Sparse Columnar Voxel Model for Robot Navigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing 3D robot navigation techniques face inefficiencies in memory consumption and computational cost due to allocating memory for every voxel in a 3D grid, regardless of occupancy, and standard ray-tracing methods are costly for detecting dynamic objects.
Innovation Solution
Organizing a voxel-based 3D model into columns with sparse data structures, where only occupied voxels are allocated memory and indexed, using linked data structures like linked lists or binary trees for efficient data management and object detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If memory is allocated for every voxel of the 3D grid regardless of occupancy, then complete 3D model coverage is achieved, but memory consumption increases significantly
Solution Approach 1:
The patent applies local quality by differentiating between occupied and unoccupied voxels in the 3D grid. Memory is allocated only to occupied voxels that contain sensor data, while unoccupied voxels are excluded from the data structure. This creates a heterogeneous memory allocation strategy where different regions of the 3D space receive different treatment based on their occupancy status, resolving the contradiction between model completeness and memory efficiency.
Solution Approach 2:
The patent segments the 3D voxel grid into occupied and unoccupied regions. By using sparse data structures that only represent occupied voxels, the system divides the complete 3D model into meaningful segments (occupied spaces) and excludes irrelevant segments (unoccupied spaces). This segmentation allows the robot to maintain a complete model of occupied regions while avoiding memory allocation for empty regions.
2Measurement precision
If standard ray-tracing techniques are used to detect dynamic objects, then object detection accuracy is maintained, but computational cost increases
Solution Approach 1:
The patent extracts and removes unoccupied voxels from the 3D data structure, creating a sparse representation that contains only occupied voxels. This extraction reduces the total number of voxels that need to be processed during ray-tracing operations, thereby reducing computational cost while maintaining detection accuracy for actual objects in the environment.
Solution Approach 2:
Instead of performing ray-tracing on the complete dense voxel grid, the patent applies partial action by limiting ray-tracing operations to only the occupied voxels that are stored in the sparse data structure. This partial processing approach maintains object detection accuracy for relevant regions while avoiding unnecessary computational expenditure on empty regions.
3Measurement precision
If comprehensive voxel-based models are used for localization, then localization accuracy is improved, but processing speed decreases
Solution Approach 1:
The patent segments the 3D environment model into occupied and unoccupied voxels, storing only occupied voxels in a sparse data structure. During localization, the robot processes only the segmented occupied regions rather than the complete dense grid, maintaining localization accuracy for actual environmental features while reducing processing time by excluding empty spaces.
Solution Approach 2:
The patent changes the density parameter of the voxel model from dense (all voxels allocated) to sparse (only occupied voxels allocated). This parameter change transforms the data structure to contain only relevant information, improving localization speed by reducing the search space while maintaining accuracy through precise representation of occupied regions.
Data Source
AI summary
Methods, apparatus, systems, and computer-readable media are provided for efficient mapping of a robot environment. In various implementations, a group of data points may be sensed by a three-dimensional sensor. One or more voxels of a three-dimensional voxel model that are occupied by the group of data points may be identified. For each occupied voxel, a column of the three-dimensional voxel model that contains the occupied voxel may be identified. Occupied voxels contained in each column may be indexed by elevation. In various implementations, one or more sparse linked data structures may be used to represent the columns.