Fusing Sound And Dead Reckoning For Multi-robot Cooperative Localization

The development of modern society and the progress of science and technology promote the development of robot industry.The robot has been a hot research field.The demand for multi-robot system is also increasing.The basic research of multi-robot system is the problem of multi-robot localization.The current main positioning methods have some disadvantages.It's not suitable for small mobile multi-robot platform.This thesis proposes a multi-robot distributed cooperative localization method based on the fusion of sound and dead reckoning.In a multi-robot system,at least one robot's pose is known as a reference.In the motion state,this algorithm can realize the real-time absolute positioning of other unknown robots.In this method,an unknown what's absolute pose is obtained through extended Kalman filter by fusing the robot's pose which is obtained by dead reckoning and the robot's relative position which is obtained by sound localization.This method can real-timely correct the accumulative error of dead reckoning,resulting in the improvement of the positioning accuracy.The absolute pose obtained by dead reckoning needs to know the pose in initial time.Then this initial pose accumulates the speed information of each period obtained by the motor encoder.In a long time of motion state,the error caused by the accumulated operation will gradually increase and the positioning accuracy is low.The relative position obtained by sound positioning is required to calculate the sound localization by the robot which is known of absolute pose.Only the relative position of the robot can be obtained.The absolute pose cannot be obtained.The sound positioning accuracy is about 15 cm.Through the fusion of the extended Kalman filter,real-time absolute pose of unknown robot can be realized.The accumulative error of dead reckoning can be correct and the positioning accuracy can be improved.Then,the computer simulation is divided into four types to verify this algorithm: the single robot in uniform circular motion;two robots in uniform circular motion;two robots in uniform linear motion;the four robots in random motion.The simulation effect is good and the error is small.In view of the above four conditions,experiments are carried out on the self-made small mobile multi-robot platform.Each robot is provided with a sound transmitting sensor,a sound receiving sensor and motor encoder.The experimental results show that the algorithm can correct the accumulative error of the dead reckoning to a certain degree and improve the positioning accuracy.