Biology-inspired Robotics
Many animals have versatile and robust sensory skills or useful and interesting behaviour that researchers would like to be able to exploit in robotic systems. Basic research in biorobotics investigates the mechanisms that enable animals to perform as they do, by building working models of the mechanisms that may be responsible for their abilities. These models are normally robots, and the fundamental thrust of the research may be described as "doing biology with robots".
A second, alternative, way of transferring good ideas from Nature to robotics is bio-inspired robotics. Here, there is less attention paid to modelling the inspirational animal(s) - the biology - and more to replicating the desired properties or skills of the animals using engineering methods.
Biorobotics and biologically-inspired robotics research takes place within the new Centre for BioRobotics, established by University of Southern Denmark (SDU) late in 2013. The new Centre is funded by a 5-year grant from SDU and is a cross-faculty collaborative venture between The Maersk Mc-Kinney Moller Institute and the Department of Biology (SDU).
The scientific goal of the Centre for BioRobotics is to tackle questions arising from the acoustic behaviours of animals which cannot be addressed completely by biology or technology alone, while also exploring applications of insights from biology in the robotics and technological domain and vice versa. Practically, the intention is to strengthen existing synergy and use it to create new opportunities for seeking external funding.
The Centre’s programme comprises work on the hearing system of lizards, the coordination of calling behaviour among large groups of frogs, and the echo-locating behaviours of bats, as well as bio-inspired embodied artificial intelligence and neurobotics including research and development on the neural control of insect locomotion using legged robotic platforms.
Core research in the Centre consists of three themes:
- Embodied Sensorimotor Behaviour Theme (Lizards)
- Embodied Communication and Acoustic Stigmergy (Frogs)
- Embodied Acoustic Scene Analysis (Bats)
Lizards have one of the most directionally hearing systems of all vertebrates, which they achieve by the trick of having a wide tube connecting their middle ears together. This allows as much sound to reach the inside of the eardrum from the opposite side of the head (through the tube) as from the outside. Mixing these two sound paths results in interference that converts a very small amplitude difference between the two sides of the head into a large amplitude difference in eardrum vibration.
A lizard robot, built in collaboration with the Department of Biology, allows us to explore the properties of this system in a real environment. The ear anatomy is modelled by a set of simple digital filters and the eardrum vibration signals generated by the model are used to control the robot behaviour, for instance to cause it to turn toward the side of the eardrum that vibrates more strongly.
Embodied Sensorimotor Behaviour Theme (Lizards)
Work in the Centre focusses on extending this lizard model to include aspects of the lizard’s neural processing of sound, and to deploy the model in a number of simple application systems - for instance, in a wheelchair that comes when you whistle, or a camera that looks at the person speaking. The ear model will also be integrated on the neurobotic walking robot referred to below, as one of its sensory input streams.
Embodied Communication and Acoustic Stigmergy (Frogs)
While lizards have acutely directional hearing, frogs make loud noises! The purpose of the frog calls, made by the males, is to attract females; however, the calling males must simultaneously avoid coming to the attention of predators. How frogs individually choose to time their calls, in the context of the large group in which they generally sit, is an interesting research question.
In the Centre, we are investigating this question from an agent-based perspective: new simulation software, RANA, has been developed which allows us to model the calling of many individual frogs in a way faithful to the timing and physical constraints of their environment. Using this software, we shall investigate the consequences for frog group behaviour of varying individual decision models.
The same software can be used to simulate collections or swarms of robots exchanging messages. Using sound for communication, perhaps combined with lizard-ear hearing, would allow a robot in a swarm to measure the bearing of a caller quite precisely, but not its distance from the listener. This suggests the possibility of an interesting set of dual algorithms to those based on wireless communication, where distance is easy to measure but bearing is hard. We plan to investigate this possibility both using the simulation tool and with robot swarms.
Embodied Acoustic Scene Analysis (Bats)
Ultrasonic sensing is used in robotics today as a simple collision monitoring system, akin to car-parking aids that monitor whether there are obstacles the driver should attend to when reversing. Bats, on the other hand, have a sophisticated ultrasound-based echo sense which they use to catch prey and to navigate in cluttered and complex environments, often while flying at several metres per second.
Our goal in the Centre is to understand how bats use their echo sense and to bring that understanding to bear on robotic problems. For example, an ultrasonic sensor with a bat’s level of sophistication would complement visual sensing in many service robotic applications, and would also work in the dark, or in smoke-filled rooms. Following on from our EU project, ChiRoPing, it is now possibly to build hardware that allows sonar sensing in air with capabilities approaching those of bats; in our future work we plan to control a rapidly moving aerial drone using (only) such a sonar.
A second spin-off of local collaborative work on bats being pursued in the Centre is technology for long-term acoustic monitoring of natural environments. A deployment of one such system in a local forest for a 2-month period revealed the relatively frequent presence of bat species previously thought to be rare in the area, as well as a lot more social interaction between bats than biologists had expected. On the other hand, the system generated 240 Gigabytes of data from which the interesting "bat events" had to be extracted automatically. With solutions to the data processing problems, the monitoring system - which consists of standard commodity components and open source software - has many applications, ranging from environmental diversity studies to monitoring the health and happiness of animals in captivity.
Contact Information
Professor John Hallam, e-mail: john@mmmi.sdu.dk