Vehicle Travel Control Using Future Image Prediction Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing travel controllers face challenges in determining the cause of inappropriate vehicle control based on predicted future surrounding conditions, making it difficult to take appropriate action.
Innovation Solution
A travel controller using two neural networks to generate a future image and control signal, where a first neural network predicts surrounding conditions and a second neural network generates the control signal based on current and future images, allowing for appropriate vehicle control by distinguishing between prediction and signal generation issues.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single neural network is used to generate control signals based on predicted future conditions, then the control system is simpler, but it becomes difficult to determine whether inappropriate control is caused by prediction errors or signal generation errors
Solution Approach 1:
The patent divides the single neural network into two separate neural networks: a first neural network dedicated to predicting future surrounding conditions and generating future images, and a second neural network dedicated to generating control signals based on current and future images. This segmentation allows independent optimization of each function and enables clear identification of whether control issues stem from prediction errors (first network) or signal generation errors (second network), thus resolving the diagnostic information loss while maintaining reasonable system complexity.
2Adaptability or versatility
If future surrounding conditions are predicted based on vicinity images, then autonomous driving capability is enhanced, but the ability to diagnose control errors is reduced
Solution Approach 1:
The patent introduces future images as an intermediary output from the first neural network that serves dual purposes: (1) providing predicted future surrounding conditions to the second neural network for control signal generation, and (2) serving as a diagnostic tool to verify prediction accuracy. By displaying or analyzing these future images, the system can determine whether prediction errors are causing inappropriate control, thus maintaining autonomous driving capability while enabling error detection.
3Measurement precision
If the predetermined period for future prediction is extended, then more accurate future conditions can be predicted, but the control response time increases
Solution Approach 1:
The patent makes the predetermined period dynamic rather than fixed. The first neural network can adjust the prediction time horizon based on current driving conditions, such as vehicle speed, distance to obstacles, and environmental complexity. This allows the system to extend the prediction period for accurate forecasting when conditions permit, while shortening it when rapid response is needed, thus balancing prediction accuracy with control response time.
Data Source
AI summary
A travel controller generates a future image by inputting a series of vicinity images representing surrounding conditions of a vehicle up to a current time into a first neural network. The future image represents predicted surrounding conditions of the vehicle at a future time that is a predetermined period after the current time. The travel controller generates a control signal for controlling travel of the vehicle by inputting a vicinity image outputted at the current time of the series of vicinity images, the future image, and the predetermined period into a second neural network different from the first neural network.