Vehicle Sensor Fusion Using Neural Adaptation for Object Positioning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for fusing environment information from motor vehicle sensors, such as cameras, radar, and lidar, often produce either fuzzy or imprecise data, which may not accurately represent the location of objects, leading to variability in object detection and tracking in varying driving conditions.
Innovation Solution
A fusion system comprising environment sensors, a neural network, and a fusion apparatus that adapts and interprets data using circumstance variables and influencing variables to provide precise environment information to driver assistance systems, combining data from multiple sensors and accounting for sensor quality and behavioral models to enhance object location accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If environment information is fused using traditional methods, then the fusion process is simple, but the object location accuracy becomes fuzzy or imprecise
Solution Approach 1:
The system segments the fusion process into two distinct components: a neural network module that processes sensor data through learned patterns, and a fusion apparatus that combines information using circumstance variables and influencing variables. This segmentation allows each component to specialize, improving object location accuracy while maintaining manageable complexity through modular architecture.
Solution Approach 2:
The fusion apparatus acts as an intermediary between the neural network and the driver assistance system. It receives circumstance variables (light conditions, weather, time) and determining influencing variables that mediate how sensor data is weighted and combined. This intermediary layer refines the fusion process to achieve higher precision without requiring complete redesign of the entire system.
2Reliability
If sensor data is processed without considering circumstance variables, then the processing speed is fast, but the reliability of environment information decreases
Solution Approach 1:
The system performs preliminary action by pre-defining circumstance variables (light conditions, weather conditions, time of day) and their associated influencing variables before actual sensor data fusion occurs. These pre-established relationships allow the fusion apparatus to quickly adjust processing parameters based on current conditions, improving reliability without requiring complex real-time analysis of every variable.
Solution Approach 2:
The fusion apparatus dynamically changes processing parameters based on circumstance variables. When light conditions change from bright to dark, the system adjusts the weighting of different sensors (e.g., increasing radar weight when camera visibility decreases). This parameter adaptation improves environment information reliability across varying conditions while maintaining efficient processing through predefined adjustment rules.
Data Source
AI summary
A fusion system for a motor vehicle includes at least two environment sensors, a neural network coupled to the environment sensors for fusing environment information from the environment sensors, a fusion apparatus for fusing environment information from the environment sensors, and a control device coupled to the neural network and the fusion apparatus. The control device is set up to adapt the environment information fused via the neural network, depending on the environment information fused by the fusion apparatus, and to provide the adapted environment information to a driver assistance system of the motor vehicle.