Differences

This shows you the differences between two versions of the page.

 — literature [2014/08/20 18:37] (current)tlund1 created 2014/08/20 18:37 tlund1 created 2014/08/20 18:37 tlund1 created Line 1: Line 1: + == Our Work from BYU == + * [http://​faculty.cs.byu.edu/​~mike/​mikeg/​papers/​CoherentTopologyProperties_v3.pdf The Technical Report] for 2010-2011 from BYU. + + == Self-Organization in Flocks and Swarms == + + * [http://​webscript.princeton.edu/​~icouzin/​website/​wp-content/​plugins/​bib2html/​data/​papers/​couzin02.pdf Couzin'​s paper] where he talks about how changing parameters leads to phase transitions (e.g., random to torus phase) in fish schooling. ​ This uses the classic three metric zones from the Boids model: repulsion, attraction, and orientation. + * [http://​ieeexplore.ieee.org/​xpls/​abs_all.jsp?​arnumber=1605401&​tag=1 Olfati-Saber'​s great paper on the theory of flocking.] ​ He defines a flock as a structured topology (an isotropic lattice) that uses only local connections. ​ Under some fairly reasonable assumptions,​ he shows that if the members of the flock always know where a leader agent is then the flock will (a) arrange itself into an isotropic lattice and (b) follow the leader when the environment has no obstacles. ​ He also shows how repelling agents can be used to push agents away from (and cause to go tangent to) obstacles in the environment. ​ Email Mike if you'd like a copy of his annotated version of the paper. + * [http://​www.ncbi.nlm.nih.gov/​pmc/​articles/​PMC2169279/​ Evolving the selfish herd] extend'​s Couzin'​s model to allow agents to adapt their individual parameters in the presence of either (a) predators or (b) food sources. ​ Under the presence of a predator, two apparently stable flocking behaviors result: a slow mill and a fast flock. ​ In terms of expressiveness,​ these two appear to be different ways in which a predator can influence what the flock does.  The fast flock could be used in HuBIRT to cause a group to fan-out as in a mine-sweeping task.  Under foraging, clumped food groups are compatible with flocks but isolated food packets promote non-flocking behavior. ​ Note that this paper identifies four different phases of possible collective behavior as a function of model parameters. ​ This is one more than Couzin identified. ​  Email Mike if you'd like a copy of his annotated copy of this paper.  ​ + * [http://​pre.aps.org/​abstract/​PRE/​v63/​i1/​e017101 Self-organization in self-propelled particles] is a physics-based version of swarming work.  Interestingly,​ their model is also additive and includes attraction, orientation,​ and repulsion. ​ They refine these ideas a bit, and provide some explanation for when various phases emerge and when these phases are stable. + * This paper [http://​www.idsia.ch/​~frederick/​taskallocation.pdf on task allocation] discusses the use of  attraction/​repulsion principles based on light sensors for team selection. Different type of lights (green, yellow) indicate different kinds of behaviors (attraction,​ repulsion, escape). Further, they study the performance achieved using attraction/​repulsion based task allocation vs broadcast based task allocation. ​ + * [http://​www.csim.scu.edu.tw/​~chiang/​course/​ComputerGameAdvance/​Collective%20Memory%20and%20Spatial%20Sorting%20in%20Animal%20Groups.pdf Modeling schooling] in fish, and getting a nice tutorial on the math of collective intelligence. + * This is a great [http://​faculty.cs.byu.edu/​~mike/​mikeg/​papers/​MOSC/​Sumpter06.pdf tutorial on common themes in self-organizing] or bio-inspired systems. ​ I've added some annotations for how humans could interact with these systems. These may be relevant when we start to apply agent-based models to the kinds of systems being built in the ONR project. + + == Moving Beyond Swarms/​Flocks to Colonies and Packs. == + + * [http://​ieeexplore.ieee.org/​xpls/​abs_all.jsp?​arnumber=5723472&​tag=1 Multi-robot system based on model of wolf hunting behavior to emulate wolf and elk interactions] By Jadden, Arkin, and MacNulty appeared in ROBIO 2010.  