Shoulder Object Passing Using Camera-LiDAR Confidence Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous vehicle systems struggle to accurately detect, classify, and respond to objects on the side of the road, particularly in complex and dynamic environments, due to limitations in detection distance, sensor reliability, and sensor fusion algorithms, leading to potential safety and efficiency issues.
Innovation Solution
An autonomous vehicle system utilizing a camera and LiDAR sensor suite to detect and classify side-of-the-road objects in two phases: initial presence detection using a camera system followed by higher-fidelity classification with LiDAR, enabling operational adjustments based on confidence thresholds to enhance safety and performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single sensor system is used for object detection, then the system complexity is reduced, but the detection accuracy and reliability deteriorate
Solution Approach 1:
The patent combines multiple sensor systems (camera, LiDAR, radar) into a unified sensor suite that operates synergistically. The camera provides initial object detection and classification, while LiDAR and radar provide complementary data for verification and enhanced characterization, resolving the contradiction by merging sensors to achieve both simplicity in operation and high detection accuracy
Solution Approach 2:
The detection process is segmented into multiple phases: initial detection by camera, verification by LiDAR/radar, and final classification. This segmentation allows each sensor to perform its optimized function at the appropriate stage, maintaining low operational complexity while achieving high measurement precision through phased processing
2Reliability
If detection distance is increased to detect objects earlier, then collision prevention capability is improved, but detection accuracy deteriorates due to lower resolution at distance
Solution Approach 1:
The camera performs preliminary detection and initial classification of objects at long distances, identifying potential hazards early. This preliminary action enables the system to prepare for closer inspection by LiDAR/radar before the vehicle reaches critical proximity, maintaining both early collision prevention capability and accurate classification
Solution Approach 2:
The camera acts as an intermediary sensor that bridges the gap between long-range detection and close-range verification. It detects objects at distance and passes preliminary classification information to LiDAR/radar, which then perform high-precision verification, resolving the contradiction between early detection and accurate classification
3Reliability
If multiple sensors are used for verification, then detection reliability is improved, but the processing time and system complexity increase
Solution Approach 1:
The system employs periodic verification where LiDAR and radar sensors are activated at specific intervals or triggers (e.g., when camera detects an object with certain confidence level). This periodic action reduces continuous processing overhead while maintaining high detection reliability through targeted verification at critical moments
Solution Approach 2:
The system applies partial verification by using LiDAR/radar only for objects detected by the camera that require further classification or verification. Not all detected objects undergo full multi-sensor verification, reducing processing time while maintaining reliability for critical cases through selective application of verification resources
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The system effectively detects and classifies side-of-the-road objects with improved accuracy, allowing for timely operational adjustments to prevent collisions and enhance overall vehicle safety and efficiency.
Implementation Method 1
the autonomous vehicle detects the presence of a side-of-the-road object from image data collected from a camera system of the autonomous vehicle
Implementation Method 2
additional sensors (e.g., LiDAR sensors or radar sensors) collect higher-fidelity image or point data to further classify the object
Implementation Method 3
additional sensors (e.g., LiDAR sensors or radar sensors) collect higher-fidelity image or point data to further classify the object
Data Source
AI summary
A method of passing an object on the side of the road by receiving a first signal from a first sensor configured to detect a presence of the object positioned on a shoulder of a road; determining a first value; assigning a first confidence value to a confidence level, wherein the first confidence value is associated with the first value; responsive to the first confidence value exceeding a confidence threshold, adjusting a first operating parameter of a vehicle; responsive to entering a threshold range of the object, receiving a second signal from a second sensor; determining a second value associated with the second signal; assigning a second confidence value to the confidence level, wherein the second confidence value is associated with the second value; responsive to the second confidence value exceeding a second confidence threshold, adjusting a second operating parameter of the vehicle.