小圷 成一

We compare the performance of four types of controllers for an autonomous mobile robot through computer simulation. Each controller consists of different kinds of architectures of neural networks (NN). Four types of NNs are as follows: Sigmoid NN, Radial NN, pulsed NN (PNN), and PNN incorporating short term synaptic plasticity (PNNSSP). In computer simulation, parameters of NNs are adjusted using genetic algorithm. The results of computer simulation show that PNNSSP is more efficient than other typs of NNs for the controller of the autonomous mobile robot.

In this paper, a Hebbian learning rule restraining “catastrophic forgetting” is proposed on pulse neural network (PNN) with leaky integrate-and-fire neurons. The strong point of this learning rule is that a learning of new pattern does not destroy past ones, and that an efficient use of synapses is enabled. First, in order to consider the function of the learning rule, a fundamental experiment is made. Next, to compare the performance between the proposed learning rule and conventional ones on the application, simulation experiments are examined using autonomous behavior robots which are forced to learn concurrently two different environments. The results of the experiments show that the proposed learning rule clearly restrains “catastrophic forgetting” and enables working of more efficient than conventional PNN learning.

In this paper we propose a discrete version of tunneling algorithm and apply it to a placement problem of VLSI layout. Although tunneling algorithm is powerful to obain a near global minimum of multimodal function optimization, it is applicable only to continuous optimization problems. We incorporate a heuristic search method into conventional tunneling algorithm so that tunneling algorithm can deal with discrete optimization problems such as VLSI placement and routing problems. We also introduce a hierarchical placement model in order to improve the search efficiency of tunneling algorithm and to reduce its computation time. Computational experiments show that proposed hierarchical method obtains better placement results with shorter running time than non-hierarchical method.

Genetic algorithms (GAs) are search procedures for combinatorial optimization problems. Unlike most of other optimization techniques, GAs search the solution space using a population of solutions. Although GAs have an excellent global search ability, it is not effective for searching the solution space locally due to crossover-based-search. Genetic Immune Recruitment Mechanism (GIRM) makes up for the week point of GAs. In GIRM, new generated solutions take an immune recruitment test. Only those solutions which are similar to the best solution in the population pass the test and survive to the next generation. As a result, GIRM promotes the search around the best solution. In this paper we apply GIRM to the floorplan design problem of VLSI layout and compare the results with the ones of GAs. Especially, we propose new affinity functions in order to evaluate the similarity between floorplans. Computational experiments show that GIRM gives equivalent or better solutions than GAs.