In this article, we will develop a realtime vision system based on FPGA's and will evaluate the performance of the developed FPGA-based vision system experimentally. First, we will describe the developed FPGA-based vision system. Secondly, we will implement three vision algorithms on the vision system: computation of the image gravity center, detection of object orientation using a radial projection, and the computation of Hough transform. We will then execute these algorithms on the FPGA-based realtime vision system.
The task understanding is completely examined for a task of crank turning by a robot manipulator. A task model embedded uncertain factors as unknown parameters is constructed. Analyzing its task model, we propose a robust task strategy turning a crank under the uncertain task conditions and a motion/force/impedance control scheme to execute the task strategy. Moreover, a task strategy with the estimation of the pivot position is addressed to realize the smoother crank turning even if the initial estimation error of the pivot position is considerably large. These validities are experimentally investigated by a 6 D.O.F industrial manipulator.
From the viewpoint of the Affordance Theory advocated by J. J. Gibson, an ecological approach for realizing artificial skills is presented in contrast to the conventional paradigm of behavior-based robots known as “sense, think and act”.A perception-driven robot demonstrates the usefulness of the ecological approach through the assembly task of mating several parts on a plane. Instead of action-rules based on sensory information, the proposed robot employs a kind of oscillator, Action Pattern Generator (APG), for taking active action before perception. The limitations of the simple robot, the actions of which are dominated by APG, are also discussed. Then, in order to compensate APG for lack of the rationality in choosing actions, two types of memories, namely Working Memory (WM) and Episodic Memory (EM), are introduced into the robot's brain. Consequently, the perception-driven robot has the abilities to distinguish the shapes of various parts, mount them in right places, and stop its motion automatically.
This paper presents a vision-speech system for service robots that can learn the user's customs and objects fixed in the environment while helping the user, and can perform their tasks more efficiently with less user's burden. We are working on a service robot that brings objects ordered by the user through speech. The robot needs vision to recognize the objects. It asks the user for help by speech if its vision fails. In early stages, the robot asks the user for help many times and it takes long time to detect the objects with vision, because the robot does not know much about the environment. We assume that humans usually put a particular object on a small number of places. The robot obtains such information and the environmental knowledge while accomplishing the user's demands. In later stages, the user does not need to say details because the robot knows where the objects usually are. Moreover, the vision processing time is greatly reduced because it knows what operations can work in such cases. Experiments using a robot system show the usefulness of the proposed system.
This paper proposes a gradual formation of a spatial pattern for a homogeneous system. The autonomous formation of spatial pattern is important for the advancement of cooperative robotic systems because pattern formation can be regarded as function differentiation of multi-agent system. When multiple autonomous robots cooperatively work on an assigned local task for a global objective like a group of insects or animals, the function differentiation is the first step and indispensable. There were a lot of papers that reported spatial pattern formation of multiple robots, but global information was supposed to be available in their approaches. It is, however, almost impossible for a small robot to be equipped with an advanced sensing system for global information because of robot's scale and sensor. The algorism for complicated pattern formation is needed even if each robot is not so comparative. We therefore propose a gradual pattern formation algorithm, i.e., the group of robots changes their pattern to goal pattern through a simple pattern like a line or a circle. Then we use the Turing Instability theory based only on local information for polygon pattern formation from circle pattern, and experimentally show the formation of several patterns with multiple autonomous robots.
We develop a cyclogyro-based flying robot with a new variable attack angle mechanism. Cyclogyro is a flying machine which is supported in the air by power-driven rotors, which rotate about a horizontal axis, like the paddlewheels of a steamboat. Machines of this type have been designed by some companies, but there is no record of any successful flights. The proposed new variable attack angle mechanism has an eccentric point that is different from a rotational point. The main feature of the mechanism is to be able to change attack of angles according to the rotational angles of the cyclogyro without actuators. The lift and drag forces of the robot are calculated and measured through simulations and experiments. The simulation and experimental results show that the new mechanism generates positive lift forces at any rotational angles of the cyclogyro.
It is important to save the dissipated energy even of a manipulator for suppressing the exhaust gas of CO2. This paper discusses the design method for the counterbalances of the links in order to save the dissipated energy of a vertically articulated manipulator with three-degree-of-freedom. If the gravity forces on the second and the third link are cancelled by the counterbalances, the minimum dissipated energy of a manipulator can be represented by a function of the inertia matrix and the movement angles in a given operating time. A new criterion that is the minimum dissipated energy normalized by the movement angles is proposed in order to design optimum counterbalances. In a cylindrical counterbalance, the optimum aspect ratio and the optimum position are obtained by minimizing the criterion. It is seen from a simulation that the optimized counterbalances make the dissipated energy almost minimal value in the various movement angles. The manipulator with the counterbalances can remarkably decrease the dissipated energy in comparison to that without counterbalances with the exception of a very quick motion.
An omni-directional wheelchair is highly maneuverable in narrow or crowded areas such as residences, offices and hospitals. However, with the increase in the degrees of freedom accompanying omni-directional wheelchairs, it has become more important than in the past to develop a smooth and vibrationless motion control system. In this paper, a novel motion control method for an omnidirectional wheelchair considering the suppression of vibration of both the wheelchair and the patient is proposed. In the proposed control system, the vibration of the wheelchair is suppressed by notching the frequency characteristics at the natural frequency of both the wheelchair and human's organs. The designed controller satisfies various control specifications such as settling time and stability by using feedback from only the wheelchair's position data. The effectiveness of the proposed system is evaluated by the output signal of a sensor attached to the wheelchair and by a descriptive inspection by several users.
This paper presents a method to update running RTLinux robot controller programs (RTU) on the fly. The proposed method exploits thread swapping. This paper focuses on the safety conditions of thread swapping, platform requirements, realtime issues, and uncertainties in RTU completion time. This paper introduces RTU-able software design by layered modules which enable the update of servo functions and robot model global variables. We have developed a RTU-able mobile manipulator controller, evaluated its performance by worst case results, and shown its practical applicability.
This paper describes a highly general discussion of the classification of the function of a quadruped walking robot, and necessity number of active degrees of freedom (DOF) for each function is revealed. Specific number of active DOF is shown for reduced function quadruped robots. For example, leg function of forward/backward/curve/turn on uneven terrain requires 2 DOF. Omni-directional walk on uneven terrain realizes by 3 DOF as for the leg function. Additional number of DOF contributes to control internal forces of the legs. Controlling internal pushing/pulling forces between 4 supporting legs requires 3 additional DOF on 3 DOF of swing leg. As for the body movement, controlling a center of the gravity (CG) in forward/backward, sideways, vertical directions requires 1 DOF individually. Body attitude control of two horizontal axes, and vertical axis requires 1 DOF each. Popular quadruped robot with 12 DOF is the one that have all above DOF. On the second, concrete mechanisms for reduced DOF quadruped are shown. A 3 DOF mechanisms for forward/backward/curve/turn locomotion on uneven terrain, and a 4 DOF one for omni-directional locomotion are designed. Latter one follows a more easily realistic 5 DOF design.A gait generating algorithm for reduced function quadruped is also shown. Finally, the experiments of trial manufactured 3 DOF and 5 DOF quadrupeds show a validity of discussions.