Grid-Based Sensor Fusion for Parking Obstacle Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional sensor fusion methods for parking assistance systems in transportation apparatuses face high computational load due to complex mathematical equations, leading to processor resets and difficulty in enhancing performance when combining ultrasonic and camera sensor data.
Innovation Solution
A sensor fusion apparatus and method that includes preprocessing and grid mapping for each sensor, followed by clustering using an integrated grid map, to reduce computational load and improve accuracy by utilizing the field of view of the ultrasonic sensor and intersection areas of direct and indirect waves.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the Probabilistic Data Association Filter (PDAF) is used for sensor fusion, then identification accuracy is improved, but computational load increases causing processor reset
Solution Approach 1:
The patent divides the sensor fusion process into discrete grid cells, where each grid independently processes sensor data. This segmentation transforms the complex global fusion problem into multiple simple local fusion tasks, dramatically reducing computational load while maintaining accuracy through the integration of grid-level results.
Solution Approach 2:
The patent changes the mathematical approach from probabilistic filtering (PDAF) to a grid-based occupancy representation. By representing sensor data as occupancy scores in grid cells rather than probabilistic distributions, the system achieves comparable accuracy with significantly reduced computational complexity suitable for embedded processors.
2Measurement precision
If the Probabilistic Data Association Filter (PDAF) is used for sensor fusion, then identification accuracy is improved, but system performance enhancement becomes difficult
Solution Approach 1:
The grid map structure serves multiple functions: it represents occupancy, stores sensor measurements, enables visualization, and supports path planning. This universal data structure allows the same system to perform sensor fusion, mapping, navigation, and obstacle avoidance without requiring separate processing pipelines, thereby enhancing adaptability and performance improvement potential.
3Device complexity
If ultrasonic sensor alone is used for obstacle identification, then device complexity is reduced, but measurement precision deteriorates
Solution Approach 1:
The patent merges data from ultrasonic sensors and camera sensors into a unified grid map representation. By combining the distance measurement capability of ultrasonic sensors with the visual recognition capability of cameras in the same coordinate framework, the system achieves superior obstacle identification accuracy while maintaining manageable system complexity through integrated processing.
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
Reduces processor load by approximately 10 percentage points and enhances identification accuracy by simplifying mathematical models, making the system more robust against noise.
Implementation Method 1
acquiring a first Time of Flight (TOF) value of a direct wave and a second TOF value of an indirect wave detected through a first sensor of the two or more sensors
Implementation Method 2
acquiring object detection information from image data detected through a second sensor of the two or more sensors through the second preprocessing
Data Source
AI summary
A sensor fusion apparatus for a transportation apparatus including two or more sensors, each of the two or more sensors having different sensing characteristics, each of the two or more sensors sensing an object, respectively, one or more processors configured to execute instructions, and a memory storing the instructions, an execution of the instructions configures the one or more processors to perform first preprocessing and second preprocessing, the first preprocessing and second preprocessing including grid mapping for each object sensed by two or more sensors, respectively and perform sensor fusion through clustering using an integrated grid map for each grid-mapped object.