Local Interactions Of Agents Research Articles

Scalable Online Planning for Multi-Agent MDPs

We present a scalable tree search planning algorithm for large multi-agent sequential decision problems that require dynamic collaboration. Teams of agents need to coordinate decisions in many domains, but naive approaches fail due to the exponential growth of the joint action space with the number of agents. We circumvent this complexity through an approach that allows us to trade computation for approximation quality and dynamically coordinate actions. Our algorithm comprises three elements: online planning with Monte Carlo Tree Search (MCTS), factored representations of local agent interactions with coordination graphs, and the iterative Max-Plus method for joint action selection. We evaluate our approach on the benchmark SysAdmin domain with static coordination graphs and achieve comparable performance with much lower computation cost than our MCTS baselines. We also introduce a multi-drone delivery domain with dynamic coordination graphs, and demonstrate how our approach scales to large problems on this domain that are intractable for other MCTS methods. We provide an open-source implementation of our algorithm at https://github.com/JuliaPOMDP/FactoredValueMCTS.jl.

Cellular Automata Modeling of Stem-Cell-Driven Development of Tissue in the Nervous System.

Mathematical and computational modeling enables biologists to integrate data from observations and experiments into a theoretical framework. In this review, we describe how developmental processes associated with stem-cell-driven growth of tissue in both the embryonic and adult nervous system can be modeled using cellular automata (CA). A cellular automaton is defined by its discrete nature in time, space, and state. The discrete space is represented by a uniform grid or lattice containing agents that interact with other agents within their local neighborhood. This possibility of local interactions of agents makes the cellular automata approach particularly well suited for studying through modeling how complex patterns at the tissue level emerge from fundamental developmental processes (such as proliferation, migration, differentiation, and death) at the single-cell level. As part of this review, we provide a primer for how to define biologically inspired rules governing these processes so that they can be implemented into a CA model. We then demonstrate the power of the CA approach by presenting simulations (in the form of figures and movies) based on building models of three developmental systems: the formation of the enteric nervous system through invasion by neural crest cells; the growth of normal and tumorous neurospheres induced by proliferation of adult neural stem/progenitor cells; and the neural fate specification through lateral inhibition of embryonic stem cells in the neurogenic region of Drosophila.

Fair Topologies: Community Structures and Network Hubs Drive Emergence of Fairness Norms

Fairness has long been argued to govern human behavior in a wide range of social, economic, and organizational activities. The sense of fairness, although universal, varies across different societies. In this study, using a computational model, we test the hypothesis that the topology of social interaction can causally explain some of the cross-societal variations in fairness norms. We show that two network parameters, namely, community structure, as measured by the modularity index, and network hubiness, represented by the skewness of degree distribution, have the most significant impact on emergence of collective fair behavior. These two parameters can explain much of the variations in fairness norms across societies and can also be linked to hypotheses suggested by earlier empirical studies in social and organizational sciences. We devised a multi-layered model that combines local agent interactions with social learning, thus enables both strategic behavior as well as diffusion of successful strategies. By applying multivariate statistics on the results, we obtain the relation between network structural features and the collective fair behavior.

非対称相互作用により動的秩序を生み出す群ロボット

Swarm robotic systems have an advantage that they can execute difficult tasks that cannot be achieved by using a single robot. Self-organization is a useful method to control swarm robots because it enables to create collective behaviors from local interactions of agents that perform simple behavior. We previously proposed a simple model for self-organization that is based on asymmetric interactions, and demonstrated via simulation that various non-trivial patterns emerge. In this study, we examine the applicability of the proposed model to a swarm robotic system. As a first step, we developed five mobile agents and demonstrated via real-world experiments that they can reproduce simulation results.

Division of labor in a swarm of autonomous underwater robots by improved partitioning social inhibition

In this paper, a distributed algorithm for adaptive task allocation and adaptable partitioning of a swarm into different work-groups is proposed and used in a swarm of underwater robots. The algorithm is based on local interactions of agents and is inspired by honeybee age-polyethism. It is adaptive to changes in the swarm size (workforce) and relative demands (workload) for different tasks and it limits the number of switchings of agents between different tasks enabling specialization of agents. The preliminary version of the algorithm was introduced previously. Here a fully decentralized version of the algorithm is proposed that improves the previous version by removing the need for global information. The algorithm is successfully implemented in swarms of physically-embodied underwater robots while the swarm size and the demands for the tasks change over the course of the experiments.

Cluster fluctuation in two-dimensional lattices with local interactions

It is known that under the local interactions of agents, we find cluster formations of states of agents, but few works exist on the changes (fluctuations) of clusters. This paper deals with the analysis of fluctuations of clusters with respect to states of agents placed on two-dimensional lattices having local interactions. We use two models; first we use Model S that consists of agents producing goods, whose states denote the activity of agents, as well as the accompanying multiple states. An agent decides the states in the subsequent time period by monitoring the states of neighboring agents. Then, we introduce Model P to cope with cases that are more realistic. In Model P, we assume that there exists a group of agents comprising a pool of labor agents and several firm agents collaborating in production, where the labor agents are free to move neighborhoods seeking better incomes. We find cluster formation and equilibrium in both models after the passage of time, but these cluster formations deform and become other formations by changing certain states of agents into random states, or by changing parameters to be random. In simulation studies, we show the conditions for the fluctuations of clusters and possible real applications.

