Vehicular Federated Learning for Accurate Function Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle control systems face challenges in adapting and updating machine learning models due to processing power and memory constraints, as well as insufficient sensor data, leading to inaccurate outputs for controlling vehicle components.
Innovation Solution
The implementation of federated learning across vehicles, where a central server or decentralized vehicle controllers maintain and update machine learning models by aggregating and distributing model parameters, enabling efficient training and adaptation of local models without relying on a single controller or server.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single controller or server is used to maintain machine learning models in a vehicle, then the system structure is simple, but the model training accuracy is insufficient due to limited processing power, memory constraints, and insufficient sensor data
Solution Approach 1:
The patent divides the centralized model training system into distributed local models across multiple vehicle controllers. Each controller maintains and trains its own local ML model using local sensor data, segmenting the training workload and data requirements across the vehicle network rather than concentrating it in a single controller or server.
Solution Approach 2:
The patent combines the computing resources and sensor data from multiple vehicle controllers to collectively train and update ML models. By merging the capabilities of distributed controllers through a controller area network, the system achieves model training accuracy comparable to centralized systems while distributing the computational burden across multiple nodes.
2Measurement precision
If more sensor data is collected to improve model training, then the model accuracy improves, but the processing load and memory requirements on individual controllers increase
Solution Approach 1:
The patent segments the data collection and processing tasks across multiple controllers in the vehicle network. Each controller collects and processes only the sensor data relevant to its local model, rather than one controller collecting all sensor data. This distribution reduces the processing load and memory requirements on any single controller while maintaining sufficient data for accurate model training.
Data Source
Figure 1
Figure 2
Figure 3
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.