By Professor Emma Hart, School of Computing, Edinburgh Napier University
In 2020, we all are familiar with the idea of robots assisting humans in many capacities, tackling tasks that are either challenging for humans or simply repetitive, for example, underwater inspection of oil-rig infrastructure or assisting in warehouse operations. Such robots are designed by engineers, drawing on expert knowledge of both robot design and the task at hand. But how can a human engineer design a robot for a task to be conducted in an environment that is completely unknown – for example, to clean up the inside of a nuclear reactor or explore the surface of an unknown planet? In such cases, it is unclear what type of morphology the robot should have (i.e. what size it should be; whether it needs wheels, legs or grippers; what sensing capabilities are required) or what type of behaviours are required (e.g. exploration, climbing, fetching…).
The ARE (Autonomous Robot Evolution) project – a 4 year collaboration between academics from the School of Computing at ENU with the University of York, University of West of England, and the Vrije University Amsterdam, funded by EPSRC - aims to answer this question by looking to biology for inspiration and asking the question: as natural evolution has produced successful life forms for practically all possible environmental niches on Earth, is it plausible that artificial evolution can also produce specialised robots for various environments and tasks? By mimicking the processes of evolution in software inside a computer in order to design robots, and coupling this with state-of-the-art techniques in 3d-printing and automated assembly, ARE aims to create a disruptive robotic technology where colonies of robots are created, reproduce and evolve in real-time and real space. The long-term vision is a technology enabling the evolution of entire robotic ecosystems, where robots are designed, manufactured, assembled, live and work for long periods in challenging environments without any need for direct human oversight. Such robots will fundamentally change the concept of machines, showcasing a new breed change their form and behaviour over time as they adapt to changes in their environment and new tasks.
The ENU team is led by Professor Emma Hart, drawing on her many years of expertise of developing artificial evolution methods that have previously been applied in diverse domains, ranging from forestry to scheduling in factories. One aspect of her team’s research is directed towards designing appropriate methods for defining the artificial genetic material that defines both a robot’s brain and the physical characteristics of its body. Another strand their work is focused on, is developing new techniques to enable newly evolved robots to quickly learn basic skills in a simple environment before being released into a more complex world - exactly analogous to the way a child learns within a nursery setting. Almost 2 years into the 4 year project, the teams in York/UWE have already built the engineering infrastructure necessary to print and assemble robots from evolved designs without any human intervention, and (Covid-19 permitting!) we are beginning to test out the first generation of newly `born’ robots in our laboratories to evaluate their ability to learn, reproduce and adapt. Already we are noticing some unconventional designs emerging!
The project will evaluate the new approach in the context of designing robots for two nuclear decommissioning scenarios, pushing new boundaries in engineering design. However, it also has the potential to bring new insights into natural evolution itself, through creating a novel research instrument that can be used to investigate fundamental issues about evolution and the emergence of embodied intelligence in the biological world.
Watch this space!
A paper that explains the concept of the project
A keynote speech by Emma Hart which discusses the history of robot evolution and provides more details about the project