Follow the Past - A Path Tracking Method Using Recorded Orientation and Steering Commands.
Ola Ringdahl, Thomas Hellström · 2005
The work presented in this paper is part of the project Autonomous Navigation for Forest Machines (Georgsson et al., 2005) at Umea University, Sweden. The goal of this project is to develop a path-tracking forest vehicle. A human driver drives the path once, while a computer is continuously recording position, velocity, orientation, and steering angle. This information is then used to control the vehicle each time it autonomously travels along the path. If the vehicle gets off course, for example as a result of avoiding an obstacle or because of noise in the positioning sensors, the developed path-tracking algorithm steers like the driver, plus an additional angle, based on the distance to the path and the error in orientation. Traditional algorithms, like Follow the Carrot (Barton, 2001) and Pure Pursuit (Coulter, 1992), use position information only, and sometimes run into problems that can be avoided, by taking into account additional information from the human driver. There are many variations on the basic Pure-Pursuit algorithm. The Adaptive Pure-Pursuit algorithm (Hebert et al., 1997) addresses computational issues and stability at high speeds. The Feedforward Pure-Pursuit algorithm (Hebert et al., 1997) simulates the outcome of various control commands, and selects the most appropriate one. In (Coulter, 1992), the average curvature of the path is used instead of the recorded steering angle in a proportional path-following algorithm. Other researchers, e.g. (Ollero et al., 2001), have approached the problem with a fuzzy-logic controller that uses the same additional information as our suggested algorithm, but has a more complex design. A brief survey of more path-tracking control algorithms can be found in (Makela, 2001).