Boosting-Based Clustering for Optimal Sensor Location Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining optimal locations for sensors or small cells in wireless communication systems, such as those in smart cities, are computationally complex and inefficient, especially for large-scale IoT scenarios, and fail to account for dynamic network changes, leading to inaccurate or prolonged computation times.
Innovation Solution
A method using boosting-based clustering techniques, where random clusters are iteratively weighted to optimize sensor placement, employing distance measures like Euclidean and Mahalanobis, and information metrics like Fisher's information, to select the best set of locations for devices while minimizing computational resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If existing optimization procedures are used to select sensor positions, then information maximization is achieved, but computation time becomes excessively long for large-scale IoT scenarios
Solution Approach 1:
The patent segments the large-scale sensor selection problem into smaller, manageable clusters of sensor positions. By dividing the entire set of potential sensor locations into multiple clusters and processing them separately, the computational complexity is reduced while still achieving comprehensive coverage and information maximization across the entire area.
Solution Approach 2:
The patent implements a dynamic approach where sensor position selections can be updated and adjusted based on changing network conditions and information requirements. The system allows for iterative refinement of sensor placements, enabling adaptation to dynamic environments without requiring complete re-computation from scratch.
2Loss of information
If a huge number of sensors are deployed to provide detailed information, then information quality improves, but cost and radiation exposure increase
Solution Approach 1:
The patent applies partial action by selecting only the essential and most informative sensor positions rather than deploying sensors at all potential locations. The clustering algorithm identifies a subset of optimal positions that provide sufficient information quality for monitoring purposes, avoiding the excessive deployment of unnecessary sensors that would increase radiation exposure and cost.
Solution Approach 2:
The patent changes the selection criteria parameters from considering all possible sensor positions to evaluating positions based on information metrics and clustering characteristics. By adjusting the selection parameters to focus on information quality and spatial distribution rather than sheer number of sensors, the system achieves effective monitoring with reduced sensor count and lower radiation exposure.
3Measurement precision
If MINLP algorithm is used to solve sensor placement optimization, then optimal locations are identified, but the algorithm is complex and difficult to update when network changes
Solution Approach 1:
The patent uses clustering algorithms that create representative copies or prototypes of sensor position groups rather than solving the complete optimization problem directly. Each cluster represents a group of similar positions, and selecting one position per cluster provides a simplified yet effective approximation of the optimal solution, reducing algorithmic complexity while maintaining practical accuracy.
Solution Approach 2:
The patent employs computationally inexpensive clustering algorithms that can be quickly executed and discarded, rather than relying on complex MINLP solvers that require extensive computation time. The clustering approach provides sufficiently good solutions for practical deployment without the heavy computational burden of exact optimization methods, enabling easier updates when network conditions change.
4Measurement precision
If manual updates are performed every time network relations change, then optimal positions are maintained, but monitoring and updating becomes very hard and time-consuming
Solution Approach 1:
The patent implements self-service by enabling the sensor position selection system to automatically adapt to network changes through iterative clustering algorithms. The system can re-evaluate and update sensor position selections based on current network conditions without requiring manual intervention, automatically maintaining optimal positions as the network evolves.
Solution Approach 2:
The patent employs periodic updates of sensor position selections based on changing network conditions. Rather than requiring continuous manual monitoring and adjustment, the system performs periodic re-evaluations using clustering algorithms, automatically updating sensor positions at appropriate intervals to maintain optimization accuracy while minimizing operational effort.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method and a determining unit (110) for determining a set of locations (m1-m8) for a number of devices (n1-n5) are disclosed. For each distance measure of multiple distance measures, the determining unit (110) generates a respective pair of clusters (c1, c2) by means of a clustering method. The determining unit (110) determines information metrics. The determining unit (110) performs some first actions, comprising identifying a replaceable respective pair based on the information metrics, calculating a weighted pair based on pairs of clusters in combination with the information metrics, and determining a further respective information metric for said weighted pair. While difference between the respective information metric and the further respective information metric is greater than a threshold, the determining unit (110) performs some second actions, comprising replacing said replaceable pair with said weighted pair, replacing the respective information metric for said replaceable pair with said further respective information metric for said weighted pair, and performing said first actions. The determining unit (110) selects the set of locations based on the weighted pair of clusters. A corresponding computer program (303) and a computer program carrier (305) are also disclosed.