Structure Aided Odometry (SAO) : A Novel Analytical Odometry Technique Based on Semi-Absolute Localization for Precision-Warehouse Robotic Assistance in Environments with Low Feature Variation
| Autor: | Suman Chakravorty, Zohaib Hasnain, Mohamed Naveed Gul Mohamed, Kartik Prakash |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
Mean squared error
Orientation (computer vision) business.industry Computer science Mechanical Engineering Industrial and Manufacturing Engineering Reduction (complexity) Odometry Artificial Intelligence Control and Systems Engineering Feature (computer vision) Robot Computer vision Ceiling (aeronautics) Artificial intelligence Electrical and Electronic Engineering business Scale (map) Software |
| Zdroj: | Journal of Intelligent & Robotic Systems. 102 |
| ISSN: | 1573-0409 0921-0296 |
| DOI: | 10.1007/s10846-021-01427-w |
| Popis: | In this study, a novel semi-absolute method of localization and subsequent odometry: Structure Aided Odometry (SAO) for precision navigation in low-feature large scale environments, is proposed and validated. Storage racks and ceiling corrugation patterns are used to calculate orientation, cross-track, and along-track estimates for a warehouse robot to produce odometry with minimal drift. A comparison is made with existing laser odometry techniques in both simulated and real (albeit controlled) test environments. The results show a promise of low drift odometry for warehouse robots inside the aisles by offering a drift reduction > 75%. A hybrid odometry technique is also proposed which combines existing odometry with structural cues. It is further validated in simulated environments containing multiple aisles as well as a full warehouse. The results show that augmenting laser with structural features reduces the average root mean square error (RMSE) in both along-track and cross-track directions by at least 70%. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full text from SpringerLink