Development And Research Of Safeguard Robot Based On Bionic Perception System

The bionic security robots have the environment perception function and play a moreand more important role in the relief work of disaster, so the research of intelligent securityrobot with bionic sensing functions is of great significance. The cross mode of science andattention mechanism from the brain and cognitive sciences is borrowed in the paper. Based onthe mobile robot’s platform, the security robot, which has certain ability of bionic sensing forsafety to the environment, is developed. Human’s joint action of visual and auditory isimitated. The coarse positioning of the sound source is realized through auditory perceptionsystem, and the target recognition and precise positioning can be achieved through visionperception system. Based on the fusion of multi-mode information the fast detection system isdesigned and realized, and robot can understand the environment better.(1) The sound attention for the robot is realized, and the generalized cross correlationpositioning algorithm based on time delay estimation and the steered beam-forming algorithmare improved. In order to improve the precision of time delay estimation, the following fourmethods are used. Firstly, the voice activity detection is adopted. Secondly, the search scope islimited. Thirdly, the cross-correlation functions are calculated in sections. Finally, themicrophone is regard as a reference alternately. As the steered beam-forming positioningalgorithm has a large computation, a positioning algorithm combining particle swarmalgorithm and genetic algorithm is proposed. In this new search algorithm the advantage ofgroup is introduced to search for the optimal solution, and the computation is greatly reduced.(2) The detection and further cognition of target are realized by using visual perception.According to the human visual perception mechanism, the bionic vision system for robot isdesigned. In this system segmentation is realized according to color, and target detection andrecognition are achieved through shape perception. The color model of flame is given aftercolor sensitivity analysis, and the expansion is combined with scintillation characteristics. Afusion detection method of multi-feature for dynamic flame is presented. Experiment resultsshow that the fusion algorithm proposed in this paper has better robustness.(3) A rapid detection system based on multi-information is designed and realized, so thesecurity robot can detect target rapidly and accurately. The robot is added with visual, auditory,olfactory, flame, temperature sensing modules, and so on. Information errors caused by singlemode can be improved through the fusion of multi-mode information. When all modulesworked synchronously the resources has a waste, and sound priority is used to improve thesituation. In this paper the phenomenon of explosion-flame is tested in the system. Simulationresults show that the fusion algorithm can satisfy the requirement of practical application.