Dual Neural-Network Perception Switching for Battery-Constrained Vehicles

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing perception systems in autonomous vehicles are power-intensive, leading to significant battery drain in battery-operated vehicles, which can reduce vehicle range and affect safety-critical functions.

Innovation Solution

Implement a dual perception system architecture with a high-power and low-power mode, optimizing machine learning models for power consumption using a cost function that balances accuracy and power usage, allowing dynamic switching between modes based on battery levels.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a high-power perception system is used to maintain high accuracy in autonomous driving, then the accuracy of environmental feature identification is improved, but the power consumption increases significantly

Engineering Contradiction:
Improveaccuracy of environmental feature identificationVSAvoidpower consumption of perception system
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The perception system is divided into multiple neural networks with different computational complexities and power consumption levels. A first neural network operates at high power for maximum accuracy, while second and third neural networks operate at lower power levels. The system segments the processing workload based on operational context, allowing the vehicle to switch between different perception modes depending on battery status and driving conditions.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system dynamically selects which neural network to execute based on real-time battery status and driving conditions. When battery charge is sufficient, the high-accuracy first neural network is used. When battery charge drops below thresholds, the system transitions to lower-power second or third neural networks. This dynamic adaptation allows the system to optimize the balance between accuracy and power consumption in real-time.

Inventive Principle:
Principle #15Dynamics

2Reliability

If the perception system operates continuously at high power, then the reliability of autonomous driving is improved, but the battery duration is reduced

Engineering Contradiction:
Improvereliability of autonomous drivingVSAvoidbattery duration
Core Design Contradiction:
ReliabilityVSDuration of action of moving object

Solution Approach 1:

The system changes operational parameters (power consumption level, neural network complexity, processing frequency) based on battery charge status. When battery charge is high, the system operates with high-power, high-reliability settings. When battery charge drops, the system adjusts parameters to reduce power consumption while maintaining acceptable reliability levels, extending battery duration without completely sacrificing autonomous driving capability.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The system continuously monitors battery status and uses this feedback to adjust perception system operation. The battery management system provides real-time charge level information, which triggers automatic switching between different neural network configurations. This closed-loop feedback ensures that the system maintains reliability when power is available while extending battery duration when charge is low.

Inventive Principle:
Principle #23Feedback

3Duration of action of moving object

If a low-power perception system is used to extend battery duration, then the battery charge preservation is improved, but the accuracy of environmental feature identification deteriorates

Engineering Contradiction:
Improvebattery durationVSAvoidaccuracy of environmental feature identification
Core Design Contradiction:
Duration of action of moving objectVSMeasurement precision

Solution Approach 1:

The lower-power neural networks perform partial perception functions rather than full analysis. The second neural network handles moderate complexity tasks with reduced accuracy, while the third neural network handles only basic detection with minimal power consumption. This partial action approach extends battery duration while maintaining sufficient accuracy for safe autonomous driving in less critical conditions.

Inventive Principle:
Principle #16Partial or excessive action

4Adaptability or versatility

If multiple neural networks with different power consumption levels are implemented, then the adaptability to different power conditions is improved, but the device complexity increases

Engineering Contradiction:
Improveadaptability to different battery conditionsVSAvoidcomplexity of perception system architecture
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

Multiple neural networks are implemented within a single perception system architecture, allowing the system to perform multiple functions (high-accuracy perception, moderate-accuracy perception, basic detection) using different network configurations. This multi-functionality approach enables the system to adapt to various battery conditions without requiring separate dedicated systems for each power level, managing complexity through unified architecture design.

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

Data Source

PatentEP3663974B1Method and apparatus for providing a low-power perception architecture
Publication Date: 2025.09.17 HERE GLOBAL BV
  • EP3663974B1 patent drawingFigure 1
  • EP3663974B1 patent drawingFigure 2
  • EP3663974B1 patent drawingFigure 3

AI summary

An approach is provided for providing a lower-power perception architecture. The approach involves, for example, determining that a device is equipped with a first perception system and a second perception system. The second perception system operates in a lower-power consumption mode than the first perception system to process image data for image recognition. The approach also involves determining a battery level of the device. The approach further involves switching from the first perception system to the second perception system based on determining that the battery level is below a threshold battery level.