Indoor Environment Modeling Based On Non-crosstalk Sonar System

With the expansion of mobile robot’s work area, it proposes new challenge to therobot’s intelligence and autonomation, which is the ability of working in an unknownenvironment. Firstly, the robot is needed to be able to model the unknownenvironment and create the environment map. This thesis introduces the design andimplementation of indoor environment modeling based on non-crosstalk sonarsystem.In the non-crosstalk sonar system, the Gold sequence combined with BFSK isapplied as the excitation sequence. The cross-zero polarity correlation method isadopted to process the echo signal and to capture the time-of-flight (TOF) ofultrasound. The hardware platform is designed based on FPGA, including transceiverintegrated circuit, signal processing circuit, RS-232communication module, etc.A special ultrasonic transducer array is designed and driven by an stepping motor,so it can achieve360°environmental scanning. During the process of scanning, FPGAcontroller sends sonar data to the computer via serial communication. Then the data issaved in a file and processed as follow: transformation of coordinates, datapreprocessing, which means the object-shape is recognized by using BP network, andRegion of Constant Depth extracting.To build a geometric feature map, a modified TBF algorithm and Least Squaremethod are applied to improve the accuracy of the map. The grid map is built by usingBayes Estimation and refined by the point estimations from modified TBF algorithm.After the comparison of two maps, the combination of the two maps can be used innavigation.