Toshiaki Iwata

Fuzzy control using neural network techniques

A method of fuzzy control using a multilayer neural network which learns fuzzy rules using the error backpropagation algorithm is proposed. To demonstrate the method, a motor servo control was simulated to confirm that tracking could be conducted. The authors also investigated the relationship between fuzzy rule number or effect of learning rules and output using a three-dimensional output expression. The more the network learns, the clearer the undulation is, but the number of rules which were learned does not affect the input-output relationship seriously if rules express a similar relationship. This system was compared with an ordinary fuzzy control method presented by E.H. Mamdani (1976). In this case, the input-output relationship is rugged. It is pointed out that one of the advantages of using a neural network is insensitivity to damage. It was found that if certain important connections are severed, the effects are critical

Smart end effector for dexterous manipulation in space

A smart end effector has been developed to add dexterous and flexible capabilities to coarse space work systems. It provides fine adjustment for precise error compensation by using a relative proximity sensor and provides delicate force control by using a force-torque sensor. It also automatically tracks a marked target and captures it with specified impedance. A new mechanism using parallel links has been devised for the end effector, and a visual proximity sensor has been developed for six-dimensional position-attitude measurement of a flying target. Attaching the smart end effector to a long arm, experiments have been carried out using a two-dimensional air-bearing test bed, and sufficient performance has been achieved.

Remote Synchronization Simulation of Onboard Crystal Oscillator for QZSS Using L1/L2/L5 Signals for Error Adjustment

A new error adjustment method for remote synchronization of the onboard crystal oscillator for the quasi-zenith satellite system (QZSS) using three different frequency positioning signals (L1/L2/L5) is proposed. The error adjustment method that uses L1/L2 positioning signals was demonstrated in the past. In both methods, the frequency-dependent part and the frequency-independent part were considered separately, and the total time information delay was estimated. By adopting L1/L2/L5, synchronization was improved by approximately 15% compared with that using L1/L2 and approximately 10% compared with that using L1/L5 and a synchronization error of less than 0.77 nanosecond was realized.

Robot control strategy for in-orbit assembly of a micro satellite

We have developed novel heuristic control strategies for fitting parts with almost no clearance and also dealing with flexible objects using visual and force feedback. We have applied these control strategies for micro satellite assembly tasks of a ground research model of an in-orbit maintenance robot system. For the in-orbit maintenance system, micro satellites should be modularized as much as possible. The in-orbit maintenance system needs to assemble the modularized parts into micro satellites. We have developed a ground research model including a two-armed robotic platform and modularized micro satellite models. To assemble a micro satellite model, two problems arise: one is to engage two parts into a tight fit and the other is to handle flexible parts. We use a heuristic approach to solve the first problem — to grope one part to find the entrance of engagement of the other and to wobble this part to make a tight fit. For the second problem, visual measurement of the parts is used to position the end-effector of a robot arm and also active limp control is extensively used to adjust any misalignment that arises from the visual measurement error. With the combination of the heuristic, visual and active limp control, the system can successfully assemble a micro satellite.

Maneuvering And Manipulation Of Flying Space Telerobotics System

A free flying telerobotics system has great potential for servicing future large unmanned space facilities because of its flexible maneuver- ability in space. We have been developing a ground model of the Astronaut Reference Flying Robot performing in-orbit servicing in the manner of an a stronaut with This paper deals with tonomous flying robot control structure of a duced for autonomous for a dynamic flying a manned maneuvering unit. control problems of the au- in a no-gravity field. The multilayered system is intro- operation. A control method maneuver in the proximity of the space facility using non-linear compensation is proposed. Moreover, simplified manipulation methods using inertial sensor compensation and ac- tive limp control are proposed for capturing a pay- load, and the effectiveness is shown through numer- ical simulation and experiments on an air bearing testbed. The demonstration experiment for the au- tonomous satellite retrieval mission was success- fully implemented on the testbed. arm using the inverse dynamics method (l, 21. Several studies on arm slew control under attitude disturbance due to arm motion were carried out using the kinematics of a base free robot (3,4). These methods require accurate knowledge of the mass/inertia property of a robot, and complex computation. Furthermore, the torque control of the actuator and thrust force of gas jet thrusters are necessary in the case of inverse dynamics control. These are not realistic for practical applications in space. Moreover, there are few papers discussing other important maneuvers before and after arm slew control required for in-orbit servicing. It is important to study control problems for overall task achievement including the flying maneuver and manipulation with momentum exchange according to practical space missions. In this paper, we deal with a series of control problems in a satellite retrieval mission which requires flying approach, capturing, docking and berthing. We propose new simplified algorithms for the maneuvers which accommodate space applications. We have been developing an experiment model of a free flying telerobotics system with a triple arm and propulsion system, working on a two-dimensional air bearing testbed. This paper presents the high-level control architecture of the robot, and presents the flying maneuver and manipulation methods under a no-gravity field. Robust coordinate control between the

