Research On Multi-sensor Information Fusion Technology For Disaster Rescue Hexapod Bio-Robot

In recent years, multi-sensor information fusion technology is a hot topic in the field ofrobot. The redundant or complementary information is combined to produce more reliable andaccurate information which provided accurate decision-making in sports for robot. Multi-sensorinformation fusion technology is a combination of control theory, signal processing, artificialintelligence, statistical probability, and it provides a technical solution for robot working incomplex and uncertain environment. The research on multi-sensor information fusiontechnology for disaster rescue Hexapod Bio-Robot provides important information forcontrolling robot and it is the key technology for improving the intelligence of robot.(1)Paper had a detailed analysis of domestic and foreign status for multi-sensorinformation fusion technology for robot, introduced our study of disaster ruscue HexapodBio-Robot, and gave the main research content of the paper.(2)Paper studied the basic theory of multi-sensor information fusion technology,introduced the definitions and methods of multi-sensor information fusion technology anddiscussed the structure and levels of the technology.(3)Based on the analysis of information fusion method applied in avoiding obstacles,studied the Back-Propagation Neural Network and the Genetic algorithm. The both wereapplied into the avoidance system of disaster rescue Hexapod Bio-Robot. According to therequirements of the robot, the ultrasonic sensors and infrared sensors were configured in theobstacle avoidance system, the neural model for avoiding obstacle was established, set tutortraining signal and trained the Neural Network. In the training process, the neural network iseasy to fall into local minimum, we introduced the genetic algorithm into the neural network toimprove the network. Then we completed simulation of robot obstacle avoidance through thecomputer technology and toolbox in MATLAB.(4)In fact, the kinematic model and the observation model of the robot is nonlinear, thereare some defects if we solve robot localization using Kalman filter. And according to thekinematics model, mileage meter model, ultrasonic observation model, noise analysis of therobot, Extended Kalman Filter was applied into robot localization. The simulation results showthat: the error of location update can be solved and the accuracy of the robot localization can beimproved through EKF.The work mentioned above provided the theoretical and technical basis for further researchof the subject.