It presents a nice model of wolf collaborative hunting behavior, identifying phases of group behavior and suggesting some individual rules by which emergent group behavior can emerge. ​ The paper is a bit sparse on results. ​ Email Mike if you'd like a copy of his annotated version. + * [http://​www.public.iastate.edu/​~kmoloney/​Instructor/​ELVISpapers/​consensus.pdf Consensus decision making in animals] is a nice framework paper that describes how animal and human groups can include flocking behavior but also extend to more sophisticated behavior. ​ They use a compelling example for how honeybees choose a new site for their hive to motivate consensus problems, and this feels to me like it opens the door for HuBIRT that works beyond flocks to more of a multi-tasking,​ quorum-signaling problem domain. ​ Note that paper gives a simple decision tree based on global/​local communication and presence/​absence of a conflict of interest; the tree produces examples from nature and identifies mechanisms for producing consensus decision-making. ​ Email Mike if you'd like a copy of his annotated version. + * [http://​rstb.royalsocietypublishing.org/​content/​364/​1518/​719.full.pdf Group decisions in humans and animals: A survey] provides a useful distinction between collective and interactive decision-making. ​ The paper also provides some useful aggregation rules for bringing together information from various members of a collective, and makes a distinction between shared, partially shared, and unshared decision processes in humans and animals. ​ Furthermore,​ the paper notes differences between human and animal rationality,​ and identifies some noteworthy theoretical concepts (e.g., Condorcet'​s jury theorem) that are applicable to animal decision-making. + + + == The Importance and Role of Variability == + * [http://​www.cs.ucf.edu/​~ecl/​papers/​1004.jaamas.pdf Multi-agent role allocation: issues, approaches, and multiple perspectives]. From Annie Wu's lab. + * [http://​www.cs.ucf.edu/​~ecl/​papers/​1110.saso.cortney.pdf On the impact of variation on self-organizing systems]. From Annie Wu's lab. + * [http://​www.cs.ucf.edu/​~ecl/​papers/​1110.collaboratecom.ramya.pdf On the relationship between response probability and redundancy in teams of collaborating agents]. ​ From Annie Wu's lab. + * [http://​www.cs.ucf.edu/​~ecl/​papers/​1110.saso.chris.pdf Using the process of norm emergence to model consensus formation]. ​ From Annie Wu's lab. + + + == Nearest Neighbor Topologies Versus Metric Topologies == + + * [http://​faculty.cs.byu.edu/​~mike/​mikeg/​papers/​MOSC/​PNAS-2008-Ballerini-1232-7_annotated.pdf Mike's annotated copy of Ballerini'​s classic paper] on how topology-based interactions between agents promote group cohesiveness. ​ In this work, Ballerini et al. measure how flocks of starlings interact and find strong evidence that their local interactions are determined by their closest 6-7 neighbors rather than by all neighbors within some kind of metric distance. + * [http://​rsif.royalsocietypublishing.org/​content/​8/​55/​301.full Limited Interactions in Flocks] replicates Ballerini'​s results on topological structure in simulation. ​ They use a probabilistic approach wherein interaction neighbors are chosen with a probability inversely proportional to their distances from the agent. ​ Once a neighbor is chosen, it influences the agent'​s behavior according to the rules associated with Couzin'​s three metric zones. ​ Send Mike an email for a copy of his annotated version of this paper. + + + == Fielded Systems == + * [http://​en.wikipedia.org/​wiki/​Multi_Autonomous_Ground-robotic_International_Challenge The MAGIC Competition] + * [http://​www.engin.umich.edu/​newscenter/​feature/​robotics/​ Michigan'​s entry into MAGIC competition] + * [http://​www.upenn.edu/​almanac/​volumes/​v57/​n13/​magic.html UPenn'​s entry into MAGIC competition] + * [http://​en.wikipedia.org/​wiki/​Heterogeneous_Aerial_Reconnaissance_Team DARPA HART Program] + * [http://​www.wpafb.af.mil/​shared/​media/​document/​AFD-070418-027.pdf AFRL has a control station called ''​Vigilant Spirit''​] + + == Leading according to need==  ​ + Email Mike if you'd like to see his annotated copy of either of the following papers. + + * [http://​www.sciencedirect.com/​science?