Adaptive Live Trip Prediction Using Confidence Rejection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional trip prediction systems rely on historical data and struggle to accurately predict a user's destination, especially when the user deviates from expected behavior, limiting their ability to provide personalized and contextual services like navigation and traffic updates.
Innovation Solution
A system that combines offline learned user behavior with real-time live trip data, using a confidence measure and rejection strategy to predict destinations, even if they are completely new, by determining the quality of predicted destinations and estimating driving directions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional systems use only historical data for trip prediction, then the system complexity is low, but the prediction accuracy deteriorates when users deviate from expected behavior
Solution Approach 1:
The patent combines offline learned user behavior models with real-time live trip data processing to create a hybrid prediction system. This merging allows the system to maintain low complexity for routine predictions while handling complex real-time scenarios when users deviate from expected behavior, thereby improving prediction accuracy without proportionally increasing system complexity
Solution Approach 2:
The system dynamically adjusts its prediction approach by switching between relying on historical patterns and processing live trip data based on the current situation. When user behavior aligns with historical patterns, the simpler historical model is used; when deviations occur, the system activates real-time processing, creating a dynamic complexity adaptation that improves accuracy while managing system resources efficiently
2Measurement precision
If the system processes real-time live trip data continuously, then the prediction accuracy improves, but the computational cost and processing time increase
Solution Approach 1:
The system applies partial real-time processing by selectively activating live trip data analysis only when necessary - specifically when user behavior deviates from historical patterns or when confidence in historical predictions is low. This partial action approach maintains high prediction accuracy for critical cases while avoiding the excessive computational cost of continuous real-time processing for all scenarios
Solution Approach 2:
The system uses feedback from comparing live trip data with historical patterns to determine when real-time processing is needed. By monitoring deviations between expected and actual behavior, the system activates computational resources only when the feedback indicates historical models are insufficient, thereby reducing overall computational cost while maintaining accuracy when it matters most
3Reliability
If the system rejects predictions with low confidence, then the prediction reliability improves, but the ability to provide service for new destinations deteriorates
Solution Approach 1:
The patent introduces an intermediary mechanism that bridges rejected low-confidence predictions and service provision. When predictions are rejected due to low confidence, the system doesn't simply fail but instead uses the partial information and directional data from the rejected prediction to still provide useful services like POI recommendations and route guidance, thereby maintaining service coverage even when full prediction reliability cannot be achieved
Solution Approach 2:
The system prepares cushioning measures in advance by having fallback service capabilities ready for when predictions are rejected. Instead of providing no service for new destinations, the system has pre-prepared alternative service modes that can operate with reduced information, ensuring that service coverage is maintained even when prediction reliability is insufficient for full functionality
Data Source
AI summary
A system, method, and non-transitory computer-readable medium are provided for implementing adaptive live trip prediction in a vehicle of a user in real-time while the user uses the vehicle. In addition to predicting trip destinations based on historical trip data, the system uses a confidence measure of each of a plurality of known candidate destinations and a rejection strategy for rejecting candidate destinations as possible destinations based on real-time vehicle location information. When the system rejects all of the known candidate destinations and determines that the user's destination is an unknown or new destination, the system determines the location of the new destination be repeatedly determining an angle between a line from the original location to the current location and a north direction over time until the new destination can be accurately predicted. Further, the system is able to self-optimize the prediction performance without an explicit interaction with the user.