Date of Award

2-1-2022

Publication Type

Thesis

Degree Name

M.A.Sc.

Department

Electrical and Computer Engineering

Keywords

Collaborative mapping, Computer vision, Map merging, Multi-vehicle, Occupancy grid mapping, SLAM

Supervisor

B. Shahrrava

Supervisor

S. Alirezaee

Rights

info:eu-repo/semantics/openAccess

Abstract

An accurate map of the environment is essential for autonomous robot navigation. During collaborative simultaneous localization and mapping, the individual robots usually represent the environment as probabilistic occupancy grid maps. These maps can be exchanged among robots and fused to reduce the overall exploration time, which is the main advantage of the collaborative systems. Such fusion is challenging due to the unknown initial correspondence problem. This thesis presents a novel feature-based map fusion approach through detecting, describing, and matching geometrically consistent features present in the overlapping region between the maps. The main drawback of usual feature-based approaches is the incapability to establish adequate valid feature correspondence primarily due to noisy sensory observation. Further, many existing map fusion approaches neglect the heterogeneity which arises due to different map resolutions and types of mapping sensors. This thesis shows that exploiting the probabilistic spatial information to refine the maps and utilizing nonlinear diffusion filtering to detect features can drastically improve the feature-matching performance. Additionally, this thesis presents a certainty grid fusion approach based on Bayesian inference to fuse pair-wise grid information. It also presents an extensive comparison of traditional feature detection methods to register map images at different scales. Finally, the effectiveness of the proposed method is illustrated based on the following map fusion assumptions using real-world data: homogeneous, hierarchical, and heterogeneous (fusing different resolution maps and maps generated using different types of mapping sensors).

Share

COinS