Control Design For Autonomous Vehicles Research Articles

Supervised reinforcement learning based trajectory tracking control for autonomous vehicles

Recent developments in autonomous vehicle technologies and applications gain a lot of interest by the public, as the popularity of both driver assistance and automated driving systems increase. One of the most promising aspect of the autonomous vehicle compared to conventional human driven vehicle is the increased level of safety. Machine learning techniques enables to achieve fast and efficient control actions compared to model based techniques. However, the advantages of a more conservative model based controller are their better robustness properties. In this paper a synergy of the two control philosophy is presented through a trajectory tracking control design for autonomous vehicles. A supervised reinforcement learning (RL) control method is introduced, where a robust Linear Parameter Varying (LPV) controller supervises the operation of the trained RL agent. Thus, in case sensor noise is detected, the guaranteed stability LPV controller takes over the steering control action. In order to demonstrate the operation of the proposed method, three different simulations have been evaluated and compared in CarSim simulation environment.

Trajectory Tracking Model Predictive Controller Design for Autonomous Vehicles with Updating Constrains of Tire Characteristics

In this paper, we address the problem of trajectory tracking control of autonomous vehicles by considering the nonlinear characteristics of tires. By considering the influence of the tires’ dynamics on steering stability, the proposed predictive controller can track the desired trajectory and desired velocity in the presence of road curvature while minimizing the lateral tracking deviation. First of all, a hierarchical control structure is adopted, in which the upper-level controller is used to calculate the desired acceleration and the desired front-wheel angle to maintain the control target, and the lower-level controller realized the command through the corresponding component devices. Moreover, a force estimator is designed based on the radial basis function (RBF) neural network to estimate the lateral force of the tires, which is incorporated into the boundary conditions of the vehicle envelope constraint to improve the adaptability of the controller to the vehicle performance. Finally, the proposed controller is tested by co-simulation of CarSim (a simulation software specifically for vehicle dynamics)/Simulink (a modular diagram environment for multidomain simulation as well as model-based design) and hardware-in-loop simulation system. The co-simulation and experimental results demonstrate the controller safely driving at the vehicle’s handling limits and effectively reduce the slip phenomenon of the vehicle.

Improved Control Design for Autonomous Vehicles

In this paper, the autonomous vehicle presented as a discrete-time Takagi-Sugeno fuzzy (T-S) model. We used the discrete-time T-S model since it is ready for the implementation unlike the continuous T-S fuzzy model. The main goal is to keep the autonomous vehicle in the centreline of the lane regardless the external disturbances. These disturbances are the wind force and the unknown curvature; they are applied to test if the autonomous vehicle moves from the centreline. To ensure that the autonomous vehicle remain on the centreline we propose two discrete-time fuzzy lateral controllers called also steering controllers.

Runtime Assurance of Learning-Based Lane Changing Control for Autonomous Driving Vehicles

Learning techniques such as deep reinforcement learning have been increasingly used in the controller design for autonomous vehicles, e.g., lane changing controllers. Although the use of learning techniques can remarkably increase the level of autonomy, their presence poses a great challenge for safety assurance due to the black-box and data-driven nature of learning techniques. In this work, we study the problem of how to safeguard the learning-based lane changing controller for collision avoidance. Our solution leverages on the runtime assurance framework, in particular the Simplex architecture to bound the behavior of the learning-based controller and to provide safety guarantees. The basic idea is to encompass the learning-based controller with a safety-by-construction controller and a decision module, which monitors the output of the learning-based controller at runtime and implements a switching logic between these two controllers according to the changing environment. We present the detailed design of the decision module and formally prove its correctness for collision avoidance. We also carry out a comprehensive experimental evaluation in a set of realistic highway scenarios using the SUMO simulator. The simulation results show that our proposed solution can not only provide safety guarantee for the learning-based lane changing controller, but also maintain a considerable level of efficiency in different volumes of traffic flow.

Discrete-Time Takagi-Sugeno Stabilization Approach Applied in Autonomous Vehicles

This paper deals with a new robust control design for autonomous vehicles. The goal is to perform lane-keeping under various constraints, mainly unknown curvature and lateral wind force. To reach this goal, a new formulation of Parallel Distributed Compensation (PDC) law is given. The quadratic Lyapunov stability and stabilization conditions of the discrete-time Takagi–Sugeno (T-S) model representing the autonomous vehicles are discussed. Sufficient design conditions expressed in terms of strict Linear Matrix Inequalities (LMIs) extracted from the linearization of the Bilinear Matrix Inequalities (BMIs) are proposed. An illustrative example is provided to show the effectiveness of the proposed approach.

Road surface estimation based LPV control design for autonomous vehicles

The paper proposes a new road surface estimation algorithm for autonomous vehicles using a machine-learning based method, which is in cooperation with the lateral control of the vehicle. The algorithm uses large datasets, which can be collected e.g. from the on-board sensors of the vehicle, or it also can be provided by vehicle dynamic simulation softwares. The result of the surface estimation is built-in the lateral control system as a scheduling parameter. Furthermore, the lateral control design is based on the Linear Parameter Varying (LPV) method, which guarantees the safe motion of the vehicle against varying parameters of the system. Finally, a comprehensive simulation is presented to show the efficiency and the operation of the proposed control system.

Control Design for Autonomous Vehicles: A Dynamic Optimization Perspective

Control design for autonomous vehicles involves a number of issues that are not satisfactorily addressed in classical control systems theory. There is typically the need for prescribing and commanding a collection of interacting dynamic control systems in order to meet the desired requirements for overall behavior, whereas conventional control design has only one system to govern. This context requires a whole new set of concepts, methods, and tools that are adapted to capture the integration of logical commands, continuous time evolution, and discrete event dynamics. A rapidly growing and extensive body of literature testifies to the valuable heritage of control systems that will address the challenges ahead. Several new issues arise in attempting a principled design of a specific system, and it is essential that they be resolved in order to ground the effort for a suitable control design framework. Furthermore, control design cannot be dissociated from the implementation phase, which introduces a wide range of challenges in communication and computation. This paper provides an overview of three control projects designed at the Underwater Systems and Technology Laboratory, and discusses the new challenges from the dynamic optimization viewpoint.