No abstract

No abstract

No abstract

This paper presents a novel Intelligent Inference System (IIS) for the determination of an optimum preshape for multifingered robot hand grasping, given object under a manipulation task. The IIS is formed as hybrid agent architecture, by the synthesis of object properties, manipulation task characteristics, grasp space partitioning, lowlevel kinematical analysis, evaluation of contact wrench patterns via fuzzy approximate reasoning and ANN structure for incremental learning. The IIS is implemented in software with a robot hand simulation.

Pneumatic Artificial Muscles (PAM) are versatile actuators having many advantages such as high force to weight ratio, soft and flexible structure, extreme safe for human, ease of maintenance and low cost. On the other hand, their inherent nonlinear characteristics yields difficulties in control actions, which is an important factor restricting wide-spread use of PAM. In literature, there are studies to resolve the control issue and their results indicate that there is still requirement for a simple and effective control system. In this work, as a first step of achieving the control goal, three common nonlinear controllers used in literature are selected for an experimental evaluation. The implemented controllers are 'Classical PID controller', 'Fuzzy PID controller' and 'Sliding-Mode Controller' (SMC). The evaluation is performed using a test rig, which is a 1-D robotic arm orthosis actuated by Festo PAMs operated with fast on/off valves. According to experimental results, a model-free Sugeno type combined fuzzy PID controller has yielded most successful performance indicating that it could be a simple and effective solution for PAM control issue.