Multi-sensor Data Fusion for Robotics Applications

Thanks to faster and better integrated processing units, as well as increasingly miniaturized and precise sensors, Robotics is experiencing a period of significant development. New technology is making robots increasingly more autonomous. Data from different sensors can be fused and processed in real time by on-board processing units in order to make the robot perform either more or less complicated tasks. Unmanned Ground Vehicles (UGV) and Unmanned Aerial Vehicles (UAV) equipped with sensors and microprocessors/microcontrollers can be remotely piloted. They can also move autonomously with the help of different navigation and localization methods. Power electronics is contributing to a transformation in many fields such as robotics, automotive and consumer devices. Some new applications of power electronics are discussed in the first chapter of this thesis. One of these describes how new power electronics devices allow the use of distributed instead of centralized control in industrial robotics. Multi-sensor data fusion theory is presented in the second chapter; the Kalman Filter and Particle Filter are described. Inertial sensors and magnetometers are introduced in the third chapter with the description of a calibration procedure for each sensor. Two new methods of multi-sensor data fusion, based on low-cost Micro Electro-Mechanical Systems (MEMS) Inertial Sensors, to estimate joint angles of industrial manipulators are described in the fourth chapter. The results of two experimental tests are also presented to evaluate and to compare the performances of the two methods. A method to estimate the attitude and heading of an UAV is described in the fifth chapter. In order to check the performance of the developed method, at the end of the fifth chapter, a comparison test with a high accuracy system is presented. A localization algorithm for a wheeled UGV equipped with a GPS and an inertial platform is described in the sixth chapter. Simulation tests and experimental results are also presented at the end of the sixth chapter.