Hybrid controller approach for an autonomous ground vehicle path tracking problem

Automotive industry is the one of the most important economic sectors according to its circulation. Starting from the last part of the 18th century, the industry keeps its up-to-dateness with the keep tracking the future technology very closely. According to these development, autonomous driving function is become hot topic when the range of the automotive industry is under consideration. Autonomous function is mainly based on driving without any labor which try to reduce faults cause by the humans. Thanks to this aim, safety, comfortable and effective transportation will offer by the future self-driver. Automated driving requires deep understanding and cooperation of many different disciplines and topics, such as sensor technologies, localization and mapping technologies, estimation and fusion algorithms, image processing algorithms, decision making and trajectory generation algorithms, vehicle controls theory and automotive engineering An ordinary driver just steers the steering wheel and apply brake or gas pedal to follow the lane and adjust the speed of the vehicle even without thinking. Nevertheless, this path following problem is under research for years as can be observed from the literature.. Path tracking methods can be divided into two main groups: Geometric and model based control methods. Geometric methods use only the geometrical relation between the path and the vehicle. In this thesis two geometric based selected Pure Pursuit and Stanley method. Then one model based method is selected as Steady State Cornering method which devoleped from linearized bicycled model. First two methods are based on the conception of an ordinary driver trying to track a given path. While the third approach is based on a simplified mathematical model of the vehicle that is trying to follow a given path. In that sense, it can be said that first two is more intuitive than the latter one, and it is obvious that each method has its own strengths and weaknesses. In contrast, Pure-Pursuit and Steady State Cornering methods look forward in order to maneuver. For that reason, the latter two methods can preview sudden changes on the path beforehand. But on a smooth path, these two methods cut corners and their performances are not as good as Stanley In order to use advantages of different methods at the same time, innovative approach that based on the combination of two method is proposed as hybrid controller. The proposed hybrid controller is using Pure-Pursuit and Stanley Method at the same time. A weight factor is adjusted depending on the smoothness of the path ahead. As the path gets smoother, the weight of the Stanley method is increased, if a sharp change is ahead the weight of the pure pursuit method is increased. To decide if the path is smooth or not, the look ahead strategy used in steady state method is implemented. After that, the proposed hybrid controller and three path tracking method performances are examined with three different path sceniros. These paths are 50 meter radius circle path, rectangular shaped path and mixed of rectangular and circle path. In comparisons, in order to see effect of speed changes on methods, simulations are done with three different speed such as 20 km/h 50km/h and 80 km/h. To sum up in this thesis, the first part will cover the literature survey on the wide range path following problem. Afterwards, the vehicle model that will be used on whole studies about the thesis will introduce at the Section-2. In Section-3, the tracking methodologies will be summarized with their mathematical backgrounds. The proposed hybrid methodology will present in Section-4. Accordingly, three different simulation studies and their comparisons will discuss in Section-5. Finally, the conclusion and planning future works on these topics will represent at the last part.