Autonomous Vehicle Positioning During Indoor LiDAR Failure
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional autonomous driving vehicles face challenges in precise position estimation and stability when a LiDAR sensor malfunctions indoors, as they rely on GPS signals and LiDAR data, and Visual-Simultaneous Localization and Mapping (SLAM) using cameras requires significant computing power.
Innovation Solution
An autonomous driving vehicle equipped with sensors for direction, position, and velocity data, a camera for tracking movement, and a processor that uses image data to estimate position through feature points, particularly employing the ORB algorithm and Local Bundle Adjustment algorithm to guide the vehicle to a safe zone when the LiDAR sensor fails.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Visual-Simultaneous Localization and Mapping (SLAM) using a camera is used for global localization, then positioning capability is improved, but computing power requirements increase significantly
Solution Approach 1:
The system performs preliminary actions by storing image data at predetermined intervals during normal operation before LiDAR failure occurs. This pre-captured image data is then reused for positioning after LiDAR failure, avoiding the need for continuous high-computation Visual-SLAM processing and reducing computing power requirements while maintaining positioning capability.
2Measurement precision
If LiDAR sensor is used for positioning, then measurement precision is improved, but system reliability deteriorates when LiDAR malfunctions
Solution Approach 1:
The system prepares compensatory measures in advance by continuously storing image data during normal LiDAR operation. When LiDAR malfunction occurs, this pre-stored image data serves as a cushion that allows the system to maintain positioning functionality without complete failure, thereby improving system reliability and stability during abnormal conditions.
Solution Approach 2:
The camera and stored image data act as an intermediary solution when LiDAR fails. Instead of direct LiDAR-to-positioning connection, the system uses pre-captured image data as a mediator to enable positioning through comparison with live camera feeds, maintaining system functionality despite LiDAR malfunction.
3Reliability
If multiple sensors are used for redundancy, then system reliability is improved, but device complexity increases
Solution Approach 1:
The camera serves multiple functions: it captures images for normal Visual-SLAM operation, stores image data for later positioning after LiDAR failure, and provides a backup positioning mechanism. This multi-functionality reduces the need for separate dedicated backup sensors, thereby improving reliability without proportionally increasing device complexity.
Data Source
AI summary
An autonomous driving vehicle includes: a sensor section installed in the autonomous driving vehicle to sense direction data, position data, and velocity data of the autonomous driving vehicle; a camera configured to track movement of the autonomous driving vehicle and to estimate a position; a LiDAR sensor installed in the autonomous driving vehicle to generate LiDAR data; and a processor configured to receive the LiDAR data to generate a map. The processor stores image data for each predetermined driving distance using the camera. The processor also estimates a current position of the autonomous driving vehicle based on the image data stored in a predetermined range, based on a position where an abnormality occurs, and based on current image data captured while being driven in a failure mode in which the abnormality occurs in the LiDAR sensor.