Intention Recognition for Multiple Agents

Abstract

Intention recognition is an important step to facilitate collaboration among multiple agents. Existing work mainly focuses on intention recognition in a single-agent setting and uses a descriptive model, e.g. Bayesian networks, in the recognition process. In this article, we develop a new approach of identifying intentions for multiple agents through a clustering algorithm. We first define an intention model for multiple agents of interest. We use a prescriptive approach to model agents' behaviours where their intentions are hidden in the implementation of their plans. We introduce landmarks into the behavioural model therefore enhancing informative features to identify common intentions for multiple agents. Subsequently, we further refine the model by focusing only action sequences in their plan and provide a light model for identifying and comparing their intentions. The new models provide a simple approach of grouping agents' common intentions upon partial plans observed in agents' interactions. Then, we transform the intention recognition into an un-supervised learning problem and adapt a clustering algorithm to group intentions of multiple agents through comparing their behavioural models. We conduct the clustering process by measuring similarity of probability distributions over potential landmarks in intention models so as to discover agents' common intentions. Finally, we examine the new intention recognition approaches in two problem domains. We demonstrate importance of recognising common intentions of multiple agents in achieving their goals and provide experimental results to show performance of the new approaches.