Dynamic, small‐world social network generation through local agent interactions

To model agent relationships in agent‐based models, it is often necessary to incorporate a social network whose topology is commonly assumed to be “small‐world.” This is potentially problematic, as the classification is broad and covers a wide‐range of network statistics. Furthermore, real networks are often dynamic, in that edges and nodes can appear or disappear, and spatial, in that connections are influenced by an agent's position within a particular social space. These properties are difficult to achieve in current network formation tools. We have, therefore, developed a novel social network formation model, that creates and dynamically adjusts small‐world networks using local spatial interactions, while maintaining tunable global network statistics from across the broad space of possible small‐world networks. It is, therefore, a useful tool for multiagent simulations and diffusion processes, particularly those in which agents and edges die or are constrained in their movement within some social space. We also show, using a simple epidemiological diffusion model, that a range of networks can all satisfy the small‐world criterion, but behave quite differently. This demonstrates that it is problematic to generalize results across the whole space of small‐world networks. © 2014 Wiley Periodicals, Inc. Complexity 19: 44–53, 2014

Genetic Team Composition and Level of Selection in the Evolution of Cooperation

In cooperative multiagent systems, agents interact to solve tasks. Global dynamics of multiagent teams result from local agent interactions, and are complex and difficult to predict. Evolutionary computation has proven a promising approach to the design of such teams. The majority of current studies use teams composed of agents with identical control rules (ldquogenetically homogeneous teamsrdquo) and select behavior at the team level (ldquoteam-level selectionrdquo). Here we extend current approaches to include four combinations of genetic team composition and level of selection. We compare the performance of genetically homogeneous teams evolved with individual-level selection, genetically homogeneous teams evolved with team-level selection, genetically heterogeneous teams evolved with individual-level selection, and genetically heterogeneous teams evolved with team-level selection. We use a simulated foraging task to show that the optimal combination depends on the amount of cooperation required by the task. Accordingly, we distinguish between three types of cooperative tasks and suggest guidelines for the optimal choice of genetic team composition and level of selection.

Self-emerging coordination mechanisms for knowledge integration processes

The increasing knowledge intensity of jobs, typical of a knowledge economy, highlights the role of firms as integrators of know-how and skills. As economic activity becomes mainly intellectual and requires the integration of specific and idiosyncratic skills, firms need to allocate skills to tasks and traditional hierarchical control results increasingly ineffective. In this work, we explore under what circumstances networks of agents, which bear specific skills, may self-organize in order to complete tasks. We use a computer simulation approach and investigate how local interaction of agents, endowed with skills and individual decision-making rules, may produce aggregate network structure able to perform tasks. To design algorithms that mimic individual decision-making, we borrow from computer science literature and, in particular, from studies addressing protocols that produce cooperation in Peer-to-Peer networks. We found that self-organization depends on the structural features of, formal or informal, organizational networks embedding both professionals, holding skills, and project managers, holding access to jobs.

FirmNet: the scope of firms and the allocation of task in a knowledge-based economy

The increasing knowledge intensity of jobs, typical of a knowledge economy, highlights the role of firms as integrators of know how and skills. As economic activity becomes mainly intellectual and requires the integration of specific and idiosyncratic skills, firms need to allocate skills to tasks and traditional hierarchical control may result increasingly ineffective. In this work, we explore under what circumstances networks of agents, which bear specific skills, may self-organize in order to complete tasks. We use a computer simulation approach and investigate how local interaction of agents, endowed with skills and individual decision-making rules, may produce aggregate network structure able to perform tasks. To design algorithms that mimic individual decision-making, we borrow from computer science literature and, in particular, from studies addressing protocols that produce cooperation in P2P networks. We found that self-organization depends on imitation of successful peers, competition among agents holding specific skills, and the structural features of, formal or informal, organizational networks embedding both professionals, holding skills, and project managers, holding access to jobs.

A model of a trust-based recommendation system on a social network

In this paper, we present a model of a trust-based recommendation system on a social network. The idea of the model is that agents use their social network to reach information and their trust relationships to filter it. We investigate how the dynamics of trust among agents affect the performance of the system by comparing it to a frequency-based recommendation system. Furthermore, we identify the impact of network density, preference heterogeneity among agents, and knowledge sparseness to be crucial factors for the performance of the system. The system self-organises in a state with performance near to the optimum; the performance on the global level is an emergent property of the system, achieved without explicit coordination from the local interactions of agents.

Agent-based Modeling and Simulation of Biomolecular Reactions

We present an agent-based, three-dimensional model of phage lambda, a virus infecting E. coli which exhibits two phases, lysogenic and lytic. This process is useful for bio-based sensors or switches and is a widely studied gene regulatory system. We model system constituents as software agents. Complex system behavior emerges from local agent interactions. Agent-based modeling lets us study how individual parameters affect overall system behavior. This bottom-up approach is an alternative to top-down approaches using differential equations and stochastic simulation. It can model any system of biomolecular reactions, with applications in physiology, pharmacology, medicine, environmental monitoring, and homeland security.