Deep Neural Path Navigation for Vehicle Pose and Lane Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous vehicle navigation systems are often inaccurate and inefficient, lacking awareness of their environment, which hinders precise control and safety.
Innovation Solution
The implementation of deep neural networks (DNNs) for analyzing image data to determine a vehicle's orientation and lateral position relative to a path, enabling real-time control and obstacle detection, thereby improving navigation accuracy and safety.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If current autonomous control implementations are used, then vehicle navigation can be performed, but accuracy and efficiency are insufficient and environmental awareness is lacking
Solution Approach 1:
The system segments the navigation task into multiple independent neural network modules: a path detection network that identifies the path from image data, a pose estimation network that determines vehicle orientation and lateral position, and a control network that generates steering commands. Each module specializes in a specific function, improving overall accuracy and reliability through divided expertise.
Solution Approach 2:
The patent introduces an intermediary coordinate system transformation layer that converts image coordinates to vehicle-centric coordinates. This intermediary representation enables the system to accurately interpret visual data relative to the vehicle's pose, bridging the gap between raw image data and navigation decisions, thereby improving both measurement precision and environmental awareness.
2Measurement precision
If deep neural networks are implemented for real-time image analysis, then navigation accuracy and environmental awareness improve, but computational complexity increases
Solution Approach 1:
The complex deep learning task is segmented into three separate neural network models, each trained for a specific function (path detection, pose estimation, control). This segmentation allows each network to be more specialized and efficient, reducing the computational burden compared to a single monolithic network while maintaining high accuracy.
Solution Approach 2:
The system performs preliminary action by pre-training the neural network models offline with large datasets. During real-time operation, the pre-trained networks perform inference with minimal computational overhead. The pose estimation and path detection are performed in advance of the control decision, enabling real-time response despite the complexity of the underlying algorithms.
Data Source
AI summary
A method, computer readable medium, and system are disclosed for performing autonomous path navigation using deep neural networks. The method includes the steps of receiving image data at a deep neural network (DNN), determining, by the DNN, both an orientation of a vehicle with respect to a path and a lateral position of the vehicle with respect to the path, utilizing the image data, and controlling a location of the vehicle, utilizing the orientation of the vehicle with respect to the path and the lateral position of the vehicle with respect to the path.