LiDAR Detection Parallelization for Real-Time Obstacle Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The reliability of laser radar systems in autonomous driving is compromised by the slow data calculation speed in the sensing process, which affects their ability to accurately detect obstacles and drivable areas in real-time.
Innovation Solution
The method involves algorithm splitting to identify sub-algorithms capable of parallel processing, followed by heterogeneous acceleration using processors that support SIMD and MIMD instructions, allowing for efficient processing of scanning data to improve data operation speed and reduce CPU occupancy rates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional sequential processing algorithms are used for laser radar data, then the detection accuracy is maintained, but the data calculation speed is slow
Solution Approach 1:
The patent divides the feature detection algorithm into multiple independent sub-algorithms (e.g., ground detection, obstacle detection, drivable area detection) that can be executed in parallel. Each sub-algorithm processes specific aspects of the point cloud data independently, allowing simultaneous computation across multiple processing units while maintaining overall detection accuracy.
Solution Approach 2:
The patent transitions from single-threaded sequential processing to multi-threaded parallel processing by utilizing heterogeneous computing architectures. This dimensional shift in processing methodology enables multiple detection tasks to run concurrently, dramatically improving data calculation speed without sacrificing detection reliability.
2Productivity
If high-performance computing devices are used to speed up data processing, then the data calculation speed is improved, but the production cost increases
Solution Approach 1:
The patent designs sub-algorithms that can execute on various heterogeneous computing platforms (GPUs, FPGAs, multi-core CPUs) without requiring specialized hardware. This multi-functionality allows the system to achieve high processing speeds using cost-effective, commercially available devices rather than expensive custom-built computing systems.
Solution Approach 2:
The patent creates portable, standardized algorithm modules that can be copied and deployed across different computing devices. These sub-algorithms are designed to be platform-independent, enabling the same detection logic to run efficiently on various hardware configurations, thereby reducing development and manufacturing costs.
3Productivity
If parallel processing sub-algorithms are implemented, then the data operation speed is enhanced, but the system complexity increases
Solution Approach 1:
By breaking down the complex feature detection task into smaller, well-defined sub-algorithms with clear input-output interfaces, the patent manages system complexity through modularization. Each sub-algorithm handles a specific detection function, making the overall system easier to understand, debug, and maintain despite the parallel processing architecture.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A detection method and a device based on a laser radar, and a computer readable storage medium are disclosed. The detection method includes: obtaining scanning data of the laser radar (S101); performing an algorithm splitting on a feature algorithm for detection based on the scanning data to obtain at least one sub-algorithm capable of parallel processing in the feature algorithm (S102); and performing a heterogeneous acceleration for the at least one sub-algorithm to process the scanning data, and obtaining a detected position of an obstacle and a detected drivable area according to a processing result (SI03).