Introduction The problem of behavior recognition has been an active research topic for a long time (Kautz 1987; Charniak & Goldman 1993; Bui 2003) and still remains very challenging. While most early systems were focused on simulations or toy examples, more recent research has begun to build behavior recognition systems that work with realworld data (Pollack et al. 2002; Patterson et al. 2004a; Liao, Fox, & Kautz 2004). There are three main technical advancements that have made these systems possible. First, various sensing technologies such as the Global Positioning System (GPS), Radio Frequency Identification (RFID) tags, digital cameras, ultrasound sensors, infrared sensors, light sensors, and motion sensors are maturing and becoming widely used. Thus, it becomes much easier for computers to sense the physical world. Second, great advances have been achieved in probabilistic reasoning, especially for large and complex systems. Third, large amount of background knowledge can be acquired through the World Wide Web and other related technologies. Behavior recognition is one of the central pieces in the Assisted Cognition project at University of Washington (Kautz et al. 2002). By combining Artificial Intelligence and ubiquitous computing technologies, the Assisted Cognition project aims to augment human capabilities, with a particular emphasis on increasing the independence of people suffering from cognitive limitations, e.g., Alzheimer’s disease patients. In this paper, we explain our recent progress on probabilistic behavior recognition and discuss future research directions. In particular, we focus on the outdoor activities; see (Patterson et al. 2004a) for more about indoor behavior recognition. Besides behavior recognition, we must also develop effective intervention strategies. For example, when and how should the system interact with users when it detects errors? In principle, we shall take a decision-theoretic perspective and compute the expected costs and benefits under uncertainty about the world and user state at hand. See (Kautz et al. 2003) for more discussions on this topic.
[1]
Hung Hai Bui,et al.
A General Model for Online Probabilistic Plan Recognition
,
2003,
IJCAI.
[2]
Henry A. Kautz.
A formal theory of plan recognition
,
1987
.
[3]
Avi Pfeffer,et al.
Probabilistic Frame-Based Systems
,
1998,
AAAI/IAAI.
[4]
Henry A. Kautz,et al.
An Overview of the Assisted Cognition Project
,
2002
.
[5]
Ben Taskar,et al.
Learning Probabilistic Models of Relational Structure
,
2001,
ICML.
[6]
Robert P. Goldman,et al.
A Bayesian Model of Plan Recognition
,
1993,
Artif. Intell..
[7]
Henry Kautz,et al.
Contextual computer support for human activity
,
2004,
AAAI 2004.
[8]
Michael P. Wellman,et al.
Probabilistic grammars for plan recognition
,
1999
.
[9]
Ben Taskar,et al.
Discriminative Probabilistic Models for Relational Data
,
2002,
UAI.
[10]
Krzysztof Z. Gajos,et al.
Opportunity Knocks: A System to Provide Cognitive Assistance with Transportation Services
,
2004,
UbiComp.
[11]
Pedro M. Domingos,et al.
Dynamic Probabilistic Relational Models
,
2003,
IJCAI.
[12]
Matthai Philipose,et al.
Mining models of human activities from the web
,
2004,
WWW '04.
[13]
Martha E. Pollack,et al.
Autominder: A Planning, Monitoring, and Reminding Assistive Agent
,
2002
.
[14]
Henry Kautz,et al.
Foundations of Assisted Cognition Systems
,
2003
.
[15]
Henry A. Kautz,et al.
Inferring High-Level Behavior from Low-Level Sensors
,
2003,
UbiComp.