R. Molina-Mesa and J. L. Martínez Andalusian Institute of Advanced Automation and Robotics. Severo Ochoa Street, No. 4 Andalusian Technological Park. 29590-Málaga, Spain Phone: +34 5 2137204. Fax: +34 5 2137203. E-mail: email@example.com
Abstract: Mobile robots must have some effective method for determining theirposition in the working environment in order to operate appropriately. In this paper, precise outdoor positioning of a vehicle is achieved by continuously fusing odometry with Differential Global Positioning System (DGPS) data through a Kalman filter. The propagation and reduction of spatial uncertainty is computed together with the mobile robot’s coordinates and heading. Here it is shown theapplication of this method to autonomous navigation of the Aurora mobile robot in a road network. Path recording and path tracking have been successfully tested in several experiments over an even surface.
Outdoor location of mobile robots is usually considered more complex than indoors since environments are more complex, less structured and distances bigger (Li and Hayashi,1998). The odometric or inertial sensors of a vehicle provides fast positional information, but involves unavoidable errors that can grow with the travelled distance. So it is necessary to periodically reduce spatial uncertainty by using external sensors. Recently, it is available world-wide and all-weather GPS receivers for outdoor navigation. Basically this is a triangulation method based on theknowledge of the position of artificial satellites that act as landmarks by broadcasting radio signals. This technique can be enhanced with differential corrections sent by a base station placed at perfect known coordinates that reduce positioning errors (Martínez et al., 2000). For mobile robot navigation Real Time Kinematic GPS is able to obtain accuracies of up 2 cm at a high cost (Rintanen etal., 1995). However, the use of commercial differential corrections from an FM link or a special satellite (Fugro, 1997) prevents the end-user from the installation of a local base station and the corresponding transmission equipment but precision decreases to 1 m approximately. In the path tracking problem the objective is to follow a path in spite of external perturbations by manipulating thevehicle’s speed and steering. Explicit path tracking is
based on fast position estimation produced by the onboard mobile robot’s sensorial system (Aono et al., 1998). In this paper it is proposed the application of a periodic sensor based localization method via a Kalman filter to the problem of appropriately registering and tracking paths for outdoor environments, which are basic capabilities forautonomous navigation of mobile robots. The rest of the paper is organized as follows. Next section reviews the Aurora mobile robot. Data fusion for outdoor location of the vehicle is addressed in section 3. Autonomous navigation is described in section 4. Then, experimental results are presented in section 5. Finally, sections 6, 7 and 8 are dedicated to acknowledgments, conclusions and referencesrespectively.
2. THE MOBILE ROBOT AURORA
The autonomous vehicle Aurora was originally conceived for agricultural operations inside greenhouses (Mandow et al., 1996). Aurora has rectangular dimensions: 1.4x0.8 m and 1 m of height (see Figure 1). A petrol-fed electric generator of 2600 W provides energy to the mobile robot. But it can also be wired into the conventional electrical network forlaboratory tests. The vehicle’s movement is obtained by the independent performance of three AC motors. Differential steering is
employed to move the lateral wheels at a maximum speed of v m=0.82 m/s when advancing in a straight line. It is also necessary to apply the corresponding steering angle to the rear and front wheels which are driven at the same time by a rigid link.