Lane keeping control algorithm using image processing

Abstract

Motivated by the ideas of an autonomously driven vehicle and driving safety issues, driver assistance systems such as active braking, cruise control and lane departure warning lane-keeping, become a very active research area. Over the years, studies have shown that several control strategies have been developed and tested with the use of sensor like LIDAR with very little accuracy. The use of LIDAR sensor could not accurately detect the lanes especially in bad weather conditions like rain and snow. Image processing technique has a great potential of solving this issue, however, very little research has been done with a camera to detect the road lane. The aim of this thesis was to develop a controller algorithm for lane keeping using image processing methods. Both simulation and experimental studies were carried out using a vehicle model based on a single-track bicycle model. An image processing algorithm was developed for lane detection using various edge detection techniques. The results show that Roberts edge detector performed better compared to other edge detectors with detection time of 0.4 sec and lane detection of 0.8 sec. A Proportional Derivative (PD) controller was simulated by manually tuning its parameters to keep the vehicle in its desired track. The model was further validated by an experimental PD study, where the PD parameters were also tuned manually. In the simulation studies, the image processing algorithm process the image by identifying the error signal, which was sent to the controller that generates the steering control command to drive the vehicle toward the desired reference track. This same method was also used in the experimental studies. The results showed that the control strategy for the PD control achieved the objective of steering the vehicle towards the reference trajectory by reducing the lateral deviation error to zero at a much lower longitudinal velocity of 1m/s and a Kp gain of 0.004 and Kd gain of 0.001 and the experimental results showed a similarity with simulations results. Further analysis found that PD controller was not robust in maintaining its performance under various conditions as a result of changing parameters. Using the same system setup, model predictive control (MPC), Fuzzy logic and Fuzzy-proportion, integral derivative (PID) control algorithms were simulated to improve lane-keeping system performance. Fuzzyproportion integral derivative showed the best performance with less overshoot, a maximum lateral deviation of 2 cm and settling time of 12 sec, Kp of 0.01, Ki of 0.01 and Kd of 0.06. Better and faster response control of the vehicle was achieved using Fuzzyproportional integral derivative (PID) controller.