Autonomous Vehicle Deep-Learning Updates via Joint Kernel Sharing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in efficiently processing large data volumes for navigation and routing, with limited storage capacity and high storage costs, necessitating an efficient continuous learning process across a fleet.
Innovation Solution
Implementing a continuous learning model on-board each vehicle, averaging deep-learning model updates into a joint kernel using knowledge distillation, and performing safety self-tests to ensure safe operation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If each autonomous vehicle processes and stores large quantities of data locally for navigation and routing tasks, then the vehicle can perform complex driving functions, but the storage capacity requirement increases and storage costs become prohibitively expensive
Solution Approach 1:
The patent merges the deep learning model across multiple autonomous vehicles by creating a shared joint kernel that aggregates learnings from all vehicles. Instead of each vehicle maintaining its own complete model, the system combines models from multiple vehicles into a unified shared resource, reducing individual storage requirements while maintaining collective intelligence.
Solution Approach 2:
The shared joint kernel serves multiple vehicles simultaneously, making the storage system universal. A single stored model can be deployed across the entire fleet, allowing one storage instance to serve multiple functions (multiple vehicles) rather than requiring separate storage for each vehicle.
2Ease of operation
If each autonomous vehicle maintains its own deep learning model, then the vehicle can operate independently, but the overall storage costs across the fleet become prohibitively expensive
Solution Approach 1:
The patent combines individual vehicle models into a shared joint kernel through knowledge distillation and averaging processes. This merging reduces the total storage required across the fleet while vehicles continue to operate independently using the shared resource, eliminating redundant storage costs.
Solution Approach 2:
Instead of each vehicle storing its own complete model, the system creates a shared master copy (joint kernel) that all vehicles reference. This single copy is distributed to all vehicles, eliminating the need for multiple duplicate copies and significantly reducing total storage costs.
3Measurement precision
If the system collects and processes large quantities of data from all vehicles, then the learning accuracy improves, but the data processing complexity and storage requirements increase
Solution Approach 1:
The patent extracts only the essential learnings from each vehicle's data and models, rather than processing and storing all raw data. By extracting and aggregating only the meaningful updates and knowledge representations, the system maintains high learning accuracy while significantly reducing data processing complexity and storage requirements.
Solution Approach 2:
The system extracts knowledge representations from individual vehicle models and consolidates them into a shared joint kernel. This extraction process filters out redundant information and retains only the essential learning patterns, improving accuracy while reducing the volume of data that needs to be processed and stored.
Data Source
AI summary
The subject disclosure relates to techniques for enabling sharing of knowledge among a fleet of autonomous vehicles. A process of the disclosed technology can include generating an update for a continuous deep learning neural network on-board the autonomous vehicle based on driving scenarios encountered by the autonomous vehicle during its deployment and providing the update for the continuous deep learning neural network to additional vehicles in the fleet on autonomous vehicles, wherein the update for the continuous deep learning neural network is configured to be incorporated into a joint kernel for use by the additional vehicles in the fleet on autonomous vehicles.