Insects SPARK robot idea
The natural world has inspired European researchers to develop intelligent, autonomous robots that not only move like insects, but also react like them.
As part of the EU-funded spatial-temporal array computer-based structure (SPARK) project, researchers created three robots to demonstrate their electronic and mechanical innovations — one featuring the complex 'brain' algorithms and two robots housing complicated control processing systems.
'The idea was to have a computational model of an insect brain embedded into a simple wheeled robot, and the other robots have six legs to mimic insect architectures,' said Prof Paolo Arena, project co-ordinator at the University of Catania in Italy.
The other project partners are Germany's University of Bielefeld, Edinburgh University, Hungary-based specialists in image processors AnaLogic, Spanish integrated circuits company AnaFocus and the Universidad Complutense de Madrid.
After observing insects the researchers decided to design robots that would mechanically mimic cockroaches, which Arena said were more robust than other insects. 'They have an optimum architecture and body in terms of autonomy and ability to run over rough terrain. The hind legs of the cockroach are very robust as they are devoted to pushing, so by copying the structure of the cockroach, we have put the heaviest load on the hind legs, for example, the batteries. They are able to carry around 50kg per centimetre,' he said.
To perceive its surrounding environment, the robot is equipped with distance and contact sensors, such as AnaFocus's Eye-RIS, a real-time visual processing system. It also has control over its directions to help it avoid large obstacles. This is supported by an avoidance reflex (found in stick insects) that boosts the robot's climbing capabilities.
As well as developing a robot that physically works like an insect, the European team designed complex algorithms to create the insect brain architecture, which was implemented and demonstrated on a wheeled robot.
'Typically the insect brain architecture is a layered structure. On the lowest level, you have parallel sensory motor pathways that directly link particular sensors to particular actuators,' said Arena. He referred to this layer as the precognitive behaviour — its basic inherited instincts. For instance, when a female cricket hears the calling song of a male cricket, she is attracted to it instinctively.
Arena described the structure of the insect brain as having, at the lowest level, the basic survival instinct to move safely in an environment, and upper layers to give meaning to the instinctive action, in order to obey an overall mission from the top.
The lowest layer of the brain architecture contains a number of parallel sensory motor pathways, such as a phonotaxis system or an optomotor reflex. A phonotaxis system enables a robot to process a sound signal and locate its source (thus mimicking crickets), while the optomotor reflex would allow a robot to maintain its sense of direction and avoid obstacles when carrying out its mission.
The activity at the lowest level is controlled and modulated by the upper layers of the brain architecture, which put the instinctive actions into context to determine the best response. The resulting action of the robot is a combination of its instinctive, precognitive behaviour and the contextualised modulation of that behaviour.
'As you go up and up in these layers, you are able to incrementally learn from experience and from the parallel sensory motor pathways,' he said.
The uppermost level of the brain is where the robot's motivation or mission is input, and which guides the learning stage.
'If you have obstacles, like a sliding door, that obscures a target, the robot is able to rearrange the behaviour modulation to give more relevance to certain kinds of parallel sensory motor pathways; for example, to reach a sound source, it weights the phonotaxis much less than avoidance. This makes the robot find an alternative path to reach the target,' said Arena.
Having established these prototypes, the researchers have recently started SPARK II, a follow-up project that aims to integrate the complex cognitive algorithms on the wheeled robot with the six-legged robot body, to make a single insect-inspired robot.
'On the tail of SPARK II, a very important challenge is to think about endowing the same architecture into two different robots and give the same mission to the robots. They will learn different things because their brains are based on different structures, so they will develop very differently but due to the fact that the mission is the same, they will automatically and autonomously co-operate with each other,' said Arena.