EV Range Estimation With Location-Based Driver Model Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional range estimation systems for electric and hybrid electric vehicles fail to accurately predict energy consumption due to slow updates in driver model adaptation when changing locations or times, requiring large databases for classification, compromising privacy, and lacking complexity in driver behavior classification.
Innovation Solution
A hierarchical learning framework in a cloud server trains location and time-based models that update onboard energy consumption estimation models in vehicles, personalized to a driver's style and common locations, ensuring accurate predictions without compromising privacy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional range estimation systems use large databases for driver model classification and adaptation, then prediction accuracy may improve, but system complexity and data storage requirements increase significantly
Solution Approach 1:
The patent segments the driver model adaptation process into distinct phases: an initial classification phase that uses a large database to identify driver behavior patterns, and a subsequent personalized phase that uses a compact learned model stored in memory. This segmentation allows the system to benefit from comprehensive database analysis while avoiding the ongoing complexity and storage requirements of maintaining large databases in the vehicle system.
Solution Approach 2:
The patent performs preliminary action by pre-processing driver behavior data and training personalized energy consumption models offline using comprehensive databases. The results of this preliminary analysis are stored as compact model parameters in the vehicle's memory, enabling accurate predictions without requiring the vehicle system to maintain or process large databases during operation.
2Measurement precision
If conventional systems use comprehensive databases for driver behavior classification, then model adaptation improves, but driver privacy is compromised
Solution Approach 1:
The patent introduces an intermediary offline training process that acts as a mediator between comprehensive database analysis and the vehicle's onboard system. This intermediary process performs the detailed classification and model training using comprehensive data, then outputs only essential model parameters to the vehicle. This intermediary layer enables accurate model adaptation while preventing direct exposure of driver data to the vehicle system, thereby protecting privacy.
3Stability of the object's composition
If range estimation systems are updated slowly when changing locations or times, then system stability is maintained, but prediction accuracy deteriorates
Solution Approach 1:
The patent implements dynamics by making the energy consumption model adaptive to changing conditions such as location and time. The system dynamically adjusts the model parameters based on the driver's actual energy consumption patterns observed in different contexts. This dynamic adaptation allows the system to maintain stability through a learned model structure while improving prediction accuracy by continuously adapting to new conditions without requiring slow updates or retraining.
Data Source
AI summary
Systems and methods are provided for estimating the energy range of an electric vehicle, including updating an energy consumption estimation model based on the location of a vehicle or based on the current time. One embodiment comprises obtaining a route of a vehicle and deploying an energy consumption estimation model to provide an energy consumption estimate along the route from a first location to a destination. In response to determining that the vehicle enters a subsequent location, the energy consumption estimation model is updated with a location model received from an edge/cloud server, to estimate an energy range along the route through the subsequent location.