Research On Population Evacuation Modeling And Simulation Based On Group Intelligence

In recent years, public security issues become increasingly prominent. Large scale crowdevacuation has become an important part in emergency treatment, which cannot be ignored.Study on behavioral characteristics and motion laws of emergency evacuation, as well as riskassessment of evacuation capacity for public places are effective ways to handle public securityproblems. Crowd evacuation under emergent cases is an extremely complex task. Despiteevacuation drill contributes to the best evacuation plan, this method has many disadvantages,such as high experimental cost, personnel security problems, and so on. The emergence ofcomputer simulation provides an effective tool for crowd evacuation simulation. Crowdevacuation simulations under emergent cases include environment modeling, path optimizationand crowd motion modeling. To solve these problems, this paper puts forwards correspondingmodeling and simulation methods.To solve the problems of high cost and low accuracy in traditional3D environmentmodeling method, this paper puts forward a monocular image based interactive3D-scenereconstruction method to achieve high fidelity with low cost. For crowd evacuation simulationunder emergent case, there exists many crowd motion models which is build up by usingcomputer simulation technology. However, few of these models introduce and quantify theimpacts of psychological factors in crowd evacuation. In this paper, we propose a combinedcrowd evacuation model. It adopts a co-evolutionary Artificial Bee Colony algorithm asmacroscopic model to instruct crowd path planning, and adopts improved social force model andParticle Swarm Optimization algorithm as microscopic model to guide individual motions.The main work and innovation of this paper are listed as follows.(1) Taking the characteristics of scene modeling into account, this paper put forwards amonocular image based interactive3D-scene reconstruction method. This method takes singleimage as input, which reduce the requirement for input. In image processing, we add humancomputer interactions and take the interaction information to instruct transformation from2Dimage to3D model, which largely decrease the time for image processing. The monocular imagebased interactive3D-scene reconstruction method is based on three kinds of primitives. Themodeling process includes primitive construction, realistic processing based on images andinteractive manipulation. Finally, we get the3D scenes which we need. Experimental resultsshow that the algorithm is low cost and robust, and the user can get the desired3D-scene quickly.(2) To simulate the grouping behavior in crowd evacuation, this paper proposes a novelmicroscopic crowd simulation model, which takes the grouping characteristics into account. Thismodel includes three parts: grouping method based on group attraction, social force modelconsidering grouping characteristics, and crowd motion model based on PSO algorithm. Thismodel divides a large crowd into small ones according to relationships, such as family,colleagues, friends, and so on. According to the grouping results, improved social force modeland PSO algorithm are used to instruct the microscopic motion process of crowd, which will contribute to the phenomenon of grouping and arching.(3) This paper introduces multi-species co-evolution into canonical ABC algorithm, andproposes a novel co-evolutionary ABC algorithm based on hierarchical communication model(HABC). This method adopts dynamical adjustment strategies on species and groups to reducethe computational complexity. Performance tests and visualization simulations show that theHABC algorithm exhibit good performance on accuracy, robustness and convergence speed. Thispaper applied HABC algorithm to the macroscopic path planning, which provide path selectionto individuals.(4) Combined with corresponding research project, we applied the above theory to thedesign of scene modeling and crowd evacuation modeling in the crowd evacuation modeling andsimulation system. In this system, we use the monocular image based interactive3D-scenereconstruction method to build environment models. HABC algorithm is used to instruct crowdpath planning and microscopic crowd simulation model is use to instruct individuals’ motion.Through crowd evacuation simulation experiments and analysis on this system, we conclude theimpacts of population density and exit number on evacuation time. Experiments show that themethod proposed in this paper is practical and efficient.