ANN-Guided Data Placement to Cut Write Amplification in AV Storage

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Autonomous vehicles face challenges in predicting component failures and scheduling maintenance, as existing systems lack efficient methods to analyze sensor data for proactive maintenance, leading to potential breakdowns during operation.

Innovation Solution

A data storage device equipped with an artificial neural network (ANN) that processes sensor data from various sources, including cameras and sensors, to predict maintenance needs by training on specific vehicle environments and habits, allowing for timely scheduling of maintenance services.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If sensor data is continuously collected and stored for autonomous vehicle operation, then the availability of data for maintenance prediction is improved, but the storage device endurance deteriorates due to frequent write operations

Engineering Contradiction:

Improvemaintenance prediction accuracyVSAvoidstorage device endurance

Core Design Contradiction:

ReliabilityVSDuration of action of stationary object

Solution Approach 1:

The system performs preliminary filtering and processing of sensor data before storage, identifying only maintenance-relevant data patterns. This preliminary action reduces the volume of data requiring frequent writes to non-volatile memory, thereby preserving storage device endurance while maintaining sufficient data availability for maintenance prediction tasks

Inventive Principle:

Principle #10Preliminary action

2Speed

If neural network models are trained offline and loaded into storage, then the processing speed during operation is improved, but the model update capability deteriorates

Engineering Contradiction:

Improveprediction processing speedVSAvoidmodel update capability

Core Design Contradiction:

SpeedVSAdaptability or versatility

Solution Approach 1:

The system implements a dynamic model architecture where neural network models exist in multiple states: a lightweight version loaded in memory for fast real-time prediction, and a comprehensive version stored in non-volatile memory for periodic updates. This dynamic arrangement allows the system to switch between speed-optimized and adaptability-optimized modes, resolving the contradiction between processing speed and model update capability

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system performs model updates periodically rather than continuously, loading updated models from storage into memory at scheduled intervals when the vehicle is stationary or during maintenance windows. This periodic action allows the system to maintain fast processing speed during operation while still incorporating updated models periodically, balancing both requirements

Inventive Principle:

Principle #19Periodic action

3Stability of the object's composition

If all sensor data is stored in non-volatile memory, then the data retention capability is improved, but the write amplification increases reducing device lifespan

Engineering Contradiction:

Improvedata retention capabilityVSAvoiddevice lifespan

Core Design Contradiction:

Stability of the object's compositionVSLoss of substance

Solution Approach 1:

The system applies different storage strategies to different types of sensor data based on their specific retention requirements. Critical maintenance-related data is persisted in non-volatile memory with high retention guarantees, while transient sensor readings are stored temporarily in volatile memory or processed immediately. This local quality differentiation reduces overall write amplification to non-volatile memory while maintaining data retention capability for essential information

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The system creates selective copies of sensor data, storing only maintenance-relevant features and patterns in non-volatile memory rather than duplicating all raw sensor data. This copying strategy maintains data retention capability for predictive maintenance while significantly reducing the write volume to non-volatile memory, thereby extending device lifespan

Inventive Principle:

Principle #26Copying

Data Source

PatentUS11650746B2Intelligent write-amplification reduction for data storage devices configured on autonomous vehicles

Publication Date: 2023.05.16 MICRON TECHNOLOGY INC

AI summary

Systems, methods and apparatus of intelligent write-amplification reduction for data storage devices configured on autonomous vehicles. For example, a data storage device of a vehicle includes: one or more storage media components; a controller configured to store data into and retrieve data from the one or more storage media components according to commands received in the data storage device; an address map configured to map between: logical addresses specified in the commands received in the data storage device, and physical addresses of memory cells in the one or more storage media components; and an artificial neural network configured to receive, as input and as a function of time, operating parameters indicative a data access pattern, and generate, based on the input, a prediction to determine an optimized data placement scheme. The controller is configured to adjust the address map according to the optimized data placement scheme.