Object Tracking Using Durative Events

Eiko Yoneki, Jean Bacon

Abstract

Traditional wireless sensor network (WSN) middleware often struggles with high data volumes and the complexity of implementing high-level logic. This paper introduces a distributed middleware architecture that transitions from a data-centric model to one based on "event knowledge." By utilizing durative events—occurrences treated as time intervals rather than instantaneous points—the proposed system significantly reduces network traffic while enhancing the contextual accuracy of object tracking in sentient computing environments.

Introduction

Sentient computing environments, powered by Wireless Sensor Networks (WSNs), typically produce an overwhelming amount of raw data. Conventional middleware approaches treat these sensors as a database to be queried. However, this "data-centric" model faces significant hurdles:

  1. Logic Complexity: Expressing complex spatial-temporal conditions (e.g., "alert if two specific individuals are in a corridor for over five minutes") is difficult with standard query languages.
  2. Data Volume: Transmitting every raw sensor "ping" to a central server creates massive bottlenecks and consumes excessive bandwidth.

To address these issues, this research advocates for a shift toward viewing events as durative states (such as "occupying a room") rather than instantaneous "ticks." This allows for sophisticated filtering and aggregation at the network's edge.

Methodology

The proposed distributed middleware is built upon a Service-Oriented Architecture (SOA), designed to decouple physical sensing from high-level application logic.

The layered middleware architecture, featuring the Event Broker and Service layers for distributed processing.

The architecture consists of several key layers:

  • Layered Decoupling: From physical sensors up to an Event Broker Layer and a Service Layer (implemented via the OSGi framework), this structure allows heterogeneous sensors to collaborate seamlessly.
  • Durative Event Model: Composite events are defined using interval-based timestamps rather than single points in time. This "interval semantics" approach is better suited for real-time monitoring in wireless environments.
  • Event Algebra: A parameterized algebra—including operators for conjunction, sequence, negation, and spatial/temporal restrictions—enables the definition of complex, high-level events.
  • Distributed Gateways: Instead of centralizing all processing, gateway nodes are deployed at specific locations (e.g., rooms). These gateways perform local correlation and filtering, transmitting only significant state changes.

Experimental Setup

The system was evaluated using real-world data from the Active BAT system at the University of Cambridge.

  • Hardware: The Active BAT system utilizes ultrasound trilateration to track tags with approximately 3cm accuracy.
  • Dataset: The study analyzed 24 hours of tracking data from May 20, 2003, involving 21 different tag holders.
  • Scale: The raw dataset comprised roughly 400,000 individual ultrasound pings.

Results and Discussion

The implementation of durative event tracking yielded several significant improvements over traditional methods.

Distribution of events across different office locations, highlighting the system's ability to track occupancy states and reduce noise.

Data Reduction and Efficiency

The most notable result was a dramatic reduction in the number of events transmitted across the network. By shifting the focus to state changes (e.g., entering or leaving a room) rather than raw pings, the system filters out local "noise" at the gateway level. For instance, minor positional fluctuations that do not change a person's room status are suppressed, preventing unnecessary network traffic.

Contextual Accuracy and Complex Detection

The system successfully identified specific user behaviors, such as tracking "Andy" primarily in his office (SN04) and "Brian" in SN10. Furthermore, the middleware effectively handled sophisticated spatial-temporal queries, such as detecting when two users occupied the same machine room or corridor simultaneously.

Conclusion and Future Work

While the prototype demonstrates the power of durative event middleware, several areas for future development remain. The high-level language for defining service compositions is still evolving. Additionally, maintaining perfect time synchronization across distributed gateways is a known challenge for temporal ordering in real-world deployments. Future work will also focus on implementing the architecture across a wider variety of communication protocols and timestamping environments.

Summary updated

Abstract

This paper presents a distributed middleware architecture based on a service-oriented approach, to manage high volume sensor events. Event management takes a multi-step operation from event sources to final subscribers, combining information collected by wireless devices into higher-level information or knowledge. An event correlation service provides sophisticated event filtering, aggregation and correlation over time and space in heterogeneous network environments. An experimental prototype in the simulation environment with real world data produced by the Active BAT system is shown.
BibTeX
@inproceedings{yoneki2005object,
  author = {Eiko Yoneki and Jean Bacon},
  title = {{Object Tracking Using Durative Events}},
  booktitle = {EUC Workshops},
  year = {2005},
  doi = {10.1007/11596042\_68},
}
Figure