Automatic Driving Model Training with On-Vehicle Data Filtering

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

The training of automatic driving models is inefficient due to the large amount of unnecessary data uploaded to cloud servers, which consumes significant bandwidth and resources, and results in low training efficiency.

Innovation Solution

A system and method where a vehicle equipment sorts and transmits only the objective driving data meeting a sample acquisition strategy to a computer device, reducing the amount of data uploaded and processed, thereby optimizing training efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If vehicles upload massive amount of driving data to cloud server, then training data volume is increased, but bandwidth and hardware/software resources are consumed excessively

Engineering Contradiction:

Improvetraining data volumeVSAvoidbandwidth and resource consumption

Core Design Contradiction:

Quantity of substanceVSLoss of energy

Solution Approach 1:

The system performs preliminary action by having the automatic driving system sort and filter driving data locally before upload, identifying and selecting only objective driving data that meets training requirements. This pre-processing step eliminates unnecessary data transmission, reducing bandwidth and resource consumption while ensuring sufficient training data volume is uploaded.

Inventive Principle:

Principle #10Preliminary action

2Quantity of substance

If all collected driving data is uploaded for training, then data completeness is improved, but training efficiency decreases due to processing unnecessary data

Engineering Contradiction:

Improvedata completenessVSAvoidtraining efficiency

Core Design Contradiction:

Quantity of substanceVSProductivity

Solution Approach 1:

The system extracts only the necessary objective driving data from the massive collected driving data using a sorting model. By taking out and uploading only the relevant data that meets training requirements, the system maintains data completeness for effective training while eliminating processing of unnecessary data, thereby improving training efficiency.

Inventive Principle:

Principle #2Taking out (Extraction)

3Quantity of substance

If cloud server receives and processes all uploaded data, then data processing comprehensiveness is improved, but software and hardware resources are wasted on unnecessary data

Engineering Contradiction:

Improvedata processing comprehensivenessVSAvoidsoftware and hardware resource consumption

Core Design Contradiction:

Quantity of substanceVSLoss of energy

Solution Approach 1:

The system performs preliminary sorting and filtering of driving data at the vehicle end before upload. This pre-processing action ensures that only objective driving data requiring cloud server processing is transmitted, maintaining comprehensive processing of necessary data while avoiding waste of software and hardware resources on unnecessary data.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS20250190860A1Method and system for training automatic driving model

Publication Date: 2025.06.12 HON HAI PRECISION INDUSTRY CO LTD

AI summary

A method for training an automatic driving model comprises: obtaining a sample acquisition strategy of the automatic driving model; transmitting the sample acquisition strategy of the automatic driving model to a vehicle equipment, the vehicle equipment being configured to sort objective driving data from a driving data set based on the sample acquisition strategy; receiving the objective driving data from the vehicle equipment; adding the objective driving data into a training set of the automatic driving model; and training the automatic driving model based on the training set. A system for training the automatic driving model is also disclosed.