Mobile Agent SLAM Mapping with Manhattan Orientation Correction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional SLAM algorithms for mobile agents suffer from orientation drift due to measurement errors, especially in environments without direct orientation measurements, such as buildings or areas with external magnetic field interference.

Innovation Solution

Integration of the Manhattan world concept into a graph-based SLAM algorithm, which includes Manhattan nodes and edges to correct pose estimation by aligning the reference coordinate system with the Manhattan orientation, using sensor data to detect and validate the Manhattan orientation in each acquisition cycle, and applying Manhattan normalization to determine the dominant orientation of rectilinear structures.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If conventional SLAM algorithms are used for pose estimation, then the system can operate without direct orientation measurements, but orientation drift accumulates due to measurement errors

Engineering Contradiction:
Improveability to operate without direct orientation measurementsVSAvoidorientation accuracy
Core Design Contradiction:
Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent introduces Manhattan orientation as an intermediary constraint that mediates between sensor measurements and pose estimation. By assuming walls are aligned with cardinal directions (North-South-East-West), the system uses this intermediate reference frame to correct orientation drift without requiring direct orientation sensors like compasses.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent changes the parameter space by introducing Manhattan orientation angles as additional parameters to be estimated. Instead of directly estimating absolute orientation, the system estimates the deviation from Manhattan alignment, which serves as a constraint to reduce orientation drift in the pose estimation process.

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If Manhattan world assumption is applied to correct orientation drift, then pose estimation accuracy improves, but the system requires environments with rectilinear structures

Engineering Contradiction:
Improvepose estimation accuracyVSAvoidapplicability to different environments
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent applies partial Manhattan world assumption by using only the extent to which walls align with cardinal directions. The system calculates a Manhattan alignment metric and applies constraints proportionally, allowing it to function partially in non-ideal environments rather than requiring perfect Manhattan structures.

Inventive Principle:
Principle #16Partial or excessive action

Solution Approach 2:

The patent applies Manhattan constraints locally to wall segments that exhibit rectilinear characteristics rather than requiring the entire environment to conform to Manhattan geometry. The system identifies local regions with strong Manhattan alignment and applies constraints selectively in those regions.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP4055459B1Computer-implemented method for creating a map of the surrounding area for the operation of a mobile agent
Publication Date: 2024.05.08 ROBERT BOSCH GMBH
  • EP4055459B1 patent drawingFigure 1
  • EP4055459B1 patent drawingFigure 2~3
  • EP4055459B1 patent drawingFigure 4

AI summary

The invention relates to a computer-implemented method for operating a mobile agent (1) in a surrounding area on the basis of a pose of the mobile agent (1), wherein a localisation of the mobile agent (1) is carried out for the operation of the mobile agent (1), with the following steps: - capturing (S1) sensor data relating to walls (2) and/or objects (3) located in a surrounding area, the sensor data specifying the alignment and distance of the walls (2) and/or objects (3) in an agent-fixed agent coordinate system (A); - determining (S11-S18) the pose of the mobile agent (1) in the surrounding area with the aid of a SLAM algorithm taking into account a Manhattan orientation.