Enabling Robot Autonomy through a Modular Software Framework

The complexity of robotic software architectures stems from the need to manage a diverse range of sensory inputs, real-time actuator control, and adaptive capabilities in dynamic environments. In order to guarantee safe operation, robots must be capable of executing tasks concurrently and asynchronously, which poses significant challenges in developing cohesive robotic software architectures. It is commonly accepted that there is no universal approach that can address the needs of all robot platforms and applications. A number of established architectures have been developed based on the publish-subscribe and action-client paradigms employed by Robot Operating System (ROS) middleware. Extending on these developments, in this research, we present a novel robotic software architecture that enables seamless integration of different robotics software components, such as Planning, Control, and Perception. The presented architecture is designed to ensure the autonomous navigation of a mobile robot operating in uneven outdoor terrains, while also supporting indoor environments with appropriate customization. Our software has been made available to the robotics community through a GitHub repository.