Environmental Detection Of Mobile Fire Fighting Robot Based On Cooperative Active Perception

With its excellent ability,mobile robot has been more and more applied in all aspects of human life.How to realize autonomous ability without artificial control is always a difficult problem for mobile robots.When mobile robots are applied in fire rescue missions,the traditional method is to conduct exploration by manual remote control robot.This passive perception method is limited by the personal quality and work experience of operators,and cannot guarantee the real-time performance of environmental information.This paper proposes a cooperative active sensing method which can be applied to environmental detection of mobile fire fighting robot.Aiming at the inherent defects of the traditional human remote control robot in performing disaster rescue tasks,a bayesian inference based state estimator and q-learning based reinforcement learning decision are proposed to enable the robot to perform active environmental exploration in an unknown environment.In order to solve the problem that the traditional path planning method is difficult to obtain the optimal solution due to the large amount of computation in the face of large-scale scenes and multi-robot system tasks,a fast traversal method based on self-organizing mapping network is designed to quickly plan the shortest path for the robot team.The main work of this paper is as follows:1.This paper introduces the active perception of this problem of the research background and current research achievements,the research status of mobile robot in fire rescue tasks,etc.,mainly including the analyzes the three stages of artificial intelligence and study actively aware of the importance,the initiative of development,and put forward by the scholars both at home and abroad related theories and achievements,and applied to mobile robot in the field of fire rescue and existing problems of the traditional methods are detailed analyzed.2.Defines the research purpose of active perception,put forward the active perception system model,analyzes the active and passive sense perception,perception and feedback control of the differences and relations actively,summarizes the active perception problem need to solve two aspects,and systematically discusses the robot operating system(ROS),architecture,and is the main feature of data processing and simulation software tools.3.The active perception model based on bayesian inference is studied,and the state estimation method of robot based on bayesian filtering is proposed,including motion distortion processing,back-end nonlinear processing method,and the idea of improved particle filter,and the simulation experiment analysis is given to realize the design of state estimator in active perception.4.This section discusses the active perception exploration strategy based on reinforcement learning method,including the basic introduction of reinforcement learning,the mathematical definition of markov process,the implementation process of q-learning algorithm,and the design of simulation experiment,and realizes the active exploration of robot in unknown environment by combining with the state estimator of bayesian inference.5.The active perception method for the robot team to quickly traverse all targets under the known environment information is analyzed,including the cooperative mode of the robot team,the mathematical model of multi-traveler problem,the basic principle and realization method of self-organizing mapping.6.The architecture and experimental design of the robot system are discussed,including hardware selection,software design and practical test.