Surfel Map Fusion for Accurate 3D Autonomous Driving Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous vehicle systems face limitations in representing three-dimensional environments, relying on either outdated 2.5-dimensional maps or real-time sensor data alone, which can lead to inaccurate navigation and inefficient processing.
Innovation Solution
The integration of surfel maps, which represent three-dimensional environments with multiple elements at each coordinate, combining existing surfel data with real-time sensor data to generate a more accurate and flexible representation of the environment, allowing for improved autonomous driving decisions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a 2.5-dimensional map is used to represent the environment, then the system complexity is reduced, but the measurement precision of three-dimensional features is insufficient
Solution Approach 1:
The patent transitions from 2.5-dimensional maps to a full three-dimensional surfel map representation. Each location in the environment is represented by multiple surfels at different elevations, allowing the system to capture complex three-dimensional structures like overpasses, tunnels, and vegetation while maintaining computational efficiency through the surfel data structure.
2Productivity
If existing surfel map data is used for navigation, then the processing speed is improved, but the reliability of environment representation deteriorates due to outdated information
Solution Approach 1:
The system dynamically integrates static surfel map data with real-time sensor data to create an updated environment representation. The surfel map provides a pre-computed three-dimensional structure that can be quickly accessed, while real-time sensor data refreshes the representation of dynamic elements, combining the speed of pre-processing with the accuracy of current information.
Solution Approach 2:
The surfel map is generated in advance through offline processing of historical sensor data, creating a pre-computed three-dimensional representation of the environment. This preliminary action allows the autonomous vehicle system to leverage existing knowledge about the environment without performing computationally expensive real-time processing, while still maintaining reliability through subsequent integration with current sensor data.
3Measurement precision
If traditional 3D mesh systems are used to represent the environment, then the measurement precision is improved, but the computational requirements and memory usage increase significantly
Solution Approach 1:
The patent segments the three-dimensional environment into discrete surfel units, where each surfel represents a small surface element with specific properties. This segmentation allows the system to represent complex three-dimensional structures efficiently, as each surfel is an independent entity that can be processed individually, reducing the computational burden compared to manipulating entire mesh structures.
Solution Approach 2:
The system changes the fundamental parameters of environment representation from traditional 3D mesh topology to surfel-based attributes including position, normal vector, and uncertainty measures. This parameter transformation enables more efficient storage and processing, as surfels require less bookkeeping and can be independently managed, reducing memory requirements and computational energy consumption while maintaining three-dimensional representation accuracy.
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on computer storage media, for using a surfel map to generate a prediction for a state of an environment. One of the methods includes obtaining surfel data comprising a plurality of surfels, wherein each surfel corresponds to a respective different location in an environment, and each surfel has associated data that comprises an uncertainty measure; obtaining sensor data for one or more locations in the environment, the sensor data having been captured by one or more sensors of a first vehicle; determining one or more particular surfels corresponding to respective locations of the obtained sensor data; and combining the surfel data and the sensor data to generate a respective object prediction for each of the one or more locations of the obtained sensor data.