In a central-air-conditioning system, a centralized unit cools or heats water before its circulation throughout a building. By using cooled or heated water, each air conditioner cools or heats each room. To control room temperature, the centralized unit needs to create heating value (heat demand) enough to cool or heat all rooms. To satisfy both indoor comfort and energy saving, the centralized unit needs to create the minimum heat demand that is necessary to satisfy indoor comfort. Heat demand is calculated from temperature difference and water flow rate. However, flow sensors are rarely attached because they are expensive. In this work, we developed a soft-sensor for predicting water flow rate. Although linear regression models have been widely used, they cannot always achieve high estimation accuracy because most of the systems are non-linear and changes in system characteristics make models not suitable. Thus, we adopted locally weighted regression, which is a type of Just-In-Time modeling methods, and achieved high estimation performance.

In this paper, we propose a simulated annealing method with an evolutionary computation for reconstructing citizens’' attributes from statistics. To implement agent-based social simulations for real communities, it is needed to reconstruct citizens'’ attributes such as age, sex, income, occupation,academic background, and so on. Although such personal data are available in the local government,they are protected for privacy reason. In order to enable any body to implement agent-based social simulation, a reliable reconstruction method is required. In this paper, we modify a previous approach using simulated annealing by incorporating a method that minimizes errors between a generated population and the real statistics. Additionally we propose an evolutionary algorithm that is applied before the proposed modified simulated annealing. We show the effectivity of the proposed two-fold algorithm using some computational experiments.

To improve development efficiency, model based techniques have been increasingly applied in embedded systems. For example, MATLAB/Simulink is regularly used for expressing physical phenomena, including signal processing and control systems. While it is very effective in these areas, it lacks object oriented capabilities that have been proven to increase development efficiency in related domains. UML is effective for descriptions of structures using object oriented abstraction, however it cannot express physical phenomenon easily. Furthermore, in conventional software engineering,code-libraries are regularly used to deliver significant development efficiency improvements across a wide range of domains, however neither MATLAB/Simulink nor UML include code-libraries. In this paper, we propose an extended UML to combine an UML model, MATLAB/Simulink models and code-libraries in a single system so as to realize efficient development. In the extended UML, an entire system is defined as a structure that consists of model parts. Each model part is integrated using UML. After generating source codes using UML, additional source code details are added from MATLAB/Simulink models and code-libraries. We demonstrate the efficiency of the extended UML with an experiment that shows behavioral equivalence between the system created with the extended UML and the original simulation model.

A stability condition is investigated for linear multi-agent systems with time delays. The condition is represented as a region in the complex plane where all of the eigenvalues of the agent interconnection matrix must exist. The region is specified by the transfer function of the agent. Furthermore, by employing the stability condition, a consensus condition is also investigated for directed networks of dynamic agents each of which is described by a linear time invariant system with a time delay.

A new estimation method of heat loss coefficient for existing buildings is proposed. The method is focused on the heat loss through convective heat transfer, because the air infiltration is negligible in the buildings and the ventilation is controlled by the building Heating, Ventilation and Air Conditioning (HVAC) controller. The heat loss coefficient (Q-value) evaluation of the existing houses are reported in the prior arts and the energy performance of the existing buildings has been reported as one of the commissioning process. The methods enable the total energy performance, but they do not clarify the week insulation elements. In order to visualize the energy flow easily, the proposed method identifies system parameters based on the data-oriented building heat loss model,using heat transfer engineering, with the interior wall surface temperatures and room temperatures that are easily obtained. The correction factor is obtained for the window natural convection heat transfer, because the empirical heat transfer equation for the week natural heat transfer contains some inaccuracy.

This paper considers a distributed unconstrained optimization problem where each agent has a local convex cost function and the sum of these functions is defined as a global cost function. We propose an event-driven subgradient algorithm based on consensus control to minimize the global cost function. Each agent has an estimate of the optimal solution as a state. In the proposed algorithm, each agent sends its state to the neighbor agents only at trigger-times when the error of its state exceeds a threshold. We show that the error between the estimate of the global cost function of the agents given by the proposed event-driven algorithm and the optimal cost is upper bounded. The simulation results show that the convergence speed by the proposed event-driven algorithm improves and the number of trigger-times can be reduced compared with the existing subgradient methods.