3D Sensor Fusion and Object Tracking for Autonomous Vehicles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems for autonomous vehicles, drones, and robots struggle to efficiently collect and process spatial information from various sensors to ensure stable and accurate operation, particularly in identifying and tracking objects in three-dimensional spaces.
Innovation Solution
A vehicle and sensing device equipped with sensors, a neural network-based object classification model, and a processor to identify, track, and distinguish objects in a 3D space, using a combination of LiDAR, radar, camera, and other sensors, and communicate with external devices to enhance spatial information acquisition and processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple sensors are used to sense 3D space successively, then measurement precision and reliability are improved, but device complexity increases
Solution Approach 1:
The patent combines multiple sensor types (LIDAR, radar, cameras, ultrasonic sensors) into an integrated sensor unit that collectively senses the 3D space. The processor merges data from all these sensors to create comprehensive spatial information, achieving high measurement precision through sensor fusion while managing complexity through unified processing architecture.
Solution Approach 2:
The processor performs multiple functions including sensing spatial information, identifying objects, tracking objects, distinguishing object areas from ground areas, and controlling vehicle operations. This multi-functional approach consolidates what would otherwise require separate systems, improving precision without proportionally increasing complexity.
2Measurement precision
If neural network based object classification model is applied to identify objects, then object identification accuracy is improved, but processing time and computational complexity increase
Solution Approach 1:
The system performs preliminary actions by first distinguishing object areas from ground areas before applying the neural network for detailed object identification. This pre-processing step reduces the data volume and complexity for the neural network, maintaining high accuracy while reducing processing time.
Solution Approach 2:
The patent segments the spatial information processing into distinct stages: sensing, spatial information extraction, object area identification, ground area separation, and final object classification. This segmentation allows each stage to be optimized independently, reducing overall processing time while maintaining accuracy through the neural network.
3Measurement precision
If object areas are clustered to distinguish individual objects, then object tracking precision is improved, but computational complexity increases
Solution Approach 1:
The patent applies segmentation by clustering object areas into distinct individual object regions. This segmentation separates overlapping or adjacent objects into discrete trackable entities, improving tracking precision while managing complexity through efficient clustering algorithms that group spatially related points.
4Speed
If real-time tracking and control is implemented, then operational responsiveness is improved, but processing speed requirements and system complexity increase
Solution Approach 1:
The system implements continuous sensing, continuous spatial information extraction, continuous object tracking, and continuous vehicle control adjustments. This continuous operation ensures real-time responsiveness by maintaining an ongoing feedback loop between sensing and control, achieving fast response through persistent processing rather than periodic updates.
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
Provided are a vehicle and sensing device for successively sensing a three-dimensional (3D) space using at least one sensor, acquiring spatial information over time for the sensed 3D space, applying a neural network based object classification model to the acquired spatial information over time to identify at least one object in the sensed 3D space, tracking the sensed 3D space including the identified at least one object, and using information related to the tracked 3D space, and a server for the vehicle and sensing device.