Masanori Goka

Constructive simulation of creative concept generation process in design: a research method for difficult-to-observe design-thinking processes

In this study, we conduct a computer simulation in order to capture characteristics or patterns in the concept generation process, which may lead to the generation of a creative design idea. This approach employs a research framework called ‘constructive simulation’, which may be effective for investigating a process that is difficult to observe both internally and externally. The simulation was conducted in four phases. In Phase 1, the virtual concept generation process was constructed on a semantic network. In Phase 2, the relevance of the process was confirmed using network theory. In Phase 3, the simulation was validated using ‘synthetic verification’ which confirms the relevance of the process with another real-life phenomenon, as well as the feasibility of creating more creative design ideas. In Phase 4, the characteristics or patterns in the actual concept generation process were inferred from those of the virtual concept generation process. The results suggest that thinking patterns in which explicit and ‘inexplicit’ concepts are continuously intertwined lead to creative design ideas.

Robust Swarm Robotics System Using CMA-NeuroES with IncrementalEvolution

Swarm robotics (SR) is a novel approach to the coordination of large numbers of homogeneous robots; SR takes inspiration from social insects. Each individual robot in an SR system (SRS) is relatively simple and physically embodied. Researchers aim to design robust, scalable, and flexible collective behaviours through local interactions between robots and their environment. In this study, a simulated robot controller evolved by a recurrent artificial neural network with the covariance matrix adaptation evolution strategy, i.e., CMANeuroES is adopted for incremental artificial evolution. Cooperative food foraging is conducted by our proposed controller as one of the most complex simulation applications. Since a high level of robustness is expected in an SRS, several tests are conducted to verify that incremental artificial evolution with CMANeuroES generates the most robust robot controller among the ones tested in simulation experiments.

Analysis of behavioural strategies on Evolutionary Swarm Robotics Systems using functional circles with Self-organizing Maps

Various techniques for developing adaptive collective behavior in Evolutionary Swarm Robotics Systems have received a lot of attention in recent years. However, it is difficult to analyze the behavior which these robots acquired through the evolutionary computation, and there was no decisive approach. We think that it is possible to analyze the characteristic traits of the strategy by understanding essential environmental information that is the base of the features of each individual robots. In this research, we attempt to develop the analysis method of the robots behavioural strategy using the description method of the perceptual world peculiar to living organisms. We propose the analysis method for the dynamics of robots behavioural strategy by creating functional circles. Moreover, these functional circles is obtained by clustering robots sensor input by Self-organizing Maps.

Apply incremental evolution with CMA-NeuroES controller for a robust swarm robotics system

A distributed autonomous robotic system consisted of many homogeneous autonomous robots without a global controller is often called a swarm robotics system. It is expected that a swarm of robots achieves much more complex tasks than the ones that a single robot can do. However, it is also well known that designing a robot controller for this system is generally quite difficult because the system level behavior is emerged from a result of many dynamical interactions within the system itself or between the system and its environment. In this paper, a robot controller is represented by a recurrent artificial neural network designed by environmental complexification with covariance matrix adaptation evolution strategy, i.e., CMA-NeuroES, adopted for the incremental artificial evolution method. Computer simulations are conducted to prove that the incremental artificial evolution with CMA-NeuroES generates the most robust robot controller among the ones tested in the simulated experiments.

Evolutionary Artificial Neural Networks using Extended Minimal Simulation on Evolutionary Robotics

In this paper, we try to construct a simulation using evolutionary artificial neural networks models for building the robot controller and adopt an expansion that introduces noise to acquire precise and advanced processing for the robot. In ER approach addressing virtual spaces that contain noise, a simulation technique is called Minimal Simulation. In some experiments, it was shown that the controller was able to be acquired by simulation, yet it does not reach the level that is necessary for the development of a general simulation methodology. Therefore, we propose Extended Minimal Simulation method that introduces alternative coding techniques and show the effect of this technique.

The Analysis of Multi-Agent Systems using the Ecological Method

An evolutionary Multi-Agent System is discussed as an autonomous decentralized system in which each agent is capable of adapting to its embodied environment by means of artificial evolution. This type of system is a promising approach to adaptive system design, although it is undoubtedly difficult to analyze what they are doing or how they interact with each other to achieve a given task. Moreover, it is also difficult to understand how they evolved their function. In this paper, the ecological method is experimentally applied to a multi-agent system to analyze the system dynamics. A food collecting problem is selected to illustrate how to use the ecological method to grasp characteristics of agents in a Multi-Agent System. After the consideration of the analytical results, we succeeded in improving the performance of the system. This proves that the ecological method for natural systems can effectively be applied to multi-agent systems.