Floor Selection via Decision Tree Stump Weighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional building automation systems face challenges in accurately determining the floor location of targets within multi-floor structures due to the complexity of radio frequency signals traversing between floors, making it difficult to update and retrain machine learning-based systems for efficient floor selection.
Innovation Solution
A floor selection system that combines radio frequency (RF), motion detection, and floor plan data using decision tree stumps and machine learning classifiers, allowing for adaptive updates and efficient computation by using only decision tree stumps as weak classifiers, enabling accurate floor determination through a weighted aggregation of features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If machine learning-based systems are used for floor selection, then adaptability is improved, but computational complexity and retraining requirements increase
Solution Approach 1:
The patent segments the machine learning model into multiple decision tree stumps arranged in sequence. Each stump makes a simple binary decision based on a single feature threshold, breaking down the complex classification task into manageable stages. This segmentation allows the system to maintain adaptability while reducing the computational burden of each individual decision step.
Solution Approach 2:
The patent implements dynamic feature selection where different features are evaluated at each decision tree stump based on the previous decisions. The system adapts the feature set being considered based on the current state, allowing the classifier to focus computational resources on the most relevant features for the current classification task, thereby reducing overall computational complexity while maintaining adaptability.
2Measurement precision
If multiple layers of sensor data are fused together, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent segments the sensor data fusion process into distinct layers: RF signal strength measurements, motion detection data, and floor plan information. Each layer processes specific types of data independently before combining results. This segmentation allows the system to integrate multiple data sources for improved accuracy while managing complexity through modular processing stages.
Solution Approach 2:
The patent introduces decision tree stumps as intermediary elements that mediate between raw sensor data and final floor location determination. Each stump acts as an intermediate processing step that evaluates specific features from the sensor data and produces simplified outputs that are easier to combine and interpret, thereby facilitating data fusion while controlling system complexity.
Data Source
AI summary
There is described a floor selection system for location tracking within a structure. A communication component receives sensor data, associated with a beacon received from a tag, from sensors. A processor identifies features based on the sensor data, identifies decision tree stumps arranged in a particular order, assigns weights to the decision tree stumps based on the features and the feature thresholds. The processor further determines whether an aggregate of the weights is greater than a resultant threshold and selects a floor location of the tag based on whether the aggregate of weights is greater than the resultant threshold. The floor location is selected from floor locations of the structure proximal to the sensors.