Research On Multi-agent Interaction Design In Task Transfer Scenarios

In recent years,with the accelerated advancement of the Internet of Things,artificial intelligence and robotics,more and more smart devices have anthropomorphic communication capabilities and highly autonomous action capabilities,and interact with users in the form of a certain agent.However,when there are more and more such smart devices around people,multi-device collaboration and cross-device interaction are becoming more common.Especially in a task transfer scenario,the user may need to interact with multiple agents at the same time.Under this circumstance,what kind of information exchange mechanism should be established between multi-agents and users,and what kind of role relationship should be established between multi-agents,these issues remain to be studied and considered.This research analyzes various agent cases in academic research and commercial applications,compares the technical characteristics of different forms of agents,discusses the design trend of multi-agent interaction,and sorts out different types and the interaction process of transfer scenarios,the significance of the multi-agents information communication process,the information communication mode of multiple agents and related concepts.Aiming at the research scenario of multi-agent task transfer,this thesis constructs a characteristic model of multi-agent interaction from the two variable dimensions of the agent's social identity and the user's focus of attention,and divides four typical information communication modes(Operational,Migratable,Representative and Social),and output the corresponding interaction design plan.In order to verify the multi-agent communication model proposed by the research,the author independently designed and developed a multi-agent interactive experiment platform,and organized 55 samples of multi-agent task transfer interactive experiments.The experimental platform with various forms of agents such as physical social robots and virtual intelligent agents can switch different preset information communication modes for task information exchange,so that multiple intelligent devices can cooperate to complete tasks.The platform has the new features of high ease of use,diversified control,diversified devices and high scalability.The experiment evaluates the four agent information communication modes from the three dimensions of usability,trust and social attributes,and analyzes the differences in the information communication modes of different agents in task transfer scenarios.Experiments show that in the task transfer scenario,the Migratable mode outperforms the other three types of modes,and the establishment of a multi-agent role relationship consistent with user expectations is the key to building a good interactive experience.These results and findings have reference significance for the research of multi-agent interaction design.