Activity Recognition for Dynamic Multi-Agent Teams

被引：4

作者：

Sukthankar, Gita ^{[1
]}

Sycara, Katia ^{[2
]}

机构：

[1] Univ Cent Florida, Dept EECS, Orlando, FL 32816 USA

[2] Carnegie Mellon Univ, Inst Robot, Pittsburgh, PA 15213 USA

来源：

ACM TRANSACTIONS ON INTELLIGENT SYSTEMS AND TECHNOLOGY | 2012年 / 3卷 / 01期

关键词：

Algorithms; Experimentation; Performance; Activity recognition; plan recognition; multi-agent systems; teamwork;

D O I：

10.1145/2036264.2036282

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

This article addresses the problem of activity recognition for dynamic, physically embodied agent teams. We define team activity recognition as the process of identifying team behaviors from traces of agent positions over time; for many physical domains, military or athletic, coordinated team behaviors create distinctive spatio-temporal patterns that can be used to identify low-level action sequences. This article focuses on the novel problem of recovering agent-to-team assignments for complex team tasks where team composition, the mapping of agents into teams, changes over time. We suggest two methods for improving the computational efficiency of the multi-agent plan recognition process in these cases of changing team composition; our proposed approach is robust to sensor observation noise and errors in behavior classification.

引用

页数：24