Research On Environmental Intelligent Perception Technology Based On Mobile Robot

As the rapid development of robotic technology and artificial intelligence, intelligent robotic systems aimed at various application scenarios have been gradually integrated into all areas such as military, industry and civil utilization, which will have great influence on the future way of production and living. The accuracy and real-time of the environmental intelligence perception which is the basic function of robots directly determines the performance of robots. This paper researches the key technologies of environmentally intelligence perception systems based on mobile robot, including the establish and update of environmental models, the modification of path planning algorithm and the display of semantic information contained environment map. The main research contents are briefly stated below:1.Aiming at the problems caused by the fixed active window of VFH+methods, this paper proposes a local path planning algorithm based on Multi-Vision and Multi-Resolution by combining the human driving experience. In this method, the active window can be changed dynamically according to the environment the robot is in. Moreover, the method enables the robot to adjust resolution size of each layer according to the obstacle complexity, to consider the cost of each alternative direction and finally to choose a target direction.The simulation shows that this method is more environmentally adaptable and solves the problems cased by fixed active window efficiently.2.The descriptions and displays of environmental information are based on the grip map. Considering the problems that the grid map weakens the overall object and that the maps do not contain object semantic information, this paper uses methods based on the morphological process and on SIFT object recognition respectively to solve these two problems. The simulation shows that the results are improved with the modified grip map.3.To solve the influences of erroneous data on the accuracy of environment model during the process of updating information, this paper designs a software filter based on the inspection of the robot current and previous situation on the grid map. Firstly, the robot predicts the next detected obstacle location according to the current movement. Secondly, it determines whether a measured value is accepted or suspended according to the deviation between predicted value and measured value. Then, it deals with the suspended value in the next cycle. The process of software filter can enhance the accuracy of environmental model so as to provide more accurate data for path planning and environment display.This paper conducts experiments and make analysis for each improvements and the whole systems on the simulation platform, and summarize in the last.