Open-Set Trajectory Classification for Unknown Mobile Edge Paths
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning models for mobile edge devices assume a static set of typical trajectories, leading to poor performance in dynamic environments where unknown trajectories can result in incorrect classifications and potentially dangerous actions.
Innovation Solution
Implementing an open-set approach using auto-encoders and class-conditioned autoencoders to classify unknown trajectories, allowing models to adapt to changing conditions with minimal human intervention and improve performance in dynamic environments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a static set of typical trajectories is assumed for classification, then model simplicity is maintained, but model reliability deteriorates in dynamic environments with unknown trajectories
Solution Approach 1:
The patent transforms the static trajectory classification problem into a dynamic one by introducing temporal context. The system maintains a history of recent trajectories and uses them to adaptively update classification behavior, allowing the model to respond to changing environmental conditions while maintaining a structured approach to unknown trajectories through distance-based metrics and context-aware decision-making
2Measurement precision
If traditional classification models are used, then computational resources are conserved, but measurement precision deteriorates when classifying unknown trajectories
Solution Approach 1:
The patent segments the classification problem into distinct components: a feature extraction module that processes trajectory data, a distance computation module that measures similarity to known classes, and a classification module that applies context-aware decision rules. This segmentation allows each component to be optimized independently, improving overall precision without requiring a monolithic complex model
Solution Approach 2:
The patent introduces intermediate representations and context variables that mediate between raw trajectory data and final classification decisions. The system computes distance metrics to known trajectory classes and uses these intermediate values, along with contextual information from recent history, to make informed classification decisions about unknown trajectories, thereby improving accuracy without directly increasing model complexity
3Loss of information
If the model set is kept static, then ease of operation is maintained, but loss of information occurs when unknown trajectory classes are encountered
Solution Approach 1:
The patent implements feedback mechanisms where the classification system continuously monitors its performance and the characteristics of encountered trajectories. When unknown trajectory classes are detected, the system uses distance metrics and contextual information from recent history to provide feedback about the nature of these unknown classes, preventing information loss while maintaining operational simplicity through automated adaptive responses
Data Source
AI summary
One example method includes training an event predictor, of a machine learning model, wherein the training includes training an event predictor of the machine learning model, and training open set trajectory classifier, of the machine learning model. After the event predictor is trained, the event predictor is operable to receive an unknown trajectory class and predict an event class for the unknown trajectory class. Further, after the open set trajectory classifier is trained, the open set trajectory classifier is operable to receive an unknown trajectory and classify the unknown trajectory with a predicted trajectory classification. Finally, during training, the output of the open-set trajectory classifier may be used as an input to the event predictor.