Engineering news
Mobile robots have come a long way over the past few decades. While many are still experimental prototypes confined to the laboratory, an increasingly significant number are venturing forth into the real world. Yet one thing common to the vast majority of robots is that they are totally dependent on their human handlers for their day-to-day operation, whether for instruction or, more importantly, for power.
This human intervention limits the extent to which these robots are truly autonomous. There are a few exceptions, such as fully autonomous solar-powered robots that can operate for extended periods, but these generally operate in constrained or unchanging environments with unobstructed access to the sun, for example in the open ocean.
My aim is to enable the creation of robots capable of ‘living’ in complex and unstructured environments while performing their intended tasks, over periods of a year or more.
Self-sufficient, ‘free range’ robots would be ideally suited to extended-duration tasks in natural or semi-natural environments.
One promising application area is agriculture. Swarms of small self-sufficient robots could perform tasks including mechanical pest or weed control, reducing the need for environmentally harmful chemicals, and artificial pollination, mitigating the risk of long-term pollinator decline. They would also minimise the risk of damage to crops, avoid soil compaction, and safely work alongside humans by virtue of their small size.
In the shorter term, agriculture is probably the most viable application area, as farms are a semi-controlled environment. In the longer term, these robots could be deployed in more natural environments, such as forests or jungles, where they could perform tasks such as monitoring biodiversity or pollution.
In developing behavioural strategies for robots of this type, we will apply what I call a ‘holistic bio-inspired approach’ that deals with the biological systems I’m drawing inspiration from at an integrated, systems level. This approach is important, because robotic self-sufficiency is a systems-level challenge. The only way to prove that all components of the solution are valid is to build an integrated system and show that it works in the real world.
In nature, mind and body co-evolve within the constraints of the environment. The fitness of individual behaviours is evaluated in the context of the full behavioural repertoire and the physics of the situated embodiment. This holistic approach can result in elegant solutions that exploit the closed-loop interaction of system components that do nothing useful in isolation.
Initially, the main focus will be on behavioural strategies for obtaining energy from the environment – early experiments will use solar panels in an environment with a patchy distribution of light – and modulating behaviour based on current energy levels and expected energy availability, creating a system that can forage for ‘food’. Preliminary work will be conducted in simulation, but will move quickly to physical experiments on a farm.
This project won’t be easy. It will start by developing a system with behavioural complexity roughly equivalent to that of a nematode worm called Caenorhabditis elegans, before moving on to something more akin to insect-level complexity. Perhaps one day we will make it as far as mammalian-level complexity, but that won’t be any time soon.