Adaptive Sensor Allocation for Event Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge lies in efficiently allocating sensor devices to detect events within a domain while balancing the need to explore unknown areas and exploit known event rates, as the cost of operation increases with the number of devices and the extent of the region monitored, leading to an exploration-exploitation dilemma.
Innovation Solution
A computer-implemented method that iteratively determines portions of a domain, estimates normalized event rates, selects subdomains based on expected rewards penalized by sensing costs, and dynamically allocates sensor devices to these subdomains, progressively increasing the number of portions to refine event detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If sensor devices are allocated to subdomains where the most events are expected to occur, then event detection efficiency is improved, but the system fails to gather event data for other portions of the domain, resulting in loss of information about event rates in unexplored areas
Solution Approach 1:
The domain is divided into multiple portions or subdomains, allowing sensor devices to be allocated to specific segments. This segmentation enables the system to focus detection efforts on high-probability areas while maintaining the ability to explore other segments, thus balancing exploitation of known event rates with exploration of unknown areas.
Solution Approach 2:
The allocation of sensor devices to subdomains is dynamically adjusted based on accumulated event data. As the system learns event rates in different portions of the domain, it adaptively reallocates sensors to optimize detection efficiency while ensuring adequate coverage of previously unexplored areas, resolving the exploration-exploitation trade-off over time.
2Area of stationary object
If sensor devices are allocated to excessively large subdomains or subdomains with too much variance, then coverage of the domain is improved, but the cost of operating the sensor devices increases
Solution Approach 1:
Different subdomains are allocated to sensor devices based on their specific characteristics and event rates. Instead of uniform allocation, the system tailors the allocation strategy to local conditions, assigning sensors to subdomains where they can achieve effective coverage with minimal operational cost. This localized optimization prevents wasteful deployment in low-probability areas while maintaining adequate coverage overall.
3Reliability
If the number of deployed sensor devices is increased to detect events throughout the entire domain, then event detection coverage is improved, but the operational cost increases
Solution Approach 1:
The system deploys sensor devices to detect events in a subset of the domain rather than attempting to monitor the entire domain simultaneously. By identifying and focusing on subdomains where events are most likely to occur, the system achieves reliable event detection for the most critical areas while avoiding the prohibitive costs of comprehensive domain-wide coverage.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
There is provided a computer-implemented method for arranging one or more sensor devices for detecting events within a domain having one or more dimensions. The method includes iteratively: determining a plurality portions spanning the domain; estimating a normalised rate of events being detected in each portion; selecting, in dependence on the estimated normalised rates, one or more of the determined portions on the basis of an expected reward associated with allocating the one or more sensor devices to the selected one or more portions; arranging the one or more sensor devices to detect events in the selected one or more portions; and receiving event data from the one or more sensor devices, the event data being indicative of dimensional co-ordinates of events detected by the one or more sensor devices within the selected one or more portions. The expected reward depends on an expected rate of events being detected by the one or more sensor devices, penalised by a sensing cost associated with the one or more sensor devices, and the number of determined portions increases with respect to the number of iterations in accordance with a predetermined relationship.