Sensor Placement Optimization via Activity Relevance Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sensor placement technologies for ambient and physiological sensing face challenges in optimizing the number and location of sensors, leading to increased costs and irrelevant data streams, which obscure salient data and hinder predictive analysis in activities of daily living and physiological monitoring.
Innovation Solution
A system and method that determine an optimal sensor arrangement by generating models for activities using time-series observations from multiple sensors, calculating likelihoods, and calculating pair-wise distances to identify the relevance of each sensor for capturing specific activities, allowing for automatic identification of a minimal set of salient sensors using regularized stochastic generative-discriminative encoding and hidden Markov models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If sensors are placed in an ad-hoc manner for ambient and physiological sensing, then sensor deployment is simplified, but the system cost increases and irrelevant data streams obscure salient data
Solution Approach 1:
The system performs preliminary analysis of sensor data relevance before final deployment configuration. By evaluating the importance of each sensor's data streams in advance using machine learning models, the system pre-determines the optimal sensor subset that will capture salient activities without generating excessive irrelevant data, thus resolving the contradiction between deployment simplicity and data relevance.
2Reliability
If a large number of sensors are deployed to capture comprehensive sensor data, then data coverage is improved, but computational and maintenance expenses increase
Solution Approach 1:
The system extracts and identifies the minimal subset of sensors that are most relevant to capturing salient activities of interest. By applying machine learning models to evaluate sensor importance and removing redundant sensors from the deployment configuration, the system maintains comprehensive data coverage for critical activities while reducing the total number of sensors, thereby lowering computational and maintenance expenses.
3Measurement precision
If redundant sensor data streams are collected to ensure complete activity capture, then measurement completeness is improved, but predictive analysis accuracy decreases
Solution Approach 1:
The system applies different evaluation criteria to different sensors based on their specific roles and locations. By analyzing the unique contribution of each sensor to activity capture and assigning relevance weights accordingly, the system ensures that each sensor in the selected subset provides high-quality, non-redundant information. This local optimization approach maintains complete activity capture while improving the overall signal-to-noise ratio by eliminating redundant data streams.
Data Source
AI summary
A controller includes a memory that stores instructions and a processor that executes the instructions. The instructions cause the controller to execute a process that includes receiving sensor data from a first sensor and a second sensor. The sensor data includes a time-series observation representing a first activity and a second activity. The controller generates models for each activity involving progressions through states indicated by the sensor data from each sensor. The controller receives from each sensor additional sensor data including a time-series observation representing the first activity and the second activity. The controller determines likelihoods that the models generated a portion of the additional sensor data and calculates a pair-wise distance between each sensor-specific determined likelihood to obtain calculated distances. The calculated distances for each sensor are grouped and a relevance of each sensor to each activity is determined by executing a regression model using the grouped calculated distances.