Font Size: a A A

Collaborative Rescue Research Based On Multi-agent

Posted on:2019-06-18Degree:MasterType:Thesis
Country:ChinaCandidate:G F MaFull Text:PDF
GTID:2428330548970110Subject:Computer Science and Technology
Abstract/Summary:PDF Full Text Request
In recent years,catastrophes such as natural disasters,terrorist attacks,leakage of dangerous goods,and explosions have frequently occurred in the world.These events have the features of suddenness,great destructive power,far-reaching influence,and difficulty in rescue,which often cause serious harm to human life and property safety.Therefore,issues such as disaster prevention,pre-disaster warning,and post-disaster relief have aroused widespread concern in the society.With the increasing perfection of Multi-Agent System(MAS)theory and the continuous improvement of simulation technology,it has provided powerful support for domestic and foreign experts and scholars to study post-disaster relief issues.The use of MAS theory and simulation technology to research on post-disaster rescue issues is mainly focused on the study of collaboration between Agents.There are two kinds of collaboration methods: cooperation between homogeneous Agents and cooperation between heterogeneous Agents.Because the research on the strategies,methods and rules of cooperation between homogeneous and heterogeneous Agents in the static environment is relatively mature,the collaboration between Agents under dynamic and complex environments is the current research hotspot.This paper builds a simulation platform in Unity to focus on the issues related to Agent's collaborative rescue in a disaster environment.In order to improve the cooperation efficiency between Agents and ensure that the group completes collaborative rescue tasks efficiently,the main contributions are as follows:(1)With regard to the problem of collaborative information filtering among individuals in MAS,We propose to use influence model to process information such as environment,individuals and tasks in the scene.The process of filtering collaborative information is divided into the following three steps: Firstly,the search Agent of the discovery task uses the first level of the influence function to obtain the type of disaster or task and then determines the number of affected people,the number of damaged houses,and the affected area.Secondly,according to the type of task,the corresponding disaster categories in the second level of the influence function are called to filter the perceived information,get specific information about collaboration tasks.This will not only ignore the invalid information while reducing the amount of collaboration information between Agents,but also improve the efficiency of the system to ensure the efficient completion of group collaborative rescue mission.Finally,the individuals further processed and stored the filtered information to ensure the uniqueness of the collaborative information.(2)Aiming at the correctness of the cooperation strategy between Agents,a state prediction algorithm is proposed.First,the state prediction function to predict the task,individual,environment and other follow-up states.Second,according to the forecast results,current information,relevant rules,etc,the algorithm to formulate preliminary cooperation strategies and determine the team command,and pass the cooperation strategy to the free individuals.Finally,the Agent that accepts the collaboration task modifies the collaboration strategy in time according to the action effect and the trigger class.On the premise of guaranteeing the correctness of the collaboration strategy,the dynamic adjustment of the collaboration strategy is implemented to ensure the successful completion of the collaboration task.(3)Combining the two major innovations,a Multi-Agent collaborative rescue algorithm is proposed and used in practical applications.Building a simulation scenario in U3 D and apply the proposed collaborative rescue algorithm to multi-agent disaster collaboration rescue;The proposed collaborative rescue algorithm was applied to multi-agent disaster recovery collaborative rescue,and the effectiveness of the algorithm was analyzed and demonstrated by the cooperation instances between agents.Through the research of heterogeneous Agent collaboration in disaster environment,the theory of Multi-Agent cooperation and rescue is enriched and perfected,which provides the theoretical foundation and decision-making support for the real disaster rescue work and expands the application field of multi-agent.
Keywords/Search Tags:Distributed Artificial Intelligence, Multi-Agent System, Collaboration and Rescue, Influence degree function, State prediction, Collaboration strategy, Unity 3D simulation
PDF Full Text Request
Related items