Graphic-simulator-augmented teleoperation system for space applications

A teleoperation system augmented by a real-time graphic simulator has been proposed for efficient operation of a space telerobot and the laboratory model has been developed. In this system, an operator teaches task sequences to a slave arm by driving the three-dimensional graphic image of the arm on the simulator display, using a master arm. The sequences are stored and edited interactively, and the modified trajectories are transmitted to the slave arm. Flexible teleoperation is realized by forward/reverse reproduction of the path at various speeds, either on-line or off-line to the slave arm. In the simulator, arm motion and interaction with the work environment are computed in real time, and the animation of the task and the pseudoforces are fed back to the operator. This simulator is integrated with a slave-manipulator system designed for a vacuum environment and a mastermanipulator system with a universal hand controller for space applications.

Prototype of an end-effector for a space inspection robot

Satellites are exposed to severe space environments such as cosmic radiation, wide temperature changes, vacuums, atomic particles and space debris. Thus sometimes encounter unexpected problems in space. Therefore, a space maintenance platform with a robot is expected to investigate the anomaly and repair the failure module of the satellite in space. We are studying the platform system that has functions for assembling, inspection and diagnosis of small satellites in orbit [1]. In this system, an end-effector of the robot arm requires various functions to provide such capabilities. On the Engineering Test Satellite VII, the fundamental functions of an end-effector have been demonstrated and evaluated in space for simple tasks such as handling an orbit replaceable unit (ORU) [2]. To perform more complex tasks for in-orbit maintenance and to increase automated operation, more sophisticated mechanisms, functions and reliabilities are required for the end-effector. Moreover, extension interfaces and functions for special inspection / diagnosis sensors are needed for the end-effector. This article describes the design and development of the end-effector for a ground test-bed of on inspection and diagnosis space robot boarded on the space maintenance platform.

Phase error reduction method for free-run QZSS clock

A method for limiting the phase error of the remote time keeping system, RTKS, clock for the Japanese quasi-zenith satellite system (QZSS), is presented. To provide a proper positioning signal, QZSS satellites need stable on-board time references. Instead of using atomic references, the recently proposed RTKS employs a remote synchronization scheme that provides an opportune synchronization/correction signal able to keep a master time reference, located on the ground, and the QZSS satellite on-board time reference constantly in lock step. One of the critic issues regarding this architecture is the loss of synchronization due to satellite communication interruption. The proposed method consists of a learning algorithm that monitors the on-board clock behavior during its regular functioning. Consequently, when synchronization becomes unavailable the QZSS onboard clock phase/frequency drift is kept contained by using a consecutive estimation of the clock phase error. The proposed system is particularly suitable for the RTKS for QZSS and is characterized by a low hardware requirement profile, particularly suitable for the RTKS satellite payload.

DIAGNOSIS OF SMALL SATELLITES USING MULTIPLE SENSORS AND A ROBOT

Four diagnostic methods for small satellite malfunction using several types of sensors and a robot are proposed and tested experimentally. The four methods are (1) averaging method of probability, (2) small probability deleting method, (3) Bayesian network method and (4) correlation matrix method. The reasoning method of the malfunction essentially has a probabilistic property. We limit the malfunction of the solar paddle to five kinds and use symptom parameters to identify the malfunction by inference. In our experiment, all four methods are effective. In particular, Bayesian network method produces excellent results; however, in this case a single failure may cause total failure. The averaging method and correlation matrix method generate rather ambiguous results; however, they are hardly affected by failure. In many cases, the small probability deleting method works well, however, a single failure may cause total failure.

Remote Synchronization Experiments for Quasi-Zenith Satellite System Using Multiple Navigation Signals as Feedback Control

The remote synchronization system for the onboard crystal oscillator (RESSOX) is a remote control method that permits synchronization between a ground station atomic clock and Japanese quasi-zenith satellite system (QZSS) crystal oscillators. To realize the RESSOX of the QZSS, the utilization of navigation signals of QZSS for feedback control is an important issue. Since QZSS transmits seven navigation signals (L1C/A, L1CP, L1CD, L2CM, L2CL, L5Q, and L5I), all combinations of these signals should be evaluated. First, the RESSOX algorithm will be introduced. Next, experimental performance will be demonstrated. If only a single signal is available, ionospheric delay should be input from external measurements. If multiple frequency signals are available, any combination, except for L2 and L5, gives good performance with synchronization error being within two nanoseconds that of RESSOX. The combination of L1CD and L5Q gives the best synchronization performance (synchronization error within 1.14 ns). Finally, in the discussion, comparisons of long-duration performance, computer simulation, and sampling number used in feedback control are considered. Although experimental results do not correspond to the simulation results, the tendencies are similar. For the overlapping Allan deviation of long duration, the stability of at 100,160 s is obtained.