Lifelong Exploratory Navigation: An Architecture for Safer Mobile Robots

This article studies a layered architecture for robotics design, proposes a contract-based interface between the layers, and shows how cross-layer adaptation between these layers can be done in response to different scenarios. Mobile robots are highly complex systems whose development covers a large number of fields, from actuator design and mechatronics to scene processing and lifelike interaction with humans. Advanced robotic systems use a layered architecture where each layer represents an abstraction level, with sensors and actuators at the bottom and semantic artificial intelligence (AI) at the top. This article presents a layered architecture, called lifelong exploratory navigation (LEN) for mobile robots, encompassing ‘stacked’ and confined layers going from the physical level to AI functionalities. LEN is capable of autonomous adaptation, so that the robot is never shut down or reset, even when it is transferred to an unknown environment.