Conventional artificial intelligence and cognitive science approaches have been arguing the emergence of intelligence by completely separating the cognitive agent from its environment. In this paper, we intensively investigate and analyze the role of the system-environment interaction through the comparison between the information processing and embodied cognitive approaches. By taking account of this interaction, we experimentally show some of the hard classification problems suffering from the large input space and the ambiguities due to the perceptual aliasing problem can be greatly alleviated. To demonstrate this concept, we apply to a garbage-collecting task as a practical example.
It is important to save the dissipated energy even of a manipulators for improving the environment of the earth which is warmed up by CO2 gas emitted from thermal power plants. Hence, this paper describes an optimal path and operating time which minimize the dissipated energy in PTP motions of a vertically articulated manipulators. A globally optimal path is roughly estimated so that the heavier link is accelerated toward gravitational direction and decelerated toward anti-gravitation. This proposed path is used as a starting function of the iteration method solving a non-linear two-point boundary-value problem. In order to avoid collision with the obstacle such as a desk laying a handled material, the obtained optimal path is modified by an optimal control theory with state constraints. The simulation results show that a proposed optimal path can reduce the dissipated energy to 1/14 -1/30 compared with conventional path when the operating time is selected so as to be optimal.
This paper describes a teleoperation system for maintenance tasks integrating planning functions based on manipulation skills. Demands for autonomous teleoperation function for maintenance tasks are increasing. We embed planning functions into a telerobotics system to make the system more flexible and robust. A motion teaching system based on contact state transition in a task in VR, a geometric modeling system using Teaching Trees, and a task execution system based on manipulation skills are integrated. The design concept of the system and essential technologies are described. An experimental task is explained to demonstrate the efficiency of the telerobotics system.
We have developed a complete automated vision based bin-picking system including vision system, planning system and manipulator system. The vision system detects and locates 3D circle on the object using stereo ranging technique. The planning system checks collisions between the manipulator and a bin with respect to each detected parts to select most suitable one for picking. Then, the manipulator successfully picks up parts one after another. This paper describes our unique approach to detect parts, picking strategy and experimental results to show its advantages.
This paper presents a new method for mechanical stiffness control of the tendon-driven joints of manipulator. First one proposes a new formula for controlling joint stiffness and joint angle independently. Next, one proposes a new elastic mechanism (Non-Linear Elastic Module, NLEM) of which nonlinear elasticity is easily designed. NLEMs situated between the actuator and the tendon play a key role for mechanical control of the stiffness. Two types of NLEM are designed and evaluated numerically and experimentally. It follows a computer simulation and experiments of the stiffness and joint displacement control, which show a validity of the proposed formula. One also proposes a simple method of vibration suppression of the tendon-driven joint, which makes use of a nonlinear elasticity of NLEMs. It also follows a computer simulation to certify its validity.
Walking robots have high adaptability for terrain variation, and thus, have been expected as effective moving platform on uneven terrain, stairs, forest, marshy surface, and on ice. On the other hand, mobile robots that perform several hazardous tasks such as mine detection or the inspection of an atomic power plant are typically controlled by operators from distant places. For a teleoperation system, use of visual information from a camera mounted on a robot body is very useful. However, unlike wheeled vehicles, the camera mounted on the walking robot oscillates because of the impact of walking, and the obtained unstable images cause inferior operation performance. In this paper, we introduce an image stabilization system for teleoperation of walking robots using a high speed CCD camera and gyrosensors. The image stabilization is executed in two phases, that is, the estimation of the amount of oscillation by the combination of the template matching method and gyrosensors, and change of the display region. Pentium MMX instruction is used for template matching calculation, and the estimated amount of oscillation is given in every 12 [msec] . Furthermore, developed image stabilization mechanism can be used an external attitude sensor from the visual information, and the damping control of the robot body while walking is also possible. Experimental results showed stabilized images that eliminates the oscillation component are taken even when the robot moves dynamically or in long distance, and verified that the performance of attitude control using the developed image stabilization system is almost same as the case using an attitude sensor.
In recent years, rover missions for exploring lunar or planetary surface has been attracting much attention over the world, culminating the success of the Sojourner rover which explored the Martian surface in 1997. This paper proposes a new path planning method for navigation of planetary rovers, the surface environment of which is a natural terrain and described by a digital elevation map (DEM) . Given a DEM, a path from a start point to a goal is calculated based on the simple rover model which consists of three parameters. An obstacle map and a new concept of an extended elevation map defined in three dimensional space are introduced which provides the capability to use the DEM height information in three dimensional configuration space. With this extended elevation map, path planning is conducted by solving an optimization problem. Numerical simulations show the effectiveness of the proposed method.
In assembly operations, there are various constraint states while a moving object is in contact with a fixed object. For different constraint states, the existent regions of contacting forces, geometrical constraints, possible object displacements, which can be expressed by polyhedral convex cones respectively, are different. We consider that the sizes of these regions are important characteristic of contact stability for planning assembly operations. This paper proposes an approach that can evaluate the sizes of these regions. Based on this approach, the magnitudes of form constraint region, applicable equilibrating force region and admissible displacement region are defined to evaluate the contact stability of constraint state. Then, their applications to planning of assembly tasks are discussed.
In this paper, we propose a novel method of supporting a specific user in the network connected multi-robot environment. We detach a robot brain from the robot body and make it an autonomous entity “Agent”. One unique software agent exclusively interacts with the user, obtains the robots dynamically and support the user physically with control of these robots. An agent obtains only necessary robots and when those are not needed, the agent releases the robots to make them available by the other agents. With a personal agent approach, the user gets supported consistently in the long term. An agent gets sensory information from the robots obtained and observes the user continuously and reorganizes the robots to follow the user in the real world. To reduce traffic and delay on the network, an agent traverses the network and always resides in the center of the controlled robots. We realized this robot environment using the distributed objects technology and the mobile agent technology. We describe the validity of the proposed method, the detailed mechanism and the experimental results using real robots.
This paper proposes multi-layered reinforcement learning by which the control structure can be decomposed into smaller transportable chunks and therefore previously learned knowledge can be applied to related tasks in a newly encountered situations. The modules in the lower networks are organized as experts to move into different categories of sensor output regions and to learn lower level behaviors using motor commands. In the meantime, the modules in the higher networks are organized as experts which learn higher level behavior using lower modules. We apply the method to a simple soccer situation in the context of RoboCup, show the experimental results, and give a discussion.
In order to keep visual tracking systems with color segmentation technique running in real environment, it should be developed on-line learning method to update models for adapting them to dynamic changes of surroundings. To deal with this problem, we propose an on-line visual learning method for color image segmentation and object tracking in dynamic environment. Our method utilizes Fuzzy ART architecture which is a kind of neural network for competitive learning. The mechanism of this neural network is suitable for on-line learning and different from that of backpropagation type neural network. In order to use Fuzzy ART architecture for color segmentation on-line, we transform the color signal that the framegrabber used yields to a particular color space called Yrθ space. To show validity of our method, we present some results of experiments using sequences of real images.