Taking inspiration from nature
Robots have, up until now, been used to great success in structured environments, such as in manufacturing plants and factories, which are tightly controlled spaces. The design of robots in manufacturing typically focuses on speed, precision and cost-effectiveness. So, sensors that would allow these robots to observe the world are unnecessary and costly, as are control systems that would make use of such sensors.
If, on a car production line, each successive car appears precisely at the same point time after time, the robot does not need a vision control system to locate the car and adjust its own position accordingly. Instead, the robot can repetitively perform the same action ‘blindly’, relying on the accurate positioning of objects in a highly structured, controlled world.
In contrast, the autonomous behaviour of robots in unstructured environments, e.g. in streets, homes and hospitals, raises a completely different set of problems. It is in these uncontrolled environments, where random events occur, that robot behaviour is not yet advanced enough to perform usefully and safely.
Natural organisms, on the other hand, have evolved to survive and flourish in unstructured environments, which are full of random, unpredictable and possibly threatening events. The success of humans and other animals in such environments is due to a number of factors, from their sensory systems and musculoskeletal structure (the hardware), to their brains (the control system), all evolved to tackle the environment in which they must survive. These attributes make natural organisms adaptable, versatile and robust. It is these design goals that researchers aim to achieve in robotics using bioinspiration.
As an example of an unstructured environment, consider a human driving down a street: other cars can pull out or brake unexpectedly, objects or people might move into the street, and any number of other unpredictable events might occur. Humans can respond instantly to such events, usually negotiating them to a safe conclusion. This process, for the driver, involves sensing, decision making and action. A driverless car would need to be able to emulate the same process to operate safely. Currently, robots are not able to do this as successfully as humans. The main deficit is in the decision-making process, which involves first processing sensory information, making sense of it, and then deciding on a safe course of action.
Given the attributes of natural organisms highlighted above (adaptability, versatility, robustness), the question is raised over what makes a suitable natural system to study for bioinspiration. Typically, for control systems, researchers will study relatively ‘simple’ systems, for instance, insects are a good choice – the number of neurons in an ant is about 250,000 compared to 90 billion in a human. On the other hand, some researchers use humans as the model system, but they tend to focus on simpler control processes such as eye movements. This is in contrast to processes like limb movements which are multi-jointed, involve a higher degree of freedom and require an understanding of spinal cord processing.
An electronically actuated robot that behaves like a realistic human eye. Learn more here.
Bioinspiration in robotic design extends to other areas outside of the nervous system and control, into the sensory, musculoskeletal systems and the body. Currently, there is much interest in emulating muscle as actuators (components that are responsible for moving or controlling) in ‘soft’ robotic systems – artificial muscle – to replace conventional actuators like motors. This is because muscles are light-weight but produce relatively high force, they perform additional roles such as springs and brakes in natural systems, and are silent in operation, reducing noise pollution.
Certain inherent, specialist characteristics of natural systems are also desirable. For instance, fish have much greater manoeuvrability than typical underwater robots due to their use of fins: this has spurred research into flippers for underwater robots.
The G6 Robotic Fish, designed by Dr Jindong Liu (Imperial College London), which moves via sensors and has autonomous navigational control. Watch the robot swim here.
Geckos can climb vertical surfaces, which has inspired the development of Gecko feet for wall climbing robots.
The Stickybot, a climbing robot designed by Professors Mark Cutkosky (Stanford University) and Sangbae Kim (MIT), which harnesses the biology of a gecko’s sticky foot. Learn more here.
Rats navigate and hunt in complete darkness, motivating the development of robot rats that rely on tactile sensing from whiskers instead of camera vision.
The Shrewbot which uses whisk contacts to build a tactile map of an area. Learn more here.
There are many more examples, but a key point to note in bioinspired design is recognising where biology has already provided a solution, evolved and refined over millions of years, to a problem that an engineer has today.
The area of bioinspired robotics has grown significantly over the last 25 years. Within the broad field specific areas have emerged with different emphases, for instance:
- Neurorobotics focuses on neural processing schemes evaluated in robotic systems.
- Biomimetics refers to accurately copying some biological system.
- Bioinspiration refers to taking some portion of design from a natural system and then integrating that design with conventional engineering.
One of the important aspects to emerge is the idea of embodied cognition, which emphasises the need to study brains and body in an integrated whole. Across all of these areas, researchers have the ambition to learn key design principles from nature that will enable robots to emerge from the factories, and take their place in wider human society, performing useful functions and operating safely.
© The University of Sheffield