Vehicular Federated Learning for ML Control Model Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Vehicle controllers face limitations in processing power and memory, making it difficult for them to individually handle the establishment, training, and updating of machine learning (ML) models, leading to inaccurate outputs for controlling vehicle functions due to insufficient sensor data accumulation.
Innovation Solution
Federated learning is employed, allowing for the aggregation and distribution of model parameters across multiple vehicles and servers, leveraging their collective processing power and memory to update and train ML models, either centrally or decentralizedly, to improve model accuracy and adaptability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If vehicle controllers individually train and update ML models using only their own sensor data, then each controller maintains independence and security, but the ML models become inaccurate due to insufficient data accumulation
Solution Approach 1:
The patent merges data resources from multiple vehicles by having each vehicle's controller contribute sensor data to a federated learning system. The controllers collectively train a shared ML model while maintaining data locality, effectively combining the data quantity of multiple vehicles to achieve accurate model training that would be impossible for any single vehicle controller alone.
2Adaptability or versatility
If vehicle controllers individually handle ML model establishment, training, and updating, then each controller operates independently, but processing power and memory limitations prevent effective model training
Solution Approach 1:
The patent combines the processing power and memory resources of multiple vehicle controllers through federated learning. By distributing the ML model training task across multiple controllers while keeping the model parameters shared, the system achieves training capabilities that exceed the individual capacity of any single controller, enabling effective model establishment and updating despite hardware limitations.
3Measurement precision
If federated learning aggregates model parameters across multiple vehicles, then ML model accuracy improves through collective data, but communication overhead and coordination complexity increase
Solution Approach 1:
The patent introduces a federated learning coordinator as an intermediary that manages the aggregation and distribution of ML model parameters across multiple vehicle controllers. This coordinator handles the complexity of synchronizing model updates, aggregating gradients or parameters from multiple vehicles, and distributing updated models back to controllers, thereby reducing the coordination burden on individual vehicle systems while enabling accurate collective learning.
Data Source
AI summary
Presented herein are systems and methods for performing federated learning across vehicles. A computing device having one or more processors coupled with memory can maintain, on the memory, a first machine learning (ML) comprising a first plurality of parameters for determining values identifying a characteristic of a vehicle function on at least one of a plurality of vehicles. The computing device can receive a second plurality of parameters generated by a second ML used by each respective vehicle. The computing device can update the first plurality of parameters of the first ML in accordance with the second plurality of parameters received from each respective vehicle of the plurality of vehicles. The computing device can transmit, to a vehicle, the updated first plurality of parameters to update the second plurality of parameters of the second ML on the vehicle.