Swarming behavior of multi—agent systems

We consider an anisotropic swarm model with an attraction/repulsion function and study its aggregation properties. It is shown that the swarm members will aggregate and eventually form a cohesive cluster of finite size around the swarm center in a finite time. Moreover, we extend our results to more general attraction/repttlsion functions. Numerical simulations demonstrate that all agents will eventually enter into and remain in a bounded region around the swarm center which may exhibit complex spiral motion due to asymmetry of the coupling structure. The model in this paper is more general than isotropic swarms and our results provide fiarther insight into the effect of the interaction pattem on individual motion in a swarm system.     

Supervised Self-Organization of Homogeneous Swarms Using Ergodic Projections of Markov Chains

This paper formulates a self-organization algorithm to address the problem of global behavior supervision in engineered swarms of arbitrarily large population sizes. The swarms considered in this paper are assumed to be homogeneous collections of independent identical finite-state agents, each of which is modeled by an irreducible finite Markov chain. The proposed algorithm computes the necessary perturbations in the local agents' behavior, which guarantees convergence to the desired observed state of the swarm. The ergodicity property of the swarm, which is induced as a result of the irreducibility of the agent models, implies that while the local behavior of the agents converges to the desired behavior only in the time average, the overall swarm behavior converges to the specification and stays there at all times. A simulation example illustrates the underlying concept.     

Stability analysis of local swarms with attraction/repulsion force

This paper considers finite sense ability into swarm system. We first construct a model of local swarms with a class of attraction and repulsion. Then, we analyze aggregation motion of the swarm. It is shown that the individuals of the swarm will aggregate and eventually enter into a bounded hypeball around the swarm center in finite time and we present the range of the finite visual radius. We finally analyze their motion in convergent region.     

Markov Chain Approach to Probabilistic Guidance for Swarms of Autonomous Agents

Motivated by biological swarms occurring in nature, there is recent interest in developing swarms comprised completely of engineered agents. The main challenge for developing swarm guidance laws compared to earlier formation flying and multi‐vehicle coordination approaches is the sheer number of agents involved. While formation flying applications might involve up to 10 to 20 agents, swarms are desired to contain hundreds to many thousands of agents. In order to deal with the sheer size, the present paper makes a break with past deterministic methods, and considers the swarm as a statistical ensemble for which guidance can be performed from a probabilistic point of view. The probability‐based approach takes advantage of the law of large numbers, and leads to computationally tractable and implementable swarm guidance laws. Agents following a probabilistic guidance algorithm make statistically independent probabilistic decisions based solely on their own state, which ultimately guides the swarm to the desired density distribution in the configuration space. Two different synthesis methods are introduced for designing probabilistic guidance laws. The first is based on the Metropolis‐Hastings (M‐H) algorithm, and the second is based on using linear matrix inequalities (LMIs). The M‐H approach ensures convergent swarm behavior subject to enforced desired motion constraints, while the LMI approach additionally ensures exponential convergence with a prescribed decay rate, and allows minimization of a cost function that reflects fuel expenditure. In addition, both algorithms endow the swarm with the property of self‐repair, and the capability to strictly enforce zero‐probability keep‐out regions. This last property requires a slight generalization of the Perron‐Frobenius theory, and can be very useful in swarm applications that contain regions where no agents are allowed to go. Simulation examples are given to illustrate the methods and demonstrate desired properties of the guided swarm.     

Self-organized motion in anisotropic swarms

This paper considers an anisotropic swarm model with a class of attraction and repulsion functions. It is shown that the members of the swarm will aggregate and eventually form a cohesive cluster of finite size around the swarm center. Moreover, It is also proved that under certain conditions, the swarm system can be completely stable, i.e., every solution converges to the equilibrium points of the system. The model and results of this paper extend a recent work on isotropic swarms to more general cases and provide further insight into the effect of the interaction pattern on self-organized motion in a swarm system.     

Training multi-agent teams from zero knowledge with the competitive coevolutionary team-based particle swarm optimiser

A new competitive coevolutionary team-based particle swarm optimiser (CCPSO(t)) algorithm is developed to train multi-agent teams from zero knowledge. The CCPSO(t) algorithm is applied to train a team of agents to play simple soccer. The algorithm uses the charged particle swarm optimiser in a competitive and cooperative coevolutionary training environment to train neural network controllers for the players. The CCPSO(t) algorithm makes use of the FIFA league ranking relative fitness function to gather detailed performance metrics from each game played. The training performance and convergence behaviour of the particle swarm are analysed. A hypothesis is presented that explains the lack of convergence in the particle swarms. After applying a clustering algorithm on the particle positions, a detailed visual and quantitative analysis of the player strategies is presented. The final results show that the CCPSO(t) algorithm is capable of evolving complex gameplay strategies for a complex non-deterministic game.     

