Federated Model Consolidation Policy for Diverse Vehicle Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing federated learning schemes for autonomous vehicles face challenges in optimally combining local model updates from diverse operating conditions, leading to suboptimal performance of the consolidated machine learning model.
Innovation Solution
A computer-implemented method for determining a consolidation policy through an optimization approach, which involves obtaining model updates from vehicles, consolidating them according to a candidate policy, evaluating the consolidated model, and updating the policy based on the evaluation to achieve optimal performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If local model updates from diverse operating conditions are combined using traditional federated learning schemes, then the consolidated model can be formed, but the performance is suboptimal due to inability to optimally balance different learning procedures
Solution Approach 1:
The consolidation policy is made dynamic by iteratively adjusting weights of different local model updates based on their evaluated performance. The system transitions from static averaging to dynamic weight adjustment where weights are modified in subsequent consolidation rounds based on how well each local model performs on the evaluation dataset, allowing the system to adaptively optimize model performance
Solution Approach 2:
An evaluation mechanism is introduced that assesses the performance of consolidated models on a held-out dataset and feeds this information back to adjust the consolidation policy. The weights assigned to different local model updates are modified based on feedback from previous evaluation rounds, creating a closed-loop system that continuously improves model performance
2Adaptability or versatility
If model updates from all vehicles are equally combined, then the consolidation process is simple, but the diverse operating conditions and scenarios experienced by different vehicles are not properly balanced
Solution Approach 1:
Different weights are assigned to different local model updates based on their specific performance characteristics and the diversity of operating conditions they represent. Instead of uniform treatment, each local model's contribution is tailored to its quality and relevance, with higher weights given to models that perform better or represent underrepresented scenarios
Solution Approach 2:
The consolidation process changes parameters (weights) of local model updates based on their performance evaluation. The weight parameters are adjusted iteratively to optimize the balance between different operating conditions, transforming the consolidation from a fixed equal-weight approach to a flexible parameter-adjusted approach
Data Source
AI summary
The present inventive concept relates to a computer-implemented method for determining a consolidation policy for forming a consolidated machine learning model from a number of local machine learning models of a fleet of vehicles equipped with an automated driving system. The method includes: obtaining two or more model updates from one or more vehicles of the fleet of vehicles, wherein each model update is a result of training a local machine learning model of the respective vehicle; consolidating the two or more model updates according to a candidate consolidation policy, thereby forming a consolidated machine learning model; evaluating the consolidated machine learning model according to an evaluation criterion; and updating the candidate consolidation policy in view of the evaluation, thereby forming an updated candidate consolidation policy. It further relates to a device thereof.