​_ob=MImg&​_imagekey=B6VRT-4W5DPDK-F-1&​_cdi=6243&​_user=456938&​_pii=S0960982209007921&​_origin=gateway&​_coverDate=04%2F28%2F2009&​_sk=999809991&​view=c&​wchp=dGLbVzz-zSkzV&​md5=169c9333b3ec1a34ec4d0fa10c972f60&​ie=/​sdarticle.pdf David Sumpter recently] called for work that combines mechanistic and functional explanations for animal interaction. ​ He says that Mechanistic explanations look at how animals interact to produce group level patterns"​ and Functional explanations are based on arguments about why a behaviour has evolved through natural selection." ​ A mechanistic explanation might be in the form of a set of rules that generate collective intelligence,​ such as the rules encoded in Couzin'​s model above, and a functional explanation might be in the form of the evolutionary forces that produce such behavior such as flocking behavior that evolves to avoid predation. ​ He cites the excellent paper by Conradt et al. + * The paper by [http://​www.jstor.org/​stable/​20491511 Conradt et al. combines a functional] description (how hungry a subset of fish are) with a mechanistic description (how assertive or speedy a fish is) for how small subgroups of fish can lead a larger school to an objective that differs from the goal of the majority. + * [http://​rstb.royalsocietypublishing.org/​content/​364/​1518/​807.full Evolution of decision sharing], combining game theory, self organizing systems, and evolutionary stable strategies. ​ Contributed by Brian Pendleton. + + + == HuBIRT is also known as ''​human-swarm interaction''​ and ''​assistive swarming''​. == + There are a handful of papers under this name in the literature. + * [http://​www.dtic.mil/​cgi-bin/​GetTRDoc?​Location=U2&​doc=GetTRDoc.pdf&​AD=ADA422540 Doug Gage'​s] work on treating a swarm as a solid, liquid, or gas. + * [http://​faculty.cs.byu.edu/​~mike/​mikeg/​papers/​MOSC/​LewisIros11swarm.pdf A cool exploratory study] from Mike Lewis' lab that looks at how humans can use various fundamental swarm behaviors and swarm topologies to perform a set of canonical tasks  This technical report is posted with Mike's permission. + * [http://​scholarsmine.mst.edu/​post_prints/​pdf/​Bashyal_09007dcc80600f1e.pdf Bashyal and Venayagamoorthy] presented an HSI approach that provided a human with a partial plan and global information,​ and then allowed the human to adjust the autonomy of a small subset of swarm members to influence swarm behavior. ​ They included a small user study that demonstrated that the HSI team performed better than the swarm alone. ​ An important theme of their paper, in addition to the user study and the ad hoc set of desirable HSI features, is that The ideal man-machine interaction is ... one that functions autonomously while providing users with a method to inject knowledge and guidance so as to improve [system performance]." ​ The take this design philosophy to an important extreme where a human'​s control over the swarm is only as much as that of any other member in the swarm,"​ meaning that the human may be able to control a single individual agent and thereby influence swarm behavior, but not perform any centralized control or exert any global influence on the entire swarm. ​ Another interesting take away from this paper is that the authors use particle swarm optimization as the basis for enforcing efficient foraging for a gradient-based radiation source. ​ [http://​faculty.cs.byu.edu/​~mike/​mikeg/​papers/​MOSC/​Bashyal_09007dcc80600f1e.pdf Mike's annotated copy]. + *[http://​citeseerx.ist.psu.edu/​viewdoc/​download?