| Summary: | This paper details the supporting of the development of a novel Simultaneous
Localization and Mapping (SLAM) algorithm for autonomous delivery robots,
emphasizing the need for a robust robotics testing platform. The research utilizes the
NVIDIA Carter Robot, an open-source Robot Operating System 2 (ROS2)-based
platform that aligns with the intended delivery robot in size and capabilities, as a
testbed for the proposed solution. The fusion of the data from LiDARs and multiple
depth sensors addresses the need for dynamic obstacle avoidance and path planning
in unseen environments.
The research encountered and addressed substantial hardware challenges, notably in
the synchronization of the drivetrain with the onboard computational unit—a critical
prerequisite for the robot's functional deployment. The need for multi-sensor and
multi-modal algorithms to be running simultaneously also introduced additional
complexities due to the limited data transfer bandwidth available. An integral
component of the research also entailed the meticulous calibration of the robot's
transformation in space within ROS to guarantee the efficacy of the SLAM
algorithms.
The results of this research demonstrate the feasibility of using multiple Intel
RealSense cameras, as well as the viability of advanced SLAM techniques in realworld scenarios, thereby paving the way for future innovations in robotic autonomy.
|
|---|