Study Of The Intelligence Perception And Behavior Coordination Framework Of Multi-robot System For Typical Tasks

How to coordinate the various behaviors of multi-robots effectively and reasonably is a key problem of cooperative control of multi-robot colony.The model based on "perception modeling,planning and movement" control architecture is a feasible solution for unmanned cluster cooperative control.The premise of cooperative action of multi-robot colony is to have the ability to collect and recognize all kinds of relevant information in battlefield space,understand the situation of battlefield and guide its decision-making and action.This paper develops an integrated design of hardware and software for active spectrum intelligent sensing for typical combat tasks.For the data collection and processing modules in integrated design,an intelligent hyperspectral data visualization processing platform is constructed,which provides two main functions of target census and specific target detailed inspection in complex environment,and realizes real-time visualization processing of hyperspectral data.As for the role management and task description and management module in integrated design,the collaborative framework of unmanned group behavior is optimized in the aspects of basic model,execution mechanism,collaboration mechanism,organization mechanism and communication mechanism.What's more,tuypical experiments are designed to evaluate the performance and practicability of the development framework.The main works are listed as follows.1.Integration of software and hardware for typical task-oriented active spectrum intelligent sensing is designed,which combines the rich information advantages of hyperspectral data,the fine resolution advantages of visible images and the temperature imaging advantages of infrared images.In the sensing system,multi-sensor data fusion and intelligent target recognition based on role management are realized,and the sensing data are processed by combining depth-learning algorithm.By processing the spatial and spectral characteristics of the target simultaneously,different functional,such as target detection,target recognition and anomaly warning,are realized.All the modules together make the sensing system to reveal camouflage intelligently and find high-value targets quickly,which enhance the overall perception of the environment for the unmanned system.(Chapter 3)2.Based on the he data collection and processing modules in integrated design,an intelligent hyperspectral data processing platform is constructed,with improve of the real-time performance of the data fusion algorithm.A hyperspectral depth tensor convolution neural network model consisting of multi-scale convolution layer,characteristic cube,full convolution layer,gradient attenuation control and over-fitting simulation is developed.Data mining for hyperspectral data cube after preprocessing is realized,which can call spectral characteristic data to achieve fast target census,target detailed investigation and other functions.(Chapter 4)3.To solve the problem of role management and task description and management module in integrated design,a typical task-oriented cooperative development framework for unmanned group behavior is developed,which supports both single robot autonomous behavior cooperation and multi-robot cooperative task cooperation.ALLIANCE model extends it to multi-robot systems,which is a fully distributed,fault-tolerant model.Experiments are carried out in simulation and real environment to verify and optimize the framework's support for single robot autonomous behavior coordination mechanism and multi-robot task coordination mechanism.(Chapter 5)...