Source-linked AI summary
Context-aware Sensor Search, Selection and Ranking Model for Internet of Things Middleware
Charith Perera, Arkady Zaslavsky, Peter Christen, Michael Compton, Dimitrios Georgakopoulos
TL;DR
CASSARAM addresses the challenge of selecting sensors as their numbers grow and users need to meet mandatory and flexible requirements. It combines user priorities with semantic querying and quantitative ranking; evaluation identifies semantic querying as a major processing-time cost, and a prototype demonstrates context-based selection.
Problem
As available sensors increase, selecting those that efficiently provide data relevant to a problem requires searching by context and distinguishing mandatory from flexible requirements.
Source: §III · p. 3Method
CASSARAM uses user priorities and sensor characteristics to search and select sensors, combining semantic requirements with priority-based quantitative techniques.
Source: §III · p. 4, §V · p. 6Results
Semantic querying takes significantly more processing time than indexing and ranking, while CPHF saves some indexing and ranking time and returns fewer sensors.
Source: §VII · pp. 7–8Takeaways & Limitations
The prototype demonstrates selecting sensors suited to user expectations and priorities using context information about sensors.
Source: §VIII · p. 8The MySQL implementation can join only 61 tables, limiting SSNO queries to a maximum of 10 context properties.
Source: §VII · p. 7
Abstract
from arXiv · showhide
As we are moving towards the Internet of Things (IoT), the number of sensors deployed around the world is growing at a rapid pace. Market research has shown a significant growth of sensor deployments over the past decade and has predicted a substantial acceleration of the growth rate in the future. It is also evident that the increasing number of IoT middleware solutions are developed in both research and commercial environments. However, sensor search and selection remain a critical requirement and a challenge. In this paper, we present CASSARAM, a context-aware sensor search, selection, and ranking model for Internet of Things to address the research challenges of selecting sensors when large numbers of sensors with overlapping and sometimes redundant functionality are available. CASSARAM proposes the search and selection of sensors based on user priorities. CASSARAM considers a broad range of characteristics of sensors for search such as reliability, accuracy, battery life just to name a few. Our approach utilises both semantic querying and quantitative reasoning techniques. User priority based weighted Euclidean distance comparison in multidimensional space technique is used to index and rank sensors. Our objectives are to highlight the importance of sensor search in IoT paradigm, identify important characteristics of both sensors and data acquisition processes which help to select sensors, understand how semantic and statistical reasoning can be combined together to address this problem in an efficient manner. We developed a tool called CASSARA to evaluate the proposed model in terms of resource consumption and response time.
I. INTRODUCTION
As sensor numbers grow, choosing suitable sensors becomes difficult, while collecting data from all available sensors may be infeasible and may not add value. CASSARAM combines semantic querying with quantitative reasoning so users can search and select sensors according to their priorities.
I. INTRODUCTION
- Growing sensor deployments make appropriate sensor selection challenging, time-consuming, and potentially infeasible at scale and cost.Collecting and storing data from all available sensors may not add value or solve efficient processing challenges.Source: §I · p. 1
- Text-based search cannot accurately capture critical sensor characteristics, while metadata approaches face interoperability problems and manual-entry errors.A study cited by the paper reports 45 mistakes by 20 participants entering weather-station metadata; duplicated entry across middleware can also be required.Source: §I · p. 1
- CASSARAM lets users prioritize sensor characteristics and combines semantic querying with quantitative reasoning to search and select appropriate sensors.The paper notes that SSNO ontologies describe sensors and relationships, support semantic querying, and can reduce error-prone manual description.Source: §I · p. 1
II. BACKGROUND AND RELATED WORK
CA4IOT envisions automated, context-aware configuration of filtering, fusion, and reasoning, with CASSARAM addressing the selection of suitable sensors. Prior middleware and search systems offer narrower search functions, while sensor selection is framed as a growing challenge as IoT scales.
