainties such as
signal disturbance, potholes and ramp, on the lane keeping performance. The above image depicts a simplified linear
two-degrees of freedom (2-DOF) bicycle model of the vehicle lateral dynamics derived.
Experimental Setup
Experimental setup at Minnesota Robotics Institute (MnRI) Drone lab.The MnRI Drone lab is equipped with PhaseSpace localization
system that allows localization of the F1TENTH vehicle in x, y, z coordinate frames. The circular reference trajectory (yellow) of radius R = 2.5m with
physical placement of obstacles ramp and planks.
Neural L1 Adaptive Control Architecture
Overview of our approach depicts the L1 adaptive
control (red region), the neural network-based adaptive
laws (green region) and the neural network training mechanism (blue region).
