The future of the automotive industry is about self-driving cars, in the recent years, numerous studies has been done on the future development of autonomous vehicles. The increase can already be seen on the streets as well. Big corporations testing their latest algorithms on the highway, autonomous features like automated parking, lane keeping algorithms, ADAS (advanced driver assistance system) appearing on the newest cars and so on.
With this sudden increase in development, an advanced simulation environment is necessary, as testing each new software feature in the real world on the test-tracks or on the streets, can often be too costly or too dangerous.
The self-driving car problem is usually decomposed into three major subsystems. First the perception, with the goal of detecting different objects in the environment. Second, the high-level path planning, that defines the desired trajectory based on the target destination and detected objects in the previous layer. Finally the motion control that defines the inputs for the actuators of the car.
In an efficient simulation environment, these subsystems can be evaluated separately, while the behaviour of the whole system can still be evaluated.
This paper presents a possible way for testing advanced autonomous controller algorithms in software in the loop environment, focusing on the third subsystem, but testing the algorithm in the entire environment. First the controller itself is tuned and evaluated, by basing on a simple linear vehicle model built in Matlab. Afterwards this linear controller is tested on a nonlinear, dynamic model. Controller inputs are obtained directly from the simulated environment, while the outputs are the steering angle and vehicle velocity, so the controller works as it would on a real car. The simulation can also be driven with a predefined “driver model”, so the algorithm is compared to the driver. The importance of a refined high-level strategy is shown, as in both cases, the control-problem is solved, yet the results, the actual trajectories are significantly different.
Trajectory Tracking Controller Testing in Software in the Loop Environment
Lect.Notes Mechanical Engineering
Vehicle and Automotive Engineering ; 2020 ; Miskolc, Hungary November 25, 2020 - November 26, 2020
2020-10-20
8 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch
Develop Trajectory Tracking Controller for 4WS Vehicle
Springer Verlag | 2020
|Trajectory tracking controller for an underwater hexapod vehicle
Tema Archiv | 2008
|On the ship trajectory tracking LQG controller design
British Library Conference Proceedings | 1999
|