II. BACKGROUND AND RELATED WORK
- CA4IOT envisions automating context-aware configuration of filtering, fusion, and reasoning, contrasting with static, a priori resource configuration.CASSARAM is positioned as the component that helps identify suitable sensors within this vision.Source: §II · p. 2
- For environmental pollution monitoring, multiple sensor types and runtime data fusion are needed, making manual sensor and operator selection complex in a dynamic IoT setting.The scenario includes land, air, and water pollution, with high-level context potentially requiring data from multiple sensors.Source: §II · p. 2
- CA4IOT must choose among alternatives with differing characteristics; one example is selecting 100 temperature sensors from 25,000 available in an area.The paper notes differences such as accuracy, reliability, and cost.Source: §II · p. 2
- Existing middleware provides constrained search: LSM offers location and type selection through SPARQL, GSN uses a sensor list, and SensorMap and COSM offer location, type, or keyword search.The paper presents CASSARAM as complementing these middleware solutions with more extensive sensor search and selection functionality.Source: §II · pp. 2–3
- Prior sensor-search work includes fuzzy similarity, location-tree search, content-based search, and Dyser’s statistical state prediction; few efforts focus on context information.The paper also compares web-service and sensor selection domains, arguing that their similarities make sensor selection comparably challenging and that IoT scale will increase its complexity.Source: §II · p. 3, Table I · p. 3
III. PROBLEM DEFINITION AND MOTIVATION
Growing sensor deployments and limited middleware search make choosing suitable sensors a challenge. CASSARAM frames selection around mandatory requirements and flexible priorities, using sensor context to support search and ranking.
III. PROBLEM DEFINITION AND MOTIVATION
- Sensor selection has two requirement types: point-based requirements must be fulfilled, while proximity-based requirements allow user-prioritized compromises.Temperature and location illustrate mandatory requirements; accuracy and reliability illustrate flexible priorities.Source: §III · p. 3
- CASSARAM is proposed as a middleware-complementary model for searching and selecting sensors according to user priorities.It aims to return sensors that provide required data and have preferred characteristics.Source: §III · p. 4
- The proposed context framework models sensor characteristics for context-based search and extends SSNO compatibly with existing SSNO-based developments.The extension is envisioned for projects such as Phenonet and OpenIoT.Source: §III · p. 4
- CASSARA lets users express priorities comparatively with sliders and presents sensors in ranked order.Source: §III · p. 4
- The Comparative Priority-based Heuristic Filtering algorithm is proposed to make sensor indexing faster and more efficient.Source: §III · p. 4
- The authors evaluate CASSARAM with a prototype tool, measuring computational resource consumption, performance, and efficiency.Source: §III · p. 4
IV. REAL WORLD CHALLENGE
Agricultural monitoring across remote plots motivates sensor search based on context rather than collecting data from every deployed sensor. The example also highlights selecting sensors to conserve energy and extend network operation.
IV. REAL WORLD CHALLENGE
- Collecting data from sensors across all plots all the time is described as inefficient, motivating context-based sensor search.The agricultural monitoring challenge is to monitor crop performance and growing conditions and make the information accessible.Source: §IV · p. 4
- Australian grain breeders plant up to 1 million 10m2 plots, often more than four hours’ travel one-way away, to identify high-yielding wheat and barley varieties.The domain also faces water scarcity and low soil fertility.Source: §IV · p. 4
- Collecting data only from sensors with more energy helps run the entire network longer without reconfiguration.Source: §IV · p. 5
V. CASSARAM: THE PROPOSED APPROACH:
CASSARAM combines exact point-based requirements with priority-weighted proximity ranking to select sensors. It also proposes heuristic filtering to reduce the work required before ranking large sensor sets.
V. CASSARAM: THE PROPOSED APPROACH:
- CASSARAM separates exact, non-negotiable point requirements from negotiable proximity requirements, selecting sensors closest to the latter.Point requirements can be expressed in SPARQL; any context property can be designated as a point requirement.Source: §V · p. 5
- Users set comparative priorities through sliders and can exclude a property; higher priority gives that property a higher weight in ranking.A property left unchecked is not considered in CPWI.Source: §V · p. 6, Figure 5 · p. 6
- Sensor descriptions and context properties use SSNO, with an extensible framework covering properties such as accuracy, reliability, battery life, and cost.The framework can be extended with additional context attributes.Source: §V · p. 6
- CPWI ranks sensors by weighted Euclidean distance from a user-defined or default ideal point in a multidimensional context-property space.Sensors closer to the user requirement rank higher.Source: §V · p. 6, Figure 6 · p. 6
- CPHF reduces ranking work by repeatedly removing bottom-ranked sensors across context properties before indexing and ranking the survivors.The example removes 40% by the highest-weighted property and 30% by the next-highest before repeating for remaining properties.Source: §V · pp. 6–7
VI. IMPLEMENTATION AND EXPERIMENTATIONS
The authors evaluated CASSARAM with the CASSARA prototype using real and synthetically generated sensor data. Their setup assumes sensor descriptions and context data are available, while leaving distributed search outside the evaluation.
