Radar Road-Edge Prediction Using Clustering and Neural Polynomials
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current driver assistance systems and autonomous vehicles face challenges in accurately predicting road courses using radar data, especially in complex scenarios like urban areas, due to reliance on manually created features and heuristics, which can lead to errors and inflexibility.
Innovation Solution
A method that processes radar data by forming clusters, transforming them into a feature space using a receptive field, and using a neural network to determine polynomials for describing road edges, allowing for accurate and flexible road course prediction without manual treatment of special cases.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If manually created features and heuristics are used for road detection, then the system can operate with simpler processing, but the accuracy and reliability deteriorate in complex scenarios like urban areas
Solution Approach 1:
The patent replaces manual feature engineering and heuristic methods with a neural network-based machine learning system. The neural network automatically learns road edge features from radar data, substituting the mechanical approach of manual feature creation with an intelligent system that adapts to complex scenarios including urban environments, thereby improving reliability without proportionally increasing processing complexity.
2Ease of manufacture
If predefined models and heuristics are used for road geometry detection, then the system is simpler to implement, but the adaptability to different road scenarios deteriorates
Solution Approach 1:
The patent implements a dynamic system where the neural network can adapt its feature extraction capabilities to different road scenarios. Instead of static predefined models, the system dynamically learns and adjusts to various road geometries, urban environments, and driving conditions, providing both ease of implementation through a unified model and high adaptability to diverse scenarios.
Solution Approach 2:
The patent changes the fundamental parameter of road detection from fixed heuristic rules to learnable neural network parameters. The neural network's weights and features are trained on diverse data and can adapt to different scenarios by changing its internal parameters, enabling the system to handle various road types, urban environments, and edge cases without requiring separate predefined models for each scenario.
3Measurement precision
If cluster-based processing of radar data is implemented, then the noise and complexity are reduced, but the computational processing time increases
Solution Approach 1:
The patent applies segmentation by dividing the radar measuring grid into clusters of radar signals that belong to the same object or structure. This clustering approach reduces noise and complexity by grouping related signals together, making the subsequent neural network processing more efficient. The segmented clustered data is then processed by the neural network to detect road edges, achieving high precision while managing computational load through the preliminary organization of data into meaningful clusters.
Data Source
AI summary
A method, a computer program with instructions, and a device for predicting a course of a road based on radar data of a motor vehicle. The radar data to be processed is received and then accumulated in a measuring grid. Subsequently, clusters are formed for objects in the measuring grid. Cluster descriptions are generated for the clusters. The resulting clusters are processed to determine polynomials for describing the road edges. The polynomials are finally output for further use.