Understanding collective behaviour in nature to design complex robotic systems

ArticleDetailDownload PDF

Collective behaviour among living things draws the interest of many researchers from different fields for its potential links to the engineering of collective artificial systems. An interesting example of a collective system is demonstrated by Dictyostelium discoideum (D. discoideum), a social amoeba that can change its behaviour to survive in response to nutrient starvation. Dr Giovanna Di Marzo Serugendo and her collaborators from the University of Geneva, Switzerland, designed an artificial collective system made of individual robotic units known as Kilobots, drawing inspiration from their studies of the D. discoideum life cycle.

The study of collective behaviour attracts researchers from the disparate fields of biology, engineering and computer science, as it provides important opportunities to understand how complex interactions emerge from the comparatively simple behaviour of individuals. A particularly fascinating form of collective system is displayed by Dictyostelium discoideum (D. discoideum), a social amoeba that alters its behaviour to survive in direct response to nutrient starvation. For most of its life, each D. discoideum individual lives in the soil as a single amoeba, feeding on bacteria. Individual cells move around on their own when there is plenty of food. Then, when food is lacking, the cells start a strange process of multicellular development. Up to a million amoeboid cells will self-aggregate, using a complex pattern formation to build a coherent super-organism, similar in some senses to a slug.

The multidisciplinary focus on collective behaviour

The type of collective behaviour adopted by D. discoideum is known as a ‘first-order emergent behaviour’ and allows the complex super-organism to acquire several properties that none of the cells could possess on their own. The ‘slug’ then moves as a collective, having acquired what is known as ‘second-order emergent behaviour’, as it looks for a suitable place for approximately 20% of its cells to die, in order to lift the remaining cells up for sporulation and dispersal in the soil for the purposes of reproduction. Interestingly, at this point the cells resume their individual behaviour.

From D. discoideum life cycle to behaviour modelling and swarm robotics.

Dr Giovanna Di Marzo Serugendo and her team from the University of Geneva, Switzerland, investigate this second-order emergent behaviour, which arises from the interactions of individuals that are themselves the result of first-order emergent societies. More specifically, within second-order emergent systems, agents recognise the existence of groups that emerged from their own collective behaviours. This relatively simple but fascinatingly powerful behaviour has inspired the research team to develop and design robotic models that can display collective behaviour. In their designs, a large number of simple homogeneous agents coordinate, self-organise, and adapt themselves in response to environmental changes.

“Collective behaviour allows complex cellular organisms to have several properties that none of the individual cells would have on their own.”

Modelling the factors leading to higher-order emergent behaviour

During the transition from growth to development, the ability of D. discoideum cells to switch from a unicellular mode of life to a multicellular entity makes it an ideal organism with which to analyse social behaviour. Collective behaviour – such as that which can be observed in ants, bees, fish, as well as during the aggregation phase of D. discoideum’s lifecycle – represents first-order emergence: a set of properties that arise from the collective behaviour of individual entities. In their studies, Di Marzo Serugendo and her colleagues are particularly interested in the transition between first- and second-order emergent behaviour, which occurs when individual entities organise themselves into ‘super-organisms’, which themselves will display swarm-like behaviour. The group investigates the social behaviour of Dictyostelium discoideum to develop collective adaptive artificial systems.

D. discoideum can be described as a ‘social amoeba’. This peculiar eukaryotic organism (an organism with a clearly defined nucleus) feeds on bacteria that can be found in the top few centimetres of soil. In the vegetative phase, individual amoeba cells move around alone, grazing for food. Once the food supplies are scarce, the individual cells produce and respond to cyclic adenosine mono phosphate (cAMP), which not only triggers the migration of individual cells to form the ‘slug’ superstructure, but also regulates the differentiation of individual cells into specialised cells so that a specific function can be fulfilled. For example, pre-stalk cells will differentiate to make the stalk, basal disk, and fruiting body, while pre-spore cells will transform into spores. These processes are all examples of first-order emergent behaviour.

Di Marzo Serugendo and her collaborators programmed robotic units called Kilobots.

Second-order behaviour

In 2020, the research team published a study describing the second-order behaviour of D. discoideum and how it can serve as a model that can be transposed into the field of robotics. In particular, the article focused on the transition of D. discoideum from first-order emergent behaviour to a second-order stage. The authors describe how, after aggregating, the amoeba cells make a coherent organisation, which is enclosed by a slime sheath and is similar to a slug without organs. These slug-like superorganisms can form chains, merging and moving away from each other; a perfect example of second-order emergence. After about 24 hours, each super-organism transforms into a new organisation: the fruiting body, consisting of spores on top of a stalk. Eventually, the spores germinate, releasing new amoeba cells that resume their individual behaviour.

In their 2020 publication, Di Marzo Serugendo and her colleagues affirm that the different models for the migration phase of D. discoideum ‘vary from mathematical models to off-lattice models involving motive force, and adhesion’. The authors describe the ‘off-lattice’ computational model as a system where cells communicate via surface molecules, reacting to different signals and a combination of internal and external forces. They explain that the model should aim to represent cell–cell adhesion and cell-signalling within a network of neighbouring cells and chemical signals. The team used agent-based computational tools to model the transition phase from first- to second-order emergence, as well as two second-order collective behaviours: the slugs merging or avoiding each other, and their collective movement toward the light. The model takes into account the concentration of chemical signals, such as ammonia, and the intensity of light using two-dimensional diffusion functions. According to the model, the length of the slugs will result in different migrating velocities. Interestingly, the team’s computational modelling results mirrored their biological experiments.

