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:
- 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.
- 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 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.
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