In certain key applications like surveillance, search and rescue, home care, modular robots should be capable of performing a wide variety of tasks ranging from sensing and manipulation to mobility in all types of terrains. In other words, there is a need for modular robots that display a set of robust behaviors that adapt to an uncertain and changing environment. In order to achieve this, existing design principles used in robot development need to be assessed, revised, adapted and possibly replaced by next-generation design knowledge, methods and tools that assist in the design and development of autonomous self-reconfigurable modular robotics. Such evolution of design principles is non-trivial since they need to account for a dual-layered complexity: first, internal complexity of the modular design in order to optimise components and interactions between modules; second, external complexity in order to maximise the capabilities of the system to sense, learn and self-reconfigure taking advantage of its modular structural and functional architectures.
In this project, we propose to develop a set of modular robots “RoboMods” and a set of design principles for modular robots that are capable of autonomous self-reconfiguration to handle adaptation required for a wide range of complex tasks in unknown environments as well as self-repair abilities across varying levels of failures. The following key objectives are identified to this end:
- Design and development of modular robot hardware with each module consisting of a set of sensors, actuators, controllers and power supply. Mechatronic design approaches will be explored that allow for seamless assembly and dis-assembly of modules in complex multi-modal structural and functional systems.
- Design and development of evolutionary behaviour control approaches for effective sensing, locomotion and navigation of the various configurations. Sensor fusion algorithms that use combination of individual intelligence and neighbourhood interactions of adjacent modules would be developed to optimise the global behavior and to enable development of new behaviours of the multi-module system.
- Design and development of autonomous self-reconfiguration algorithms that enable optimal transformation from one form to another. Behavioural algorithms targeted at optimal configuration selection (energy/terrain/task/time), self-repair abilities across varying failure units and parallel-motion will be developed. This functional architecture will enable the combination and adaptation of behaviours as a function of the interaction of the multi-module system with a dynamic environment.
- Design and development of Human-Robot Interaction (HRI) principles, methods and tools that enable operators to control, monitor and interact with the RoboMods so that the levels of autonomy (from direct control to full autonomy as well as various semi-autonomous levels) and the levels of controllable combinations (from controlling individual modules to controlling a fully-combined RoboMod system as well as controlling configurations in between) can be effectively, easily and gracefully switchable and handled during the course of the robots’ continuously evolving formations and actions.
- Evaluate, adapt and/or replace current design principles used in robotics design in order to support the design process of autonomous, self-reconfigurable modular robots. There are a number of challenges involved in this evolution of design knowledge:
- Existing design principles are based on the assumption that a robot is an autonomous self-contained individual unit that interacts with the environment, which may include terrain, humans and other equally autonomous robots. Here we perceive an opportunity to develop next-generation design principles to support the design of inter-dependent robotic modules that self-organise and self-repair in a system with a common set of goals and are embedded in a shared environment.
- To the best of our knowledge, little or no design principles exist that support the design of modular structural and functional architectures designed for self-assembly and dis-assembly as a function of their ongoing interaction with the environment.
- Existing design principles largely fail to account for autonomous learning capabilities and their role in combining and evolving pre-defined behaviors in modular architectures.
This design-centered project has the potential to propose new meta-design principles, i.e., knowledge, heuristics, methods and tools for the design of machines that are able to repair, change and ultimately self-design themselves.