​doi=10.1.1.95.7486&​rep=rep1&​type=pdf Introduces the term] ''​assistive swarming''​. ​ I don't know the real reference for this and haven'​t read it carefully, but it seems like the name should be mentioned in literature reviews. + * [http://​ieeexplore.ieee.org/​xpls/​abs_all.jsp?​arnumber=4745884&​tag=1 The GUARDIANS project] seeks to use swarm robotic technology to support firefighters. ​ These two papers present both the use-cases for HSI as well as an artificial potential field-based implementation. ​ The robots include the ability to autonomously respond to obstacles and to follow either a proximate human or a remotely controlled virtual avatar through an environment. ​  This ability to proximate track the robot may be unique in the literature. ​ Although these papers claim that a stability analysis has been performed, I think it is more accurate to say that some simulations have been performed that subjectively establish that the HSI is successful.  ​ + * Work by MaryAnne Fields on controlling swarms, so-called "​Soldier-Robotic Swarm Interaction"​. + ** [http://​schmidtcds.com/​uav/​wp-content/​uploads/​2010/​01/​Swarm-formation-control-utilizing-ground-and-aerial-unmanned-systems.pdf Abstract on UAV-UGV swarm formation control]. + ** [http://​med.ee.nd.edu/​MED13/​papers/​T33-004-301.pdf Unmanned Ground Vehicle Swarm Formation Control Using Potential Fields]. + ** [http://​www.dtic.mil/​cgi-bin/​GetTRDoc?​Location=U2&​doc=GetTRDoc.pdf&​AD=ADA523930 Technical report on Soldier-Robotic Swarm Interaction]. + * [http://​www.springerlink.com/​content/​9m71g0028322v217/​ A Generalized Graph-Based Method for Engineering Swarm Solutions to Multiagent Problems] by Wiegand, et al., explores adapting a type of potential field-like agent controller in a multi-objective problem. ​ Use graph-based method for designing behaviors, suggesting that an organizational approach (e.g., graph topology) may be instrumental in design collective behaviors. + *[https://​facwiki.cs.byu.edu/​HCMI/​index.php/​File:​KiraPotterICARA09.pdf Exerting Human Control Over Decentralized Robot Swarms]. Zsolt Kira and Mitchell A. Potter. ​ 2009. + * A copy of the [http://​faculty.cs.byu.edu/​~mike/​mikeg/​papers/​MOSC/​SMC2010.pdf paper that we submitted to SMC2010]. ​ It includes Jon Whetten'​s work as well as a formalized version of Yisong'​s work.  ​ + * [https://​facwiki.cs.byu.edu/​HCMI/​index.php/​File:​KiraPotterICARA09.pdf Exerting Human Control Over Decentralized Robot Swarms]. Zsolt Kira and Mitchell A. Potter. ​ 2009. + + == Physicomimetics == + + * [http://​www.idsia.ch/​~frederick/​taskallocation.pdf An implementation of physicomimetics.] ​ Sujit says "Its a nice way of implementing the physiomimetics with light sensors for the follower robots. Yellow light to attract while green light to repel. This is another form of no-broadcast communication."​ + + + == Other Examples from Biology or Human Behavior == + + * This is a great [http://​faculty.cs.byu.edu/​~mike/​mikeg/​papers/​MOSC/​Sumpter06.pdf tutorial on common themes in self-organizing] or bio-inspired systems. ​ I've added some annotations for how humans could interact with these systems. These may be relevant when we start to apply agent-based models to the kinds of systems being built in the ONR project. ​ + * [http://​www.sciencemag.org/​cgi/​content/​full/​329/​5996/​1194 Science article] on how network structure influences how behavior diffuses through a social network. ​ Mike's annotated version is [http://​faculty.cs.byu.edu/​~mike/​mikeg/​papers/​SpreadofBehaviorinSocialNetworkExperimentScience.pdf available here]. + * [http://​www.nature.com/​nature/​journal/​v442/​n7106/​abs/​nature05014.html Cell proliferation] affects the way that tissues are organized, including strong topological properties. + + == Concepts from Organizational Behavior, Physics, and Engineering == + + * [http://​www.springerlink.com/​content/​9m71g0028322v217/​ A Generalized Graph-Based Method for Engineering Swarm Solutions to Multiagent Problems] by Wiegand, et al., explores adapting a type of potential field-like agent