Use of remote sensing in urban area navigation of mobile robots

Navigation focuses on the process of monitoring and controlling the movement of a vehicle from one place to another. It can refer to any study that involves the determination of position and direction. In a broather sense, navigation is a process of trying to answer the questions where am I ?, where will I go ? and how will I go ? The answer given to the first phase of the process, where am I ? will determine the accuracy of the navigation process. Regardless of the sophistication level of the navigation path planning algorithms used, it is obvious that routes will be calculated incorrectly when wrong initial position data are used as input.
In the field of outdoor navigation of mobile robots in urban areas, GPS data as well as LASER scanners, SONAR, RADAR and LIDAR data are becoming increasingly important. These sensors are mainly used for detection of obstacles around a mobile robot, obstacle avoidance, and real-time positioning and mapping. Relationship between robot's surroundings and its location and orientation is obtained using these sensors and is used in navigation process.
In high budget mobile robot projects, very accurate and expensive GPS receivers and other sensors are used. Precise and consistent navigation operations can be performed by using these accurate sensor data. Nowadays, low cost GPS receivers used in low to medium budget robot projects have approximately a position accuracy of ± 10 meters in urban areas. This accuracy enables the determination of the position of the GPS receiver roughly in a 400 m2 area which we will define as error probability area. For small or medium sized mobile robots, in some scenarios, an area of this size can cover the entire study area. In this case robot’s exact location in the study area cannot be determined and thus navigation operations cannot be carried out precisely. Therefore increasing accuracy of the GPS location used in mobile robot navigation is of great importance.
In this paper we describe the design and implementation of a unique algorithm to increase accuracy of the GPS location that utilizes 2D laser scanner data and high resolution satellite imagery. By combining digital image processing techniques, template matching techniques and laser image, high resolution satellite imagery, compass and inertial measurement unit (IMU) data we developed an algorithm to calculate GPS error correction vector to correct erroneous GPS position obtained from low cost GPS receivers.
2D laser edge map is produced by plotting angle and distance data pairs obtained from a laser scanner. Compass and IMU data are used to correct orientation of laser edge map. Online mapping service data are used to match ground resolution of laser edge map with of satellite imagery.
Our method uses high resolution satellite imagery and Canny edge detection method to produce edge map of the robot’s surroundings and uses template matching techniques to align laser edge map produced from 2D laser scanner. In this alignment process GPS error correction vector is calculated by using X and Y axis translations. This method is demonstrated in computer simulations and field experiments. We believe that while a single method cannot be used to create a robust and efficient mobile robot localization and navigation system, our method will improve low cost GPS receivers’ position accuracy in urban areas to be used as an input to localization and navigation planning algorithms.