Swarm Robotics: From sources of inspiration to domains of application Erol Sahin KOVAN Dept. of Computer Eng. Middle East Technical University Ankara, Turkey http://www.kovan.ceng.metu.edu.tr
What is Swarm Robotics? Yet another novel approach to the control of large group of robots! Study of multi-robot coordination strategies inspired from social insects. Engineering self-organization in physically embodied swarms. Application of Swarm Intelligence to the control of a group of robots. 2/29
Attempting to define the term Swarm Robotics? Need to identify aspects that make swarm robotics approach novel and desirable. aspects that distinguishes swarm robotics from other related studies. 3/29
What s novel and desirable in the Swarm Robotics approach? Emphasis on the system-level functioning properties observed in social insect systems: Robustness Flexibility Scalability Essential for deploying large numbers of robots. 4/29
Robustness Social insects can continue to operate despite large disturbances. Redundancy Decentralized coordination Simplicity of the individuals Distributed sensing 5/29
Flexibility Social insects can offer modularized solutions to tasks of different nature by utilizing different coordination mechanisms. 6/29
Flexibility same swarm, different tasks Foraging Prey retrieval Chain formation 7/29
Scalability Social insects are observed to be able to operate under a wide range of group sizes. That is, coordination mechanisms are rather independent of the number of individuals in the group. 8/29
Putting swarm robotics in place Where is Swarm Robotics placed in relation to other related studies? Aspects that distinguish swarm robotics studies from: other flavors of multi-robot studies other related studies such as Swarm Intelligence, Sensing networks, etc.. 9/29
0 - Individuals should be robots! Individuals should be autonomous robots. Individuals should be situated and autonomous be able to physically interact Mobility of individuals is sufficient, but not required. Metamorphic robotic systems? Yes Sensor networks? No 10/29
1 - Large number of robots The study should be relevant for the coordination of large numbers of robots. Why relevancy? How large is large? 11/29
2 - Few homogeneous groups of robots The robotic system should consist of few homogeneous groups and that the number of robots in each group should be large. Teams are not swarms. Hierarchical robotic systems (for instance swarms with a designated queen ) are less `swarm robotic. What s a homogeneous group? How about individual adaptation? 12/29
3 - Relatively incapable of inefficient robots The robotic system should utilize relatively incapable or inefficient robots with respect to the task at hand. The robots should have difficulties in carrying the task on their own. The deployment of a group of robots should improve the performance of system. The deployment of a group of robots should improve the robustness of the system. 13/29
4 - Robots with only local sensing and communication abilities For coordinating their actions, the robots should utilize only local sensing and communication capabilities. Locality promotes scalability. Existence of global communication channels not used for coordination among the robots does not violate. 14/29
Criteria for Swarm Robotic systems A swarm robotic system should consist of large numbers of robots, few homogeneous groups of robots, robots that are relatively incapable or inefficient, robots with only local sensing and communication abilities. Not a checklist for evaluating a study. But as yardsticks to evaluate how `swarm robotic a given study is. 15/29
Finally a definition Swarm robotics is the study of how large number of relatively simple physically embodied agents can be designed such that a desired collective behavior emerges from the local interactions among agents and between the agents and the environment. 16/29
Sources of inspiration Self-organizing natural systems Social insect systems: ants, termites, wasps, bees, cockroaches, locusts Animals with social behaviors: penguins, birds, fish, sheep... Unicellular organisms: Amoebae, bacteria, viruses Artificial self-organizing systems Amorphous computing Self-assembly of materials 17/29
Aggregation of amoebae into slime mould When food is abundant, amoebae (D. discoideum) acts independently of others, feeding and multiplying (Bonner;1967, Goldbeter;1996). When food supply is depleted amoebae release camp ( a chemo-attractant for amoeba) into the extracellular environment. Amoebae aggregate forming a slug, a multi-cellular organism which can move and sporulate. Summarized from Self-Organization in Biological Systems by Scott Camazine, Jean-Louis Deneubourg, Nigel R. Franks, James Sneyd, Guy Theraulaz, and Eric Bonabeau 18/29
