We present a novel method of machine learning toward autonomous developmental systems. The method is based on a growing neural network that initially produces senseless signals but later associates rewarding signals and quasi-rewarding signals with recent perceptions and motor activities and, based on these associations, incorporates new cells and creates new connections, which results in more structured output patterns. The rewarding signals are produced in a device called “pleasure center”, while the quasi-rewarding signals (that represent pleasure expectation) are generated by the network itself. The network was tested using a simulated mobile robot equipped with a pair of motors, a speaker, a set of touch sensors, and a camera. Despite a lack of innate wiring for any purposeful behavior, the robot developed from scratch, without any external guidance (except hardwired perception-pleasure patterns), a set of perception-reaction patterns. The emerging patterns include obstacle avoidance, vocalization of interest, and approaching an object of interest, which are fundamental for creatures and usually handcrafted in traditional robotic systems.
This paper studies a method for blind identification based upon independent component analysis. By observing vibration of a mechanical system subject to unknown but independent signals, the method makes it possible to identify partial data of the system parameter and to estimate the unknown input signals. Then, by monitoring independence of the estimated signals, the paper gives a method for detecting the change in mechanical parameter. This can be applied to fault detection of operating machines without any special sensor for the fault. An experiment with a flexible structure is carried out to verify these methods.
Gaming is one of the good tools to understand or study complex phenomena through experiences in a virtual world. Now, computer agents are beginning to join gaming as substitutes for human players. To help finding strategies through a gaming, this paper proposes an agent-based model for gaming-simulation. In this model, each agent has its own neural-networks for predicting behavior of other agents, including itself. In addition, each agent has a classifier model for tactical decision-making, and to achieve tactical target, the agent uses neural-networks to get an optimal answer.These agents try to find tactical rules with playing the game that aims at the second place. It is shown that this three-model structure enables us to monitor behavior of agents easily, and it enables us to consider strategies in the world of gaming.
This paper is concerned with PID control for n-state, r-input and m-output linear system (MIMO system). We propose a new PID parameters determination method based on an eigenvalue assignment. At first a new eigenvalue assignment method with a static output feedback is proposed. Then we regard a PID control as the static output feedback and apply its eigenvalue assignment method in order to determine the static output feedback gain related to PID parameters. Consequently the PID parameters are determined by matrices operation after giving an appropriate parametrized vector under a condition 2m + r ≥ n + 1. The effectiveness of the proposed method is confirmed with the simulation results for unstable MIMO systems.
This paper provides a new solution to the output feedback H∞ control problem for descriptor systems. Unlike previous results, the proposed criterion for existence of H∞, suboptimal controllers does not depend on the choice of the descriptor realization. The criterion is given in terms of LMIs, whose solution yields an H∞ suboptimal controller.