Asynchronous parallelization of particle swarm optimization through digital pheromone sharing

In this paper, a model for sharing digital pheromones between multiple particle swarms to search n-dimensional design spaces in an asynchronous parallel computing environment is presented. Particle swarm optimization (PSO) is an evolutionary technique used to effectively search multi-modal design spaces. With the aid of digital pheromones, members in a swarm can better communicate with each other to improve search performance. Previous work by the authors demonstrated the capability of digital pheromones within PSO for searching the global optimum in both single and coarse grain synchronous parallel computing environments. In the coarse grain approach, multiple swarms are simultaneously deployed across various processors and synchronization is carried out only when all swarms achieved convergence, in an effort to reduce processor-to-processor communication and network latencies. However, it is theorized that with an appropriate parallelization scheme, the benefits of digital pheromones and communication between swarms can outweigh the network bandwidth latencies resulting in improved search efficiency and accuracy. To explore this idea, a swarm is deployed in the design space across different processors. Through an additional processor, each part of the swarm can communicate with the others. While digital pheromones aid communication within a swarm, the developed parallelization model facilitates communication between multiple swarms resulting in improved search accuracy and efficiency. The development of this method and results from solving several multi-modal test problems are presented.     

基于动态多粒子群的多目标优化算法

为了让多目标粒子群优化算法在运行过程中保持粒子的多样性,提出了一种初始化方法和动态多粒子群协作的多目标优化算法。根据粒子群在决策空间中的分布情况动态增加或者减少粒子群数量;为避免粒子收敛速度过快,改进了决定粒子飞行速度的因素,速度值依赖于粒子当前速度惯性、粒子最优值,群最优值和所有群最优值。用五个测试函数对算法进行了测试并与多目标粒子群优化进行了比较,测试结果表明提出的算法优于多目标粒子群优化算法。     

Inherent Noise‐Aware Insect Swarm Simulation

Collective behaviour of winged insects is a wondrous and familiar phenomenon in the real world. In this paper, we introduce a highly efficient field‐based approach to simulate various insect swarms. Its core idea is to construct a smooth yet noise‐aware governing velocity field that can be further decomposed into two sub‐fields: (i) a divergence‐free curl‐noise field to model noise‐induced movements of individual insects in a swarm, and (ii) an enhanced global velocity field to control navigational paths in a complex environment along which all the insects in a swarm fly. Through simulation experiments and comparisons with existing crowd simulation approaches, we demonstrate that our approach is effective to simulate various insect swarm behaviours including aggregation, positive phototaxis, sedation, mass‐migrating, and so on. Besides its high efficiency, our approach is very friendly to parallel implementation on GPUs (e.g. the speedup achieved through GPU acceleration is higher than 50 if the number of simulated insects is more than 10 000 on an off‐the‐shelf computer). Our approach is the first multi‐agent modelling system that introduces curl‐noise into agents' velocity field and uses its non‐scattering nature to maintain non‐colliding movements in 3D crowd simulation.     

Automatic calibration a hydrological model using a master–slave swarms shuffling evolution algorithm based on self-adaptive particle swarm optimization

Parameter estimation for hydrological models is a challenging task, which has received significant attention by the scientific community. This paper presents a master–slave swarms shuffling evolution algorithm based on self-adaptive particle swarm optimization (MSSE-SPSO), which combines a particle swarm optimization with self-adaptive, hierarchical and multi-swarms shuffling evolution strategies. By comparison with particle swarm optimization (PSO) and a master–slave swarms shuffling evolution algorithm based on particle swarm optimization (MSSE-PSO), MSSE-SPSO is also applied to identify HIMS hydrological model to demonstrate the feasibility of calibrating hydrological model. The results show that MSSE-SPSO remarkably improves the calculation accuracy and is an effective approach to calibrate hydrological model.     

Sliding mode control-based algorithms for consensus in connected swarms

In this article, finite-time consensus algorithms for a swarm of self-propelling agents based on sliding mode control and graph algebraic theories are presented. Algorithms are developed for swarms that can be described by balanced graphs and that are comprised of agents with dynamics of the same order. Agents with first and higher order dynamics are considered. For consensus, the agents' inputs are chosen to enforce sliding mode on surfaces dependent on the graph Laplacian matrix. The algorithms allow for the tuning of the time taken by the swarm to reach a consensus as well as the consensus value. As an example, the case when a swarm of first-order agents is in cyclic pursuit is considered.     

Coordinating metaheuristic agents with swarm intelligence

Coordination of multi agent systems remains as a problem since there is no prominent method suggests any universal solution. Metaheuristic agents are specific implementations of multi-agent systems, which imposes working together to solve optimisation problems using metaheuristic algorithms. An idea for coordinating metaheuristic agents borrowed from swarm intelligence is introduced in this paper. This swarm intelligence-based coordination framework has been implemented as swarms of simulated annealing agents collaborated with particle swarm optimization for multidimensional knapsack problem. A comparative performance analysis is also reported highlighting that the implementation has produced much better results than the previous works.     

Leader-based versus soft control of multi-agent swarms

We focus on the control of heterogeneous swarms of agents that evolve in a random environment. Control is achieved by introducing special agents: leader and infiltrated (shill) agents. A refined distinction is made between hidden and apparent controlling agents. For each case, we provide an analytically solvable example of swarm dynamics.     

