Position Estimation and Autonomous Control of a Quad Vehicle 2016-01-1878
The major contribution of this paper is the general description of a complete integrating procedure of autonomous vehicle system. Using Robot Operating System (ROS) as the framework, process from senor integration to path planning and path tracking were performed. Based on an off-road All-Terrain Vehicle, an Extended Kalman filter based autonomous control strategy was developed on the ROS. Both the position estimation and autonomous control were performed on the ROS platform. For the position estimation phase, sensory measurements from GPS, IMU and wheel odometry were acquired and processed on ROS. In accordance with the ROS architecture, separate packages were developed for each sensor to gather and publish corresponding measurements. Furthermore, Extended Kalman filtering was performed to fuse all sensory measurements to achieve an optimizing accuracy. Necessary conversion and normalization were also conducted prior to data fusion, such as conversion from LLA to ENU for GPS measurements. For the autonomous control phase, a ROS package was first developed to generate target path. By defining several control points, the algorithm performed cubic spline interpolation to smoothly connect every two adjacent control points. Only the coefficients of the computed splines were transferred to other packages to be more efficient. With the predefined target path, the Pure Pursuit algorithm was implemented to compute desired steering angle. Meanwhile, a simple longitudinal control strategy, which defines velocity as a function of steering angle, was also developed to calculate throttle value. All developed algorithms were first validated in simulation, and some preliminary field tests were also conducted to examine the performance of the proposed approach.
This is a general guidance of implementing autonomous control with low cost sensors on the ROS platform. With the mentioned general procedures, this paper can allow readers to have an understanding of the autonomous control architecture. Meanwhile, the adopted algorithms can facilitate the development of similar applications. The proposal of using low cost sensors would also contribute to increase the acceptance of autonomous vehicles.