Biological illustration: D. discoideum streaming (aggregation phase).

Swarm robotics: Current and future developments

Collective behaviour in nature provided the team with a vital source of inspiration for engineering artificial systems involving robotics. The field of swarm robotics aims to design systems where a large number of simple robots coordinate themselves, self-organise, and adapt to changing or complex environments.

“Di Marzo Serugendo and her collaborators designed an artificial collective system made of individual robotic units known as Kilobots.”

In their 2020 publication, Di Marzo Serugendo and her colleagues refer to several examples of swarm robotic applications that have already been designed by other researchers around the world. Some complex systems present a biologically-inspired hierarchical approach, implementing an artificial leadership model, where individual robots react to certain signals and ‘follow’ other robots that are higher up in the hierarchy. While most swarm robotics have a simple set-up, where the robots are of the same type, heterogeneous swarm robotic systems are made of many robots of different types.

For example, a particularly intriguing and complex system was developed to mimic the ‘stimulus and response’ action of the nervous system. In this type of design, referred to by Di Marzo Serugendo and colleagues, the robots physically dock, sharing a joint architecture for sensing the environment and for simulating a decision-making process. This system could be referred to as a robotic nervous system, where one robot acts as a brain unit to make the decisions for the collective.

Biological illustration: D. discoideum – moving slugs (migration phase).

Di Marzo Serugendo and her collaborators designed an artificial collective system made of individual robotic units known as Kilobots, drawing inspiration from their study of the D. discoideum life cycle. To programme Kilobots they used infrared transmitters, which also allowed them to programme all the units collectively, at the same time. Communications were transmitted by pulsing messages in a range of 10cm by infrared LED light, which enables the robots to get information uniformly from all directions.

The robots were able to synchronise themselves in a swarm and to select a leader, which will also serve as the centre of aggregation. The robots self-organised into ‘slugs’ displaying behaviour such as avoiding each other, or collectively moving towards a source of light. In future experiments, the team is planning to scale up the investigation by employing a larger number of Kilobots, and will importantly look to reproduce in the robots the transition phase from first- to second-order emergent behaviour.

The team’s research is importantly yielding a nuanced understanding of D. discoideum individual cell behaviours at each phase of its life cycle, through biological experiments. It is also finessing the translation of first- and second-order emergent behaviour into swarm robotics, further highlighting the fruitful interrelation of natural collective behaviours and artificial robotic imitation.


How will swarm robotics advance technology? Do you have any specific applications in mind?

Swarm robotics is particularly useful when the robots are on their own in uncertain or unknown environments. They need to collectively adapt to or overcome the ongoing situation, such as in disaster scenarios (eg, earthquakes) or for space exploration.

 

References

  • Parhizkar, M, Di Marzo Serugendo, G, Nitschke, J, et al, (2020) First-order agent-based models of emergent behaviour of Dictyostelium discoideum and their inspiration for swarm robotics. Artificial Life Robotics 25, 643–655.
  • Parhizkar, M, Di Marzo Serugendo, G, Nitschke, J, et al, (2020) Second-order agent-based models of emergent behaviour of Dictyostelium discoideum and their inspiration for swarm robotics. Artificial Life Robotics 25, 656–665.
DOI
10.26904/RF-139-2126282534

Research Objectives

Dr Giovanna Di Marzo Serugendo and her team designed an artificial collective system made of individual robotic units, based on the collective behaviour of Dictyostelium discoideum.

Funding

This project was partly funded by Swiss National Science Foundation SNSF fund.

NB: 205321_179023 – Dicty – Social Amoeba Dictyostelium discoideum as an inspiration for Higher-Order Emergence in Collective Adaptive Systems.

Collaborators

  • Professor Thierry Soldati, University of Geneva
  • Professor Salima Hassas, Université Claude Bernard Lyon 1
  • Dr Mohammad Parhizkar
  • Mr Jahn Nitschke, University of Geneva
  • Dr Assane Wade, University of Geneva
  • Mr Louis Hellequin

Bio

Giovanna Di Marzo Serugendo holds a PhD in software engineering from EPFL. She is full professor and Director of the Computer Science Center of the University of Geneva. Her research relates to the engineering of decentralised software with emergent behaviour. Giovanna co-founded the ACM Transactions on Autonomous Adaptive Systems (TAAS).

Dr Giovanna Di Marzo Serugendo

Contact
Centre Universitaire d’Informatique
Battelle, Bâtiment A
7, Route de Drize
CH-1227 Carouge (GE)
Switzerland

E: giovanna.dimarzo@unige.ch
T: +41 22 379 00 72

Related posts.

Comments.

Leave a Comment

Your email address will not be published. Required fields are marked *

Share this article.

Share on facebook
Share on twitter
Share on whatsapp
Share on linkedin
Share on reddit
Share on email