Recurrent Neural Network for Autonomous Driving Mode Switching

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional autonomous vehicles require multiple sensors and significant computing resources to switch between driving modes, increasing manufacturing costs and resource consumption.

Innovation Solution

A recurrent neural network is used to integrate multiple camera information from front and rear cameras to determine hazardous situations, allowing safe switching between autonomous and manual driving modes without the need for extensive sensor arrays or high computational resources.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If multiple sensors are installed to acquire driving information for mode switching, then the reliability of autonomous driving safety check is improved, but the manufacturing cost increases

Engineering Contradiction:
Improveautonomous driving safety checkVSAvoidmanufacturing cost
Core Design Contradiction:
ReliabilityVSEase of manufacture

Solution Approach 1:

The patent merges multiple camera information sources (front camera, rear camera, and potentially other cameras) into a unified processing system that uses a single recurrent neural network to perform hazard recognition and safety checks. This consolidation achieves reliable autonomous driving monitoring while avoiding the need for multiple separate sensor systems, thereby reducing manufacturing costs.

Inventive Principle:
Principle #5Merging (Combining)

2Measurement precision

If various sensors are used to acquire driving information, then the accuracy of hazardous situation recognition is improved, but the device complexity increases

Engineering Contradiction:
Improvehazardous situation recognitionVSAvoidsensor array
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The recurrent neural network is designed as a universal processing system that handles multiple camera inputs and performs multiple functions including hazard recognition, safety assessment, and mode switching recommendations. This multi-functional approach achieves accurate hazardous situation detection while avoiding the complexity of dedicated specialized sensors for each function.

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Reliability

If a large amount of sensor information is processed to change driving mode, then the reliability of mode switching decision is improved, but the computing resources required increase

Engineering Contradiction:
Improvemode switching decisionVSAvoidcomputing resources
Core Design Contradiction:
ReliabilityVSUse of energy by moving object

Solution Approach 1:

The system extracts only the essential features and hazard-related information from camera inputs using the recurrent neural network, rather than processing all raw sensor data. This extraction approach maintains reliable mode switching decisions by focusing computation on critical safety parameters while significantly reducing overall computing resource requirements.

Inventive Principle:
Principle #2Taking out (Extraction)

4Loss of information

If multiple sensors are installed for comprehensive driving information, then the completeness of driving information is improved, but the quantity of sensors increases

Engineering Contradiction:
Improvedriving information completenessVSAvoidnumber of sensors
Core Design Contradiction:
Loss of informationVSQuantity of substance

Solution Approach 1:

The patent combines information from multiple camera sources into a unified processing framework, achieving comprehensive driving information coverage through software integration rather than hardware proliferation. This approach maintains information completeness while minimizing the physical number of sensor components required.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentEP3690709B1Learning method and learning device of recurrent neural network for autonomous driving safety check for changing driving mode between autonomous driving mode and manual driving mode, and testing method and testing device using them
Publication Date: 2023.11.01 STRADVISION
  • EP3690709B1 patent drawingFigure 1
  • EP3690709B1 patent drawingFigure 2
  • EP3690709B1 patent drawingFigure 3

AI summary

A method for learning a recurrent neural network to check an autonomous driving safety to be used for switching a driving mode of an autonomous vehicle is provided. The method includes steps of: a learning device (a) if training images corresponding to a front and a rear cameras of the autonomous vehicle are acquired, inputting each pair of the training images into corresponding CNNs, to concatenate the training images and generate feature maps for training, (b) inputting the feature maps for training into long short-term memory models corresponding to sequences of a forward RNN, and into those corresponding to the sequences of a backward RNN, to generate updated feature maps for training and inputting feature vectors for training into an attention layer, to generate an autonomous-driving mode value for training, and (c) allowing a loss layer to calculate losses and to learn the long short-term memory models.