Vehicle Future Path Estimation Using Road-Scene Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing advanced driver assistance systems (ADAS) and autonomous vehicle (AV) systems face limitations in handling the infinitesimal variety and dynamic nature of road environments, such as moving vehicles and shadows, which affect their ability to accurately detect and respond to objects.
Innovation Solution
A system utilizing multiple cameras and processors to process images, employing neural networks and deep learning algorithms to estimate a vehicle's future path, incorporating piece-wise affine functions, convolutions, and max pooling, and rectifier linear units (ReLU) for enhanced object detection and path prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If preconfigured object classifiers are used to detect objects in road environments, then the system can identify predefined objects, but the system fails to handle the infinitesimal variety and dynamic nature of road environments effectively
Solution Approach 1:
The patent transitions from static preconfigured classifiers to dynamic deep learning models that adapt to varying road conditions. The system uses neural networks that can dynamically adjust to different environments, handling moving vehicles, shadows, and other dynamic elements through continuous learning and processing of image data.
Solution Approach 2:
The system changes the parameters of object detection by moving from fixed classifier configurations to adaptive neural network parameters. The deep learning model adjusts its internal parameters through training on diverse road environment data, enabling it to handle infinitesimal variety and dynamic conditions effectively.
2Measurement precision
If traditional object detection methods are used, then the system can detect predefined objects, but the system lacks the ability to accurately estimate future paths in dynamic environments
Solution Approach 1:
The patent adds a temporal dimension to path estimation by using recurrent neural networks (RNNs) and long short-term memory (LSTM) networks. These models process sequences of image data over time, enabling accurate prediction of future vehicle paths by analyzing historical motion patterns and environmental dynamics.
Solution Approach 2:
The system segments the path estimation task into multiple components: object detection, trajectory prediction, and future path estimation. Each component is handled by specialized neural network modules, making the overall complex system manageable and accurate through functional decomposition.
3Productivity
If deep learning systems are implemented to improve path estimation, then the ability to predict future paths improves, but the computational requirements and processing time increase
Solution Approach 1:
The system performs preliminary action by pre-training deep learning models offline on extensive road environment datasets. This pre-training enables the models to make accurate predictions during real-time operation with reduced computational burden, as the heavy learning phase is completed beforehand.
Solution Approach 2:
The patent replaces traditional mechanical and rule-based path prediction systems with intelligent software-based deep learning models. This substitution enables more accurate and adaptive path estimation while optimizing computational efficiency through algorithmic improvements and hardware acceleration.
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
A system and method estimate a future path ahead of a current location of a vehicle. The system includes at least one processor programmed to: obtain an image of an environment ahead of a current arbitrary location of a vehicle navigating a road; obtain a trained system that was trained to estimate a future path on a first plurality of images of environments ahead of vehicles navigating roads; apply the trained system to the image of the environment ahead of the current arbitrary location of the vehicle; and provide, based on the application of the trained system to the image, an estimated future path of the vehicle ahead of the current arbitrary location.