Research On Obstacle Avoidance Decision Of Mobile Robot Based On Multi-sensor Information Fusion

With the deepening of research on mobile robots,it has more and more functions and can complete different tasks.Among them,Multi-sensor information fusion and obstacle avoidance decision-making are two of the important research directions in the application of mobile robot technology.This article is aimed at the problem of low fusion accuracy of mobile robots in information fusion and low efficiency of obstacle avoidance decision-making methods in known and unknown environments.The main research contents are as follows:Firstly,it constructs a mathematical model of the mobile robot movement principle,and completes the construction of the mobile robot system and device selection.For the external environment information collection part,the sensor for measuring obstacle distance information uses ultrasonic and infrared-based sensors.Secondly,the research on multi-sensor information fusion methods includes the fundamental concepts,fusion structure and levels,and basic methods of multi-sensor information fusion.According to the system structure of the built mobile robot,a centralized fusion structure and data hierarchical fusion are adopted,and the information measured by the sensor is directly processed by combining the weighted average method and the Kalman filter method.In a static measurement environment,a weighted Kalman filter information fusion method is proposed.The extended Kalman filter information fusion method is used for the dynamic measurement environment.Thirdly,the study of the mobile robot's obstacle avoidance decision-making process in a known environment,focusing on the basic method and obstacle avoidance technique for obstacle avoidance decision-making.Comprehensive basic obstacle avoidance decision-making methods,mainly research on artificial potential field method.Aiming at the local minimum problem in the traditional artificial potential field method,As a consequence,an artificial potential field method based on the simulated annealing algorithm is proposed.The main process of this method is to use the simulated annealing algorithm to add random target points near the position where the local minimum appears,and to direct the mobile robot to gradually escape from the region and continue to move forward.The proposed method not only enables the mobile robot to escape the local minimum position and successfully reach the target point position,but also takes a shorter time and is more stable than other methods by the MATLAB simulation.Fourthly,the research on the decision-making method of mobile robot avoiding obstacles in unknown environment introduces the main content of reinforcement learning.It primarily studies the Q-learning algorithm in reinforcement learning.Aiming at the problem of the lack of cognitive information in the early unknown environment,the mobile robot has a slow search speed in the early stage and cannot quickly find the shortest path.Therefore,an improved Q-learning algorithm is proposed.The main principle of this method is to use the potential energy of the gravitational field in the artificial potential field method as the early cognitive information of the unknown environment to initialize the Q value function,which solves the problem of slow initial search speed and convergence efficiency.According to the simulation experiments,the improved Q-learning algorithm obtains a better optimization effect under the condition of less training times,and has a higher learning efficiency.Finally,experimental verification and interpretation of the results.A multi-sensor information fusion experiment,divided into static measurement fusion experiment and dynamic measurement fusion experiment,verify that the fusion method improves the measurement fusion accuracy of mobile robots in different situations.The obstacle avoidance decision-making experiments of mobile robots in known and unknown environments have verified the feasibility and superiority of the two improved obstacle avoidance decision-making methods in different environments.