Vehicle 3D Map Alignment for Asynchronous Point Clouds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for generating and updating global 3D maps using multiple tridimensional sensors on moving vehicles face challenges in aligning and merging point clouds due to sensor misalignment, non-overlapping fields of view, and synchronization issues, especially in convoys of vehicles, leading to inaccurate map generation and limited usability.
Innovation Solution
A method and system where multiple tridimensional sensors generate continuous streams of point cloud frames asynchronously, with a central processing unit aligning and merging these frames into a global cumulated map in a common coordinate system without requiring synchronized data acquisition or overlapping fields of view, using techniques like Iterative Closest Point algorithms to handle misalignment and non-overlapping data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If multiple tridimensional sensors are mounted on distant vehicles in a convoy, then the coverage and scope of the generated map is improved, but the sensors experience significant relative motion and non-overlapping fields of view, making point cloud combination impossible
Solution Approach 1:
The system dynamically adjusts the transformation between sensor coordinate systems based on real-time pose estimation. Instead of relying on fixed overlapping fields of view, the system continuously updates the relative positions and orientations of sensors on different vehicles, allowing point clouds to be combined even when fields of view do not overlap at any given moment.
Solution Approach 2:
The system introduces an intermediary global coordinate system that mediates between multiple sensor coordinate systems. By transforming all point clouds into this common global reference frame using estimated pose information, the system enables combination of point clouds from sensors with non-overlapping fields of view, as the global coordinate system acts as a bridge connecting disparate sensor viewpoints.
2Measurement precision
If sensors are carefully synchronized to compare point cloud frames, then accurate displacement determination is possible, but this is difficult to meet in practice especially when sensors are mounted on distant vehicles
Solution Approach 1:
The system uses self-service by leveraging the point cloud data itself and environmental features to automatically determine sensor poses and transformations without requiring external synchronization signals. The pose estimation algorithm autonomously identifies corresponding features between point clouds and computes the transformation, eliminating the need for complex synchronization hardware or protocols.
Solution Approach 2:
The system replaces mechanical synchronization mechanisms with computational methods. Instead of using synchronized clocks or trigger signals to coordinate sensor acquisitions, the system uses algorithmic pose estimation that can handle asynchronous data. The computational approach substitutes for physical synchronization infrastructure, reducing device complexity while maintaining measurement precision.
3Manufacturing precision
If 3D measurement tools and interfaces are used to determine sensor positions and orientations, then calibration accuracy is improved, but the system becomes difficult to manage for layman operators
Solution Approach 1:
The system performs self-calibration by automatically estimating sensor poses using point cloud data and environmental features. The pose estimation algorithm autonomously computes transformation parameters without requiring manual intervention or specialized calibration tools. This self-service approach maintains calibration accuracy while eliminating the need for operators to manage complex 3D measurement interfaces.
Solution Approach 2:
The system discards the need for manual calibration procedures and specialized measurement tools, recovering calibration accuracy through automated pose estimation. By replacing manual calibration processes with algorithmic methods that continuously estimate sensor positions and orientations from operational data, the system achieves both high precision and ease of operation.
4Reliability
If sensors are mounted on a single rigid structure, then sensor relative positions remain fixed and stable for reliable merging, but this precludes use in convoys of independently moving vehicles
Solution Approach 1:
The system transitions from static sensor mounting to dynamic pose estimation. Instead of relying on fixed relative positions within a rigid structure, the system continuously estimates the time-varying poses of independently moving vehicles. This dynamic approach maintains merging reliability by compensating for vehicle independence through real-time transformation calculations based on GPS, inertial, and visual data.
Solution Approach 2:
The system introduces a global coordinate system as an intermediary that connects sensors on independently moving vehicles. By transforming all sensor data into this common reference frame using estimated poses, the system enables reliable point cloud merging despite the lack of a rigid connecting structure. The global coordinate system mediates between disparate sensor platforms, maintaining reliability while allowing vehicle independence.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method and a system for dynamically generating and updating a global tridimensional map of an environment surrounding one or several moving vehicles (10) on which N tridimensional sensors (21) are mounted and communicates with a central processing unit (22). Each sensor (21) generates a continuous stream of point cloud frames, in parallel and asynchronously with the other sensors, the point cloud frames are representative of object surfaces located in a local volume of the environment surrounding each sensor. The central processing unit continuously receives the continuous streams from the sensors, store them in a memory and, for each newly received point cloud frame of each stream, generates or updates a global cumulated tridimensional map of the environment of said at least one vehicle by determining an aligned point cloud frame in a global coordinate system of the environment, and updating the global cumulated tridimensional map by merging the aligned point cloud frame.