VI. IMPLEMENTATION AND EXPERIMENTATIONS
- The CASSARA prototype used Java, MySQL, Apache Jena, and Apache Commons Mathematics, and experiments were averaged over 100 runs.Testing used an Intel Core i5-2557M 1.70GHz computer with 4GB RAM.Source: §VI · p. 7
- The CASSARA interface captures context-property preferences, slider priorities, ideal values, and SPARQL point requirements, then displays ranked sensors.Users can adjust slider scale and sensitivity.Source: Figure 7 · p. 7
- Synthetic data supplied context attributes absent from available public real-world datasets, allowing control over data characteristics for different scenarios.The authors combined real and synthetic data to create larger datasets and vary sensor context information.Source: §VI · p. 7
- The work assumes sensor descriptions and per-sensor context data are already available to software systems; distributed sensor search is left for future work.The authors note that sensor data may be distributed across systems such as GSN instances.Source: §VI · p. 7
- The evaluation used a dataset of one million sensor descriptions and context records, synthetically assembled from multiple real datasets.Sources included Bureau of Meteorology environmental linked data, Phenonet, and Linked Sensor Middleware datasets.Source: §VI · p. 7
VII. EVALUATION AND DISCUSSION
Evaluation identifies semantic querying as the largest processing-time component, while CPHF reduces processing and resource use at an accuracy trade-off as sensor scale grows.
VII. EVALUATION AND DISCUSSION
- Semantic querying takes significantly longer than indexing and ranking, and execution time rises significantly as a query retrieves more context properties.MySQL's 61-table join limit restricts retrieval to at most 10 context properties in the SSNO data structure; Figure 8 compares processing times across steps.Source: §VII · p. 7, Figure 8 · p. 8
- Indexing is faster with fewer than 10,000 sensors, while processing time starts increasing significantly after 100,000 sensors.Figure 9 measures indexing time as sensor and context-property counts increase, using synthetic context properties.Source: §VII · p. 8, Figure 9 · p. 8
- CPHF accuracy increases as margin of error M increases, whereas lower M reduces resource consumption, creating an accuracy-resource trade-off.Figure 11 examines accuracy as M changes while the number of sensors increases; the authors suggest dynamically learning an optimum M from user priorities and normalized weights.Source: §VII · p. 8, Figure 11 · p. 8
- CPHF returns a limited sensor set and saves time in indexing and ranking, while the non-CPHF approach returns all available sensors and uses more processing time, memory, and temporary storage.Figure 10 compares selection time with and without CPHF; the experiments hold the requested sensor count at 50.Source: Figure 10 · p. 8, §VII · p. 8
- The authors identify closer integration of semantic and quantitative reasoning as future work.They also propose investigating multiple queries as an alternative that can be efficient with fewer sensors.Source: §VII · p. 8
VIII. CONCLUSIONS AND FUTURE WORK
The conclusion presents context-based, priority-driven sensor selection and a working prototype, with middleware integration and closer reasoning integration left for future work.
VIII. CONCLUSIONS AND FUTURE WORK
- CASSARAM uses sensor context information together with user expectations and priorities to search for and select sensors suited to user requirements.The authors built a working prototype to demonstrate the model in realistic applications.Source: §VIII · p. 8
- The authors plan to incorporate CASSARAM into GSN, SensMA, and OpenIoT to support automated sensor selection in distributed environments.They plan further evaluations to understand how automated selection would complement middleware solutions.Source: §VIII · p. 8
- Future work will investigate how semantic and quantitative reasoning can work more closely together to improve efficiency and functionality.The paper motivates prioritizing operational and conceptual sensor context properties when similar sensors are available.Source: §VIII · pp. 8–9