Animal‌ and‌ robot‌ Societies‌ ‌Self-organise‌
and‌ Integrate‌ by‌ Social‌ Interaction‌ (bees‌ and‌ fish)‌

ASSISI_bf presented at Living Machines in London, 2013


Two PIs in the ASSISI_bf project organized a workshop on bio-hybrid societies and bio-hybrid systems at the Living Machines conference in London, 2nd of August, 2013. At this workshop titled “Emergent social behaviours in bio-hybrid systems” was held at the Imperial College London. It was co-organized by Stuart P. Wilson, José Halloy and Thomas Schmickl. A distinguished selection of key scientists in the fields of bio-hybrid systems and emergence in social systems showcased about their approaches towards bio-hybrid systems to the community and discussed implications of their work and their future work plans with the audience. The group focussed on bio-hybrid systems, which are systems that consist of a mechanical (robotic) part and a natural (organic) part. This holds for products from synthetic biology, electro-mechanic implants in organisms, organic implants in machines to mixed societies consisting of animals and robots. In addition to that, the scope of the workshop stretched also to bio-inspired and bio-mimetic swarm robotics and artificial evolution systems.

- Stuart P. Wilson (University of Sheffield) showed his work on self-organized “huddling” behaviour of rat pubs, which aggregate densely in cold temperatures but spread in the nest at warmer environmental conditions. Even in the dense aggregate, they tend to move periodically from inside to the outside and vice versa.

- Terence W. Deacon (University of California, Berkeley) referred to the necessity of developing a novel information theory that takes into account the significance of the message. He proposed a novel concept of “referential information” by which the receiver alters the work done by the emitter after the information traversed the channel. To illustrate this, he showcased a molecular model system in which specific molecules turned out to be such a “reference”, meaning that they can act as a signal or indicator for something in the environment. In some sense, they get a “meaning” within the system. He argued that it is possible to connect information and thermodynamical work. According to T. Deacon, informational significance can be measured in terms of work saved in achieving some functional end with respect to work required in the absence of that information. If such theory works it will make the link between information, thermodynamics and significance i.e. a novel information theory that goes beyond Shannon’s theory of signals.
- Tim Landgraf (Free University Berlin) replaced the invited speaker Andrew Adamatzky (University of the West of England), who couldn’t travel to London in time. Tim presented his work of a dancing robot that could perform the waggle-dance language of honeybees. Following previous work on dancing bee-robots by Esch (1950s) and Michelsen (1990s), a modern variant of the robot, which was, for example, computerized to allow a vision system to prevent collisions. The robot could also feed sugar water to the bees. In addition, image-analysis, tracking of dancing bees and computational motion models helped to tailor the dancing robot to mimic the bees’ behaviours, whereby the robot was implemented by re-using a plotting device. The scientists used the carriage of the plotter to move the dancing robot. Tim showed that the acceptance of his robot by bees is actually pretty high, resembling the natural statistical distribution of dance followers concerning the number of dance rounds they did follow.
- Chrisantha Fernando (Queen Mary University of London) talked about the algorithmic basis of values. First, he gave an overview about the concept of the “Darwinian brain”. This concept assumes the existence of self-replicating structures in the brain, namely self-copying connection-patterns among sets of neurons. Thus, its is not material that replicates, it is connectivity patterns that replicate. The concept of the Darwinian brain assumes now that selection, mutation and inheritance might generate a process that resembles the evolution of natural organisms in Darwinian evolution. To investigate the potential of such a process, the scientists developed a system that combines genetic programing principles with a library of pre-formulated “atoms” that contain higher-level arithmetic operations/functions but also functionality like learning or search. Their system is evolving solutions by playing “games” and the talk was discussing what makes a game a good game, that is one that yields interesting or valuable solutions.

