In this paper, we propose a high quality sampling technique of Particle-based volume rendering (PBVR) which is suitable for handling a complicated ultra-scale volume dataset composed of multiple sub-volumes. In PBVR, we estimate a number of particles by integrating a particle density function derived by an opacity transfer function and generate particles by using a stochastic technique. Previously, we approximated the density function using a uniform function in a volume cell which may result in an inaccurate estimation of the particle number and a sampling failure when we deal with a transfer function with drastic changes. To solve the problem, we employ a Gibbs sampling technique and apply the proposed technique to several hexahedral unstructured volume datasets in order to confirm the effectiveness.

抄録全体を表示

  Product recall, which is to be executed by the self-judgment of product producers, is a social scheme for protecting consumers from unanticipated damage due to defective products. As people's demands for safety have increased, it is required socially as well as administratively to establish an effective and trustworthy scheme for product recall. Institutional design of product recall, however, has not yet been studied with rationale; some method for doing it is highly desired.
  In this research, we applied a method of social simulation that is based on a multi-agent model to product recall in order for developing a simulation method useful for its institutional design. The simulation model consists of agents representing producers, consumers, and news media, and producers and consumers behave seeking their own profits. By observing interacting behaviors of the agents in a virtual society, it is possible to search for the conditions where the designated goals of product recall are achievable. As a result of test simulation, it was demonstrated that the proposed simulation is a useful mean for institutional design of product recall.

抄録全体を表示

  One of the solutions to traffic issues is to suppress the dependency of society to car traffic, to create a human-centered road space, and to construct new traffic systems that consider global environment as well as mobility-impaired people. In a new traffic system, existence of pedestrian becomes more and more important. However there are few examples of mixed traffic simulations of pedestrians and cars in which subjectivity of pedestrian is treated in the same way as cars. We have been developing a Multi-Agent based Traffic and Environment Simulator, MATES. In this research, we newly construct the simulation method in which coexistence of pedestrians and cars can be expressed. Especially, we newly propose the multilevel pedestrian model to achieve pedestrian's collective behaviors and their interaction with cars. In the model, pedestrian agents are grouped together according to the particle size of agents to interact, and the constructed group is modeled as a virtual agent. Agents recognize their surrounding circumstances and interact among others through the virtual agent. We implement the multilevel pedestrian model to MATES and verify its performance.

抄録全体を表示

  The Look-Back GMRES (k) method with restart coefficient k which determines an initial guess by an approximate solution vector obtained at a previous restart was proposed by Imakura et al. After that, we proposed Multi-Restarts Look-Back GMRES (k, m) method by an approximate solution vector obtained at previous multi restarts as an extension of the Look-Back GMRES (k) method. It was also clarified that the Multi-Restarts Look-Back GMRES (k, m) method outperforms the original Look-Back GMRES (k) method. In this paper, we consider and propose a separative type of the Multi-Restarts Look-Back GMRES (k, m) method in order to enhance the convergence rate. Numerical results will make clear effectiveness of the separative technique for the Multi-Restarts Look-Back GMRES (k, m) method.

抄録全体を表示