Aggregation mechanism Amoebae secrete camp leading to spiral waves that streams the cells to the center Positive feedback mechanisms Release camp with a certain period (oscillatory mode). If hit by a camp pulse, amplify it (relay mode). Negative feedback mechanism High camp concentrations briefly desensitize the receptors. Amoebae moves in the direction of increasing camp concentration (at 1/10 of the camp wave speed). Cell-to-cell adhesion makes amoeba clumps persistent. 19/29
Take-home lesson/inspiration The mechanism discussed aggregate 10,000-100,000 cells! In a recent study (Dorigo et al.;2004), it is shown that aggregating tens of robots (equipped with simulated speakers and microphones) is very challenging. Stigmergy seems to be a key element for scalability! Stigmergy in a swarm of robots Natural stigmergy: Using water, chemicals, etc.. Stigmergy using embedded systems: e.g. Gnats (Balch;) Stigmergy using robots: Use some of the robots as the medium while others aggregate. 20/29
Quorum sensing in bacteria Bacteria seem to have interesting communication mechanisms to increase their survival. V. fischeri produces light when its population reach a critical size. V. cholarae delays the production of virulence factor until they reach a certain mass, to ensure a successful infection against the infection system. Recent studies show that bacteria use certain auto-inducers to detect their density in the environment. B.L. Bassler, How bacteria talk to each other: regulation of gene expression by quorum sensing. Current Opinions in Microbiology 1999 Dec;2(6): 582-7. 21/29
Take home lesson/inspiration Quorum sensing will be an essential problem for swarm robotic systems. The density of individuals is an important parameter in natural swarms which can lead to bifurcations in swarm behavior. Density measurement w/o stigmergy is likely to be an interesting challenge. 22/29
Information exchange in bacteria Bacterial colonies can be more resistant to antibiotics than bacteria living in suspension! Hypothesis: Bacteria form a genomic web communicating with each other: Inducive communication: a chemical signal triggers a certain action in other bacteria. Informative communication: the message received is interpreted by the cell, and its response is determined by its history as well as its current state. E. Ben-Jacob, Bacterial self-organization: co-enhancement of complexification and adaptability in a dynamic environment. Phil. Trans. R. Soc. Lond. A, 361, pp 1283 1312, 2003. 23/29
Take home lesson/inspiration In real life, some individuals of a swarm robotic systems will probably discover certain hazards the hard way. Individuals should be able to pass lastminute signals and information to the rest of the swarm. 24/29
Amorphous computing Challenge: How can prespecified, coherent behavior be engineered from the cooperation of vast numbers of unreliable parts interconnected in unknown and time-varying ways? Medium: a system of irregularly placed, asynchronous, locally interacting computing elements. Inspiration and approach: morphogenetic processes in biological systems such as tissue growth. Amorphous Computing, Abelson et al, Communications of the ACM, Volume 43, Number 5, May 2001. 25/29
Take home lesson/inspiration Amorphous computing nodes [if and when they become available] can be active intelligent pheromones of swarm robotic systems. Swarm robotic systems, when immobile, are amorphous computing mediums and can utilize their programming paradigms. 26/29
Self-assembly Self-assembly: self-organization by making physical bond formation Individuals lose some of their motility. This creates some interesting dynamics. Social insects and breakable bonds in chemistry Self-assembly of materials is described as the autonomous organization of components into patterns or structures without [external] intervention. Whitesides and Grzybowski (Science; 2002) Self-assembly is a promising method for fabricating regular structures: nano-scale self-assembly is promising for building large numbers of micro- electro-mechanical systems (MEMS), improving the robotic assembly processes. 27/29
Take home lesson/inspiration Use of templates for scaffolding the selfassembly/organization process to reduce defects in the structure. Catalytic agents to improve the selfassembly process. 28/29
Thanks for listening..