Co-operation and swarming
As robots increase in number, their activities will need to be co-ordinated. There will be opportunities for robots to co-operate on tasks, and to have similar robots act as a collective (robot swarming) to complete tasks. Nature has been used as an inspiration for multi-robot co-ordination. Robot swarms allow simple, inexpensive modular robotic units to work as a team. Swarms can act as force multipliers, able to do the same task(s) as larger more expensive single robots and, by co-ordination, solve complex problems. The advantages of multi-robot systems include:
- redundancy – if one robot breaks down, the mission can continue
- scale – multiple robots can cover large spatial distances or domains
- flexibility – by distributing capabilities across multiple platforms, more flexible and adaptive robot systems are produced
- adaptability – robots can come together in the best configuration to solve one task, then reconfigure for a different task [USR16].
Swarms rely on evolutionary algorithms and decentralised intelligence to produce complex behaviours, but these have generally only been applied to swarms of the same type of robot [USR16]. Depending on the task, it makes sense for some swarms to contain a range of robots with complementary skills to give them more flexibility [USR16]. Imagine combining UAVs to give an aerial view of the environment, with ground robots that can be tasked to take action depending on the information they receive from the sky. The resulting system is a heterogeneous team, that may consist of different dynamic configurations, sensing capabilities, spatial footprints, or behavioural strategies [USR16].
Centralised intelligence that works in real-time is required to provide a means of communication between multiple agents, while allowing local control of single robots to adapt to unexpected events. The coordination of multiple robots requires robust and systematic communication protocols, which have not yet been widely agreed [AAS18]. The multi-disciplinary nature of multirobot systems is challenging, making it hard to design robust decision-making strategies in a dynamic environment where communication links may be broken. An emerging research area is communication-aware robotics – robots that actively move to maintain a communications network. Advances in the science of robot swarms requires swarms to be responsive to human commands, adaptable to changing conditions, robust to disturbances, and resilient to failure and change [SM18].