Autonomous Mobile Mapping With Adaptive 2D/3D Data Resolution

Solution Overview

Problem

Existing environment recognition techniques for autonomous mobile bodies, such as voxel grid and mesh, require high calculation costs, while height maps are insufficient for operations like 'going under', necessitating a method to perform accurate recognition with reduced computational expense.

Innovation Solution

An accumulation method setting unit adjusts data resolution based on the drive mechanism, movement performance, peripheral environment, and task content of the autonomous mobile body, allowing for three-dimensional data accumulation only where necessary, with fine resolution in critical areas and coarse resolution elsewhere, thereby reducing data volume and calculation costs.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If three-dimensional environment recognition techniques (voxel grid, mesh) are used, then environment recognition accuracy is improved, but calculation cost increases

Engineering Contradiction:

Improveenvironment recognition accuracyVSAvoidcalculation cost

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent applies local quality by using different data structures for different spatial regions: three-dimensional voxel grids are used only in regions where the mobile body may travel (nearby areas), while two-dimensional height maps are used for distant regions. This selective approach maintains recognition accuracy in critical areas while reducing overall calculation cost.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The patent segments the environment recognition space into multiple regions based on distance from the mobile body. Nearby regions are processed with high-resolution 3D voxel grids, while distant regions use lower-resolution 2D height maps. This segmentation allows the system to allocate computational resources efficiently across different spatial zones.

Inventive Principle:

Principle #1Segmentation

2Measurement precision

If three-dimensional data accumulation is performed throughout the entire space, then environment recognition accuracy is improved, but data volume and calculation cost increase

Engineering Contradiction:

Improveenvironment recognition accuracyVSAvoiddata volume

Core Design Contradiction:

Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent applies local quality by using different data structures for different spatial regions: three-dimensional voxel grids are used only in regions where the mobile body may travel (nearby areas), while two-dimensional height maps are used for distant regions. This selective approach maintains recognition accuracy in critical areas while reducing overall calculation cost.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The patent segments the environment recognition space into multiple regions based on distance from the mobile body. Nearby regions are processed with high-resolution 3D voxel grids, while distant regions use lower-resolution 2D height maps. This segmentation allows the system to allocate computational resources efficiently across different spatial zones.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS20260093039A1Information processing apparatus, information processing method, and autonomous mobile body

Publication Date: 2026.04.02 SONY GROUP CORP

AI summary

An accumulation method for distance measurement data is set on the basis of a drive mechanism, movement performance, a peripheral environment, or a task content of an autonomous mobile body. For example, necessity of three-dimensional data accumulation is determined, and setting is made to perform two-dimensional data accumulation in a case where it is determined that the three-dimensional data accumulation is unnecessary.Furthermore, for example, resolution of the three-dimensional data accumulation is set for each axial direction of three axes X, Y, and Z or in accordance with a height.