Bioinspired Control of Electro-Active Polymers for Next Generation Soft Robots

Lead Research Organisation: University of Sheffield
Department Name: Psychology

Abstract

The current generation of robots is typically built of hard, inflexible and insensitive materials, in contrast to animal bodies which use soft and elastic elements with the capacity for sensing and shape change. A revolution is beginning in robotics in which new soft-smart materials more similar to animal tissues are being introduced. One such material is the class of electroactive polymers (EAPs)-flexible plastics that can change shape when an electric current is passed through them and thus can act rather like animal muscles. However, in order to use these new materials in robotics we need to understand how to control them. Soft-smart materials pose particular problems because their response to stimulation may be complex, their material properties can change over time, and they are often fabricated with wide tolerances. Since these are the same problems posed by animal tissues it is natural to look to biology as a source of inspiration. In mammals, a part of the brain called the cerebellum, or little brain , is particularly associated with skilled control of movement, the integration of sensing with action, and the capacity to adapt to change. In this project we therefore seek to derive new ways of controlling soft materials from our knowledge of cerebellar function. We believe this work can unlock the potential of such materials leading to machines with the fluidity and grace of movement we associate with animals. This change will have many knock-on effects. Robots with soft body parts and skilled movement will be able to go places that are currently impossible for them. For instance, exploring disaster sites or remote worlds, picking their way through boulders, squeezing though crannies, whilst automatically adapting to wear-and-tear. Robots with softer bodies and flexible limbs will also be safer for humans to interact with, and thus can play a wider role in our homes, schools and hospitals.The first step will be to explore how soft materials, in this case EAP actuators, can be used as artificial muscles on two robot platforms. One will acquire tactile information from its surroundings by copying the facial whiskers of animals such as mice and rats. These animals move their whiskers back-and-forth at high-speed while exploring-in our robot these accurate, high-velocity whisking movements will be performed by EAP actuators. Our second platform will have an EAP-controlled artificial eye designed to acquire visual information during robot movement. To provide a stable visual image this robot needs to generate rapid changes in camera position that counteract movements of its body and head. The second step is the development of cerebellar-inspired control algorithms. This will lead to a software cerebellar chip that can be plugged into different tasks without prior knowledge of all the requirements-the chip will learn online how best to control the soft materials. These algorithms will be shown to be effective in solving two fundamental problems for active sensing in our EAP-based robots-recognising spurious sensory signals caused by your own movement, and controlling sensor position so as to maintain a steady gaze.The final objective is to demonstrate the versatility of the cerebellar chip by applying it to a task that combines soft-control of both tactile and visual sensing systems. Here we will train a new robot to orient accurately to novel targets whilst it also learns about the properties of its sensing and actuation systems. The underlying representation will be a head-centred sensory map of the environment that is adaptively generated and modified by experience. Performing this task in a life-like way will not only demonstrate the capacity of cerebellar algorithms to effectively control smart materials but will also show that these two components combined can bring about a step-change in robotics that will lead to a future in which we have more versatile, agile, and user-friendly robots.

Planned Impact

The central aim of this project is to develop control strategies suitable for robots constructed from soft materials using a brain-based 'universal' adaptive control algorithm. Project outcomes will impact on a wide range of robot application domains where lighter weight, lower power, adaptability to environmental and internal change, compliance and safety are important factors. Many such applications are likely to be in service or field robotics and thus outside the factory settings that provide the main market for most existing robotic systems. This quickly growing market is already worth 1B annually, with some application areas seeing a 40% p.a. growth (World Robotics 2007). The UK may have a narrow lead in some areas of non-manufacturing robotics, but this is a new and heavily research-dependent sector of the economy. Companies who could benefit from advances made in this project include those in the specialist robotics market and larger enterprises interested in increasing the role of automation in relevant application domains. We list a number of potential partners in the Pathways to Impact plan. Soft-smart materials are likely to be particularly important in service robotics where machines will be increasingly required to interact closely with people, for instance as domestic or office assistants or as hospital care-workers. In these applications safety is a prime concern, generating a need for lighter, more flexible robots with compliant control. In field robotics the potential impact is through increased reliability, autonomy, energy efficiency and reduced cost. Applications lie in agriculture, search and rescue, bomb-disposal, and operations in remote or hazardous environments (such as beneath the ocean, in disaster sites, or on other planets). A further contribution of the project will be through advances in whisker-based tactile sensing. Touch is an under-developed artificial sensing modality and many applications in field robotics will benefit from improved tactile proximal sensing capabilities, for instance where robots are required to operate in darkness, or in visually-occluded environments (e.g. dust or smoke-filled buildings, or turgid water). By advancing our understanding of the computations performed by the cerebellum the project will have significant impact in the area of brain-machine interfaces and intelligent prostheses. For instance, medical engineers are working towards controllable artificial limbs for amputees that have a bi-directional interface with the nervous system. A major hurdle is to effectively integrate sensory signals from the prosthetic device with motor control data from the device. In natural sensing this functionality is provided by the cerebellum so its absence in prosthetics could make such devices difficult to control or deliver unwanted and meaningless signals. Cerebellar filtering of sensory data and sensory correction of motor errors are core topics of our project that will speak directly to these issues. Advances in these areas may also lead to new hands-free or teleoperation technologies. The project will extend our basic knowledge of the cerebellum leading to better understanding of neurological conditions, such as dyslexia and autism, that involve this structure. The longer-term societal impact could include improved ability to diagnose/treat these disorders. Finally, biomimetic robots are an excellent way in which to capture the imagination of both children and adults. The investigators have a strong track record of involvement in public awareness events with schools and museums, and of disseminating scientific results through the media. Project outreach activities will help to promote public interest in soft robotics and to raise awareness of related growth areas of science and engineering such as soft-smart materials and computational neuroscience.
 
Description The cerebellum is the part of the brain responsible for learning skilled movements in humans and animals. We have developed computer algorithms for robotic control which mimic the operation of the cerebellum. These algorithms have been applied successfully to the control electro-active polymer (EAP)actuators, often called artificial muscles, which are expected to have extensive applications in the next-generation of 'soft' robots.
We have also developed new configurations and control paradigms for EAP devices which bring them closer to commercial utility.
Exploitation Route The algorithms we are developing have applications in all areas of adaptive robotics. After further case studies we are planning to release an algorithmic toolbox which will make them available in a convenient form.
Sectors Aerospace, Defence and Marine,Healthcare,Other

URL http://www.bellaproject.co.uk/
 
Description The grant is still in progress and the findings have not yet been used. We are discussing commercialisation opportunities with the University of Sheffield.
 
Description Human Brain Project Systems and Cognitive Neuroscience
Amount € 415,000 (EUR)
Organisation European Commission 
Department Horizon 2020
Sector Public
Country European Union (EU)
Start 04/2016 
End 03/2018
 
Company Name Consequential Robotics 
Description Company formed as a collaboration between the University of Sheffield and the Industrial Designer Sebastian Conran to commercialise developments in biomimetic and assistive robotics. 
Year Established 2014 
Impact Consequential Robotics is launching its first product in 2016, a biomimetic mammal-like robot.
Website http://consequentialrobotics.com/