A distributed autonomous approach to adaptive system design is investigated through the Cooperative Behavior Acquisition Problem (CBAP) by a homogeneous autonomous robot group. A task is supposed to be given only to the group of robots, for the purpose of putting the main interest on how to design on-line specialization mechanisms, which should be autonomous and adaptive. The robot group dealt by this paper is comprised of same autonomous robots connected by a triangular board. Reinforcement learning (RL) is adopted for a basic framework of the robot's decision-making mechanism, so that quick online learning can be expected. However, since RL in a simple form is not effective in developing a stable cooperative behavior in a multi-agent environment, a novel decision-making mechanism is designed using two RL units, of which the first RL unit is for predicting its partners' next states, and the other is for generating an action of its own. Several empirical experiments for three connected robots are conducted on a computer in order to investigate the effectiveness of the proposed mechanisms.
This paper addresses an adaptive generation method of desired velocity field for leader-follower type cooperative mobile robotic systems with decentralized passive velocity field control (PVFC) handling a common object in coordination. The proposed control method for cooperative mobile robotic systems is constructed by extending PVFC. A common desired velocity is generated by the initial states of the robotic systems and the interactions. Moreover, the whole system becomes passive against human interaction and the whole motion can be controlled by a leader which can be a human being. In the design, the stability and boundedness of the signals in the resultant system with the proposed control algorithm is also guaranteed. Finally, the proposed control algorithm is examined by computer simulations for cooperative tasks with 3-wheeled mobile robotic systems. The simulation results illustrate the validity of the proposed control algorithm.
The authors previously developed a length measurement system for straight pipes using Kalman filters and maximum likelihood method. In the system, the stationary waves in pipes were represented by state variables of a linear dynamic system, and on-line accurate length measurement of straight and curved pipes was achieved. In many situations, however, there are necessities of length measurement not only for straight and curved pipes but also for branch pipes. The paper proposes a measurement system of pipe length and branch point for branch pipes. A criterion for optimal modes of the stationary wave is also presented, and experiments show that an accurate measurement is realized with the method.
One of major research topics for behavior-based AI is to construct an appropriate sensor-motor, relation for an autonomous moving robot in an embedded environment, hopefully, with less preliminary setting by an autonomous robot designer. This paper proposes a new reinforcement learning algorithm, which is called the Continuous Space Classifier Generator (CSCG), for this problem. The major attraction of CSCG is that the state space and the action space of a learning agent are segmented simultaneously with the process of its behavior rule acquisition in the embedded environment. This means that a robot designer can be released from the segmentation of those two spaces, which is often crucial to the success of reinforcement learning. After showing the detail of CSCG, not only computer simulations but also experiments using a small real robot are conducted in order to illustrate the learning process of the proposed method.
In this paper, the numerical solution method of an on-ramp traffic control problem is discussed from the viewpoint of the application of evolutionary computation. The problem is formulated as a discrete-time nonlinear optimal control problem. The optimal control problem is then transferred to an optimization problem with successive constraints. A parallel-adaptive genetic algorithm is proposed to obtain the numerical solution of the problem. A real-time practical control method is also proposed. Numerical simulations for assessing the performance of the proposed methods are performed using observed data of Hanshin Expressway. As a result, the proposed method obtains the solution at comparatively small generations. This means shortening the calculation time, and is useful for the real-time control. Moreover, it is shown that the proposed real-time control is effective to reduce heavy congestion on the expressway.
This paper describes a parallel algorithm of Krawczyk's method that lends itself most naturally to the original sequential algorithm extended by R.E.Moore and S.T.Jones. It also presents an improvement to achieve efficient computation in parallel. The effectiveness is confirmed by an example.
A stabilizing control law is obtained for a pendulum whose weight can be moved based on a condition for suppressing vibrations of the swing derived by an energy-based method, and experiments are performed to examine the usefulness of the controller. The controller is realized by a servo system for the position of the weight that simply consists of a second-order lag and a bang-bang reference input obtained from the pendulum trajectory. The control system approaches the optimal one that minimizes the damping ratio of the pendulum when the time constant of the servo system approaches zero. The results of experiments show the effectiveness of the proposed method.