- José Halloy (Université Paris Diderot, Paris-7) gave an excellent overview of animal-robot interaction, ranging from the original experiments of N. Tinbergen (1950s) with sticklebacks to modern bio-hybrid social systems built by autonomous robots and animal societies. Jose suggested 3 methods to introduce robots into animal societies: (1) Have a mobile robotic swarm, (2) have an array of static devices, or (3) mounting devices on the social animals themselves. A first task is to replace one natural member of the society by an autonomous device that is accepted by the society, so that the behaviour of the total group does not change. He reported from an experiment in which cockroaches and small autonomous robots merged into one bio-hybrid society and performed collective shelter choice. Re-programming the robots could influence this decision-making, thus showing the possibility of outside control on such mixed societies. It also allowed novel insights into the collective decision making process of cockroaches, e.g. the emergent bifurcation in collective shelter choice when varying the size of shelters. In addition to cockroaches, Jose and his colleagues researched animal-machine interaction in chicken, focussing on the filial imprinting of the freshly hatched chicks to the robot. By using video tracking techniques, the autonomous robot (acting as a mother) was used to trigger several behaviours, like following, avoiding, turning or similar. Computer algorithms produced machine-made time-event plots and time budgets from the observed data, thus a novel method for ethological research was prototyped by using an autonomous robot. Finally, Jose raised the question whether or not we are facing the dawn of the age of “ethomics”, which is automatized derivation of behavioural models of animals to test genetic mutants and making the link with genomics and neuroscience.
- Roderich Gross (University of Sheffield) presented his studies on robotic swarm system. He highlighted his quest for a minimalistic swarm system, involving a very simple robot with minimal sensor input and minimalistic actuation to solve complex task (like the “rendezvous problem”). His work is based on the e-puck robot, which was artificially restricted to a one-pixel vision upwards and a one-pixel vision forward. Also motion was restricted to two specific motion patterns: straight motion and a pre-defined backwards curve. He showed that his beautifully simple swarm algorithms allow the robot swarm to perform some interesting collective motion tasks: Aggregating in one cluster and forming a rotating circle of moving robots. After showing the swarm research in his lab, Roderich focussed on artificial evolution of passively moved self-replicating building blocks. Artifacts of certain shapes are placed on air-hockey tables and floating around, basically randomly propelled by the airflows. They can hook into each other and form higher-level shapes. Modules can fulfil several tasks, like energy harvesting, energy sharing, reacting to stimuli, replication. This way, it is expected that evolutionary processes kick into action through emergence, so that “species” of such multicellular robot organisms appear. Finally, he presents a concept of a (yet simulated) machine that can autonomously perform science around animal behaviour. It can automatically produce hypothesis, setup environments, make observations and generate a behavioural model. As Jose before did, Roderich points to the well-known Turing test, this time performed with an animal and a machine (robot).


- Thomas Schmickl (Karl-Franzens University Graz) presented the historical pathway to the ASSISI_bf project by highlighting the key results of the biological experiments on honeybees: Groups of young bees are capable of performing collective decision making concerning their aggregation place in a temperature field to a much higher extent as single bees can do. They can pick out the best of all offered option from several sub-optimal options with some simple, yet smart and robust behavioural interaction patterns. Such abilities are often referred to as “swarm intelligence”. He showed that simple mechanical devices can interfere with this collective decision making, driving the swarm of bees to other, yet novel, solutions. After a brief overview of the project ASSISI_bf, which operates with fish and bees, showing also videos from the first experimental setups with both animals species, he turned to another question: Why would we create bio-hybrid social cyborgs and what service do these systems have to our society. He pointed out various benefits ranging from pest control to agriculture, but finally presented our modern, social-media affected society, to be a social cyborg in itself. Thus, he concluded, having model systems in the lab for studying such mixed-society cyborg consisting of robots, algorithms and social organisms is crucial to understand our own human society.
At the end of the workshop, an open discussion on bio-hybrid systems yielded several interesting questions:

The question was raised whether or not modern human society can be seen as a social cyborg. While the present group of expert scientists could not agree on the question whether social internet services like facebook, google or twitter are just a modern version of a hammer, which is mainly an “enforcement” tool, they agreed on several other aspects: such automatized services are network changers, that change the topology of the social network structure and thus can have significant impact onto society through feedback (loops) that are only partially understood. Also other autonomous services like “smart” traffic control and “smart grid”, as well as automatized stock exchange have societal consequences concerning stability and flexibility. Internet services and databases also have an impact on the time scales on which some social processes operate, thus they did also transform society into a new, yet only partially understood form. In consequence, some members of the workshop pointed out that social sciences and psychology should continue to study these phenomena of social cyborgs, but that there should also be simple model systems of group living animals interacting with machines to study simpler social cyborgs deeper. Finally, the group agreed that mixed societies, animal-machine interaction and social cyborgs constitute a new research field that should grow into a bigger research community by organizing similar workshops in future.

Besides the workshop, two posters were presented at the Living Machines conference main track (29th of July to 2nd of August 2013), shown within the robot exhibition event at the London Science Museum on the 1st of August 2013.