Research On Mobile Robot Odor Source Search Based On Gas Sensor Array

In recent years,with the development of industry,the leakage accidents of the strong toxic substances stored in the pressurized tanks have occurred frequently during the transportation.Therefore,it is of great practical significance to improve the accuracy of gas detection and develop effective search strategies.Based on the gas sensor array,the odor source searching of mobile olfactory robot was researched.The main research contents are as follows:The research status of artificial olfactory system and olfactory robot at home and abroad was analyzed.Based on the artificial olfactory system,the gas sensor array module was designed,and the one-to-one response mode corresponding to the measured gas was established with its inherent cross sensitive characteristics.Based on the mechanical system,the hardware circuit was designed,which can send signals to the core chip by carrying a variety of sensors for comprehensive processing,so as to the mobile robot was controlled to realize the search of odor source.The quantitative identification algorithm was analyzed and the improved BP algorithm was proposed based on cuckoo search algorithm and simulated annealing algorithm.The intelligent algorithm test functions were used to compare and analyze the test results of different algorithms.It is proved that the improved algorithm reduced dependence on the initial weights and thresholds and overcame the shortcomings of easy to trap into the local optimal value.The gas detection system designed with MATLAB software was used to obtain the detected gas concentration information,and then the concentration change curve may be got.The experimental scheme was designed and the experimental platform was built.The sensor voltage and gas concentration information obtained from the experiment were normalized to meet the input and output requirements of the neural network.With the neural network,concentration forecast of experimental data selected randomly was made and the evaluation indicators were defined.The two neural networks are applicable through analyzing and comparing their predicition results.The neural network was transplanted to the core chip,and then the neural network solidified into the embedded chip was used to realize the quantitative identification of the gas concentration.The search algorithm and search strategy were designed,and the experimental verification was performed for the known and unknown smoke plume models.When the smoke plume model was known,the Gaussian plume model was selected as the plume diffusion model.Different initial angles were adopted,and different stages were divided to search for different length and steering angle according to different concentration thresholds.It was shown that the zigzag search phase with initial angle 45o was with fewer steps,faster search speed,and more accurate position to find the highest concentration point.When the smoke plume model was unknown,the search strategy was developed and the hardware driver was designed.The upper computer interface designed by MATLAB software was used to receive the measured gas concentration information,and then the concentration change curve may be obtained.It was demonstrated that the overall trend of gas concentration detected by the mobile robot increases gradually,and the concentration reaches the maximum after the search was completed.