controller in a multi-objective problem. ​ Use graph-based method for designing behaviors, suggesting that an organizational approach (e.g., graph topology) may be instrumental in design collective behaviors + * [http://​faculty.cs.byu.edu/​~mike/​mikeg/​papers/​MOSC/​CollectiveIntelligence.pdf Evidence for a Collective Intelligence Factor in the Performance of Human Groups]. ​ Is there something called collective intelligence,​ can it be measured, and what makes one group more collectively intelligent than another? ​ Mike's annotated copy. + * [http://​faculty.cs.byu.edu/​~mike/​mikeg/​papers/​LewisSMC2010.pdf Teams Organization and Performance in Multi-Human/​Multi-Robot Teams]. ​ Mike Lewis et al. discuss how teams of people can self-organize in a shared pool of operators, and how the way that they self-organize affects performance especially for teams that share authority. ​ This feels a lot like the effect that Jon Whetten found for teams that specialize perform better. ​ The copy here is Mike G's annotated copy. + * [http://​www.sciencemag.org/​cgi/​content/​full/​329/​5996/​1194 Science article] on how network structure influences how behavior diffuses through a social network. ​ Mike's annotated version is [http://​faculty.cs.byu.edu/​~mike/​mikeg/​papers/​SpreadofBehaviorinSocialNetworkExperimentScience.pdf available here]. + ** [https://​facwiki.cs.byu.edu/​HCMI/​images/​8/​89/​KnudsenLevinthal07.pdf Two Faces of Search: Alternative Generation and Alternative Evaluation] by Knudsen and Levinthal. ​ Keywords: organizational search, bounded rationality,​ organizational decision-making + * [https://​facwiki.cs.byu.edu/​HCMI/​images/​8/​89/​KnudsenLevinthal07.pdf Two Faces of Search: Alternative Generation and Alternative Evaluation] by Knudsen and Levinthal. ​ Keywords: organizational search, bounded rationality,​ organizational decision-making + * A classic paper on [http://​faculty.cs.byu.edu/​~mike/​mikeg/​papers/​MOSC/​QueuingTheoryFlightManagement.pdf ​ using queuing theory] to model pilot decision-making. ​ This paper introduces a lot of the time-scheduling models used in our work. + * An excellent paper defining the concept of a [http://​faculty.cs.byu.edu/​~mike/​mikeg/​papers/​MOSC/​SharedMentalModels.pdf shared mental model]. ​ This includes both models for how the group work as well as models for what is happening within the situation. ​ The experiment results isn't that interesting,​ but the paper provides a potential way to formalize the problem, either as an optimization problem, and agent-based simulation, or as a mechanism design problem. + * A copy of the [http://​faculty.cs.byu.edu/​~mike/​mikeg/​papers/​MOSC/​SMC2010.pdf paper that we submitted to SMC2010]. ​ It includes Jon Whetten'​s work as well as a formalized version of Yisong'​s work.  ​ + *[https://​facwiki.cs.byu.edu/​HCMI/​index.php/​File:​Tucker-mag.pdf Executive Decision Support]. Xu Chu Ding, Matthew Powers, Magnus Egerstedt, Shih-Yih (Ryan) Young, and Tucker Balch. ​ 2009.  [http://​faculty.cs.byu.edu/​~mike/​mikeg/​papers/​SpreadofBehaviorinSocialNetworkExperimentScience.pdf Mike's annotated version is available here.] + + == Human Factors == + + * [http://​www.ise.ncsu.edu/​nsf_itr/​794B/​papers/​Bainbridge_1983_Automatica.pdf Ironies of Automation] by Lisanne Bainbridge. ​ Discusses human factors in the context of automated systems and points out some ironies of increased automation. ​ For example, fully automated systems may require much more operator training than semi-automated systems due to a lack of hands on experience. ​ + * [http://​www.jstor.org/​stable/​1416145 A Theoretical Field-Analysis of Automobile-Driving] by James J. Gibson and Laurence E. Crooks. A classic paper 1937 paper detailing human factors involved in automobile driving. + + == Miscellaneous == + + * [http://​www.math-info.univ-paris5.fr/​~bouzy/​publications/​bouzy-metivier-icml10.pdf A comparison of MAL algorithms.] ​ Jake Crandall'​s M-Qubed algorithm does very well in this comparsion.