Decentralized coordination of autonomous swarms using parallel Gibbs sampling

In this paper we present analysis of a discrete-time, decentralized, stochastic coordination algorithm for a group of mobile nodes, called an autonomous swarm, on a finite spatial lattice. All nodes take their moves by sampling in parallel their locally perceived Gibbs distributions corresponding to a pairwise, nearest-neighbor potential. The algorithm has no explicit requirements on the connectedness of the underlying information graph, which varies with the swarm configuration. It is established that, with an appropriate annealing schedule, the algorithm results in swarm configurations converging to the (global) minimizers of a modified potential energy function. The extent of discrepancy between the modified and original potential energy functions is determined by the maximum node travel between time steps, and when such distance is small, the ultimate swarm configurations are close to the global minimizers of the original potential energy. Simulation results are further presented to illustrate the capability of the sampling algorithm in approximate global optimization for swarms.     

确定性核粒子群的粒子滤波跟踪算法及其CRLB推导

针对运动声阵列在有色噪声环境中的非线性滤波跟踪问题,提出一种确定性核粒子群的粒子滤波算法.该算法通过确定性初始化核粒子集、确定性后验概率密度函数及粒子群与核粒子集更新方式来提高跟踪的精度,并推导出该算法的理论误差性能下界.与传统的粒子滤波算法相比,仿真结果表明了所提出算法的有效性和优越性.     

A particle swarm optimisation algorithm with interactive swarms for tracking multiple targets

We propose a novel particle swarm optimisation algorithm that uses a set of interactive swarms to track multiple pedestrians in a crowd. The proposed method improves the standard particle swarm optimisation algorithm with a dynamic social interaction model that enhances the interaction among swarms. In addition, we integrate constraints provided by temporal continuity and strength of person detections in the framework. This allows particle swarm optimisation to be able to track multiple moving targets in a complex scene. Experimental results demonstrate that the proposed method robustly tracks multiple targets despite the complex interactions among targets that lead to several occlusions.     

Analysing robot swarm behaviour via probabilistic model checking

An alternative to deploying a single robot of high complexity can be to utilise robot swarms comprising large numbers of identical, and much simpler, robots. Such swarms have been shown to be adaptable, fault-tolerant and widely applicable. However, designing individual robot algorithms to ensure effective and correct overall swarm behaviour is actually very difficult. While mechanisms for assessing the effectiveness of any swarm algorithm before deployment are essential, such mechanisms have traditionally involved either computational simulations of swarm behaviour, or experiments with robot swarms themselves. However, such simulations or experiments cannot, by their nature, analyse all possible swarm behaviours. In this paper, we will develop and apply the use of automated probabilistic formal verification techniques to robot swarms, involving an exhaustive mathematical analysis, in order to assess whether swarms will indeed behave as required. In particular we consider a foraging robot scenario to which we apply probabilistic model checking.     

基于均匀设计的粒子群算法及其在飞控系统中的应用

将粒子群算法应用于飞行控制系统的优化设计中,需要解决两个问题：如何选择目标函数和如何确定初始种群和算法运行参数。针对这两个问题,分别提出了基于参考模型的飞行控制系统优化策略和基于均匀设计的粒子群算法初始种群和运行参数的选择方法。仿真结果表明,本文所提出的优化策略能够有效地解决飞行控制系统的优化设计问题,粒子群初始种群分布均匀,收敛速度快。     

多Agent 主从粒子群分布式计算框架

面向大规模复杂优化问题,提出了一个基于并行粒子群优化的分布式Agent计算框架.框架中使用一个主群(master swarm)来演化问题的完整解,并使用一组从群(slave swarm)来并行优化一组子问题的解,主群和从群通过交替执行来提高问题的求解效率.采用异步组结构,主群/从群中的各类Agent共享一个解群,并通过相互协作,对解群进行构造、改进、修补、分解和合并等演化操作.该框架可用于求解复杂的约束多目标优化问题.通过一类典型运输问题上的实验,其结果表明,所提出的方法明显优于另外两种先进的演化算法.     

A MULTISWARM COMPETITIVE PARTICLE SWARM ALGORITHM FOR OPTIMIZATION CONTROL OF AN ETHYLENE CRACKING FURNACE

A fuzzy C-means (FCM) multiswarm competitive particle swarm optimization (FCMCPSO) algorithm is proposed, in which FCM clustering is used to divide swarms adaptively into different clusters. The large-scale swarms are according to the standard particle swarm optimization (PSO) algorithm, whereas the small-scale swarms search randomly in the neighborhood of the optimal solution to increase the probability of jumping out of the local optimization point. Within every cluster, the adaptive value gained by competitive learning is respectively found and arranged in order. Swarms of small adaptive value were integrated with the neighboring swarms of large adaptive value to search the optimal solution competitively by the swarms. The algorithm's validity was tested by benchmark functions and compared with other PSO algorithms. Furthermore, an integrated FCMCPSO-radial basis function neural network was studied for nonlinear system modeling and intelligent optimization control of cracking depth of an ethylene cracking furnace application in a chemical process.