Visual Simultaneous Localization and Mapping (VSLAM) is a crucial algorithm used in
mobile robots to determine
their position relative to the encirclement environment. In multirobot architectures,
mobile robots communicate with a base
station or a centralized edge device. However, the computational complexity nature
of the VSLAM algorithm, due to
the processing of massive visual information for tracking and
mapping, poses challenges for deploying the algorithm entirely
on the mobile robot. To address these limitations, this paper
introduces an efficient architecture that partitions the V-SLAM
framework between mobile robots and a centralized edge. This
approach complies with resource limitations and optimizes energy
consumption while minimizing the robot’s size and weight. In
addition, the paper proposes a communication module comprising an encoding and decoding
framework to obtain effective
data communication between the mobile robots and the edge
device. The performance of the proposed system is evaluated
and compared with a corresponding architecture that employs
the baseline JPEG technique in terms of trajectory accuracy,
data quality, and execution time.
