Research On Ultrasonic Network Positioning Method For Indoor Mobile Robot

Self-localization is the most important and fundamental function ofautonomous mobile robots and it is the precondition of completing a series ofintelligent tasks. Compared to other environmental perception sensors, ultrasonicsensors, because of its low price, easy implementation and mature technology, etc.,have been widely used in mobile robot indoor positioning systems.In this paper, ultrasonic network localization method for mobile robot in aknown indoor environment was studied. There are two disadvantages of ultrasonicnetwork localization methods. One is its positioning failure with the presence ofobstacles which blocked the signals. The other is its foundational hardware facilities.Aiming at solving the above mentioned two issues, a multi-sensor fusion algorithmwas used in the integrated positioning system utilizing ultrasonic sensor-basednetworks, encoders and electronic compass. The following aspects are investigated.Kinematic model of the double-encoder-based differential driving mobile robotwas established. Through analyzing the principle and error of ultrasonic networkpositioning system and digital compass, the observation model was established.Multi-sensor data is fused via the extended Kalman filter such that ultrasonic sensornetwork functions well even if signal is lost. A computer simulation was conducted.A robot self-localization method using the sparse ultrasonic sensor network,encoders and digital compass was proposed to reduce the number of sensors. Thismethod makes use of asynchronous Time-of-Arrival (ATOA) measurements toobtain the position of robot. The principle and error of the positioning method aredescribed. A double-layer Kalman filter (DLKF) for data fusion is proposed. Thecomputer simulation was also conducted.Active and passive ultrasonic network positioning were analyzed and compared.An improved active positioning method was proposed. An experimental robotsystem was designed and an ultrasonic network positioning system was build. Thesystem is adjusted and modified to achieve a good performance. The robotlocalization experiments employing dense and sparse sensor network wereconducted and led to a positioning accuracy of2cm and5cm, respectively. Thefactors affecting the positioning accuracy were investigated and the effectiveness ofthe positioning method is verified.