Machine learning has been studied for more than 20 years. Most of the recent research contributions are basic and come from the areas of Cognitive Science and Artificial Intelligence. The classical domain of machine l...
Machine learning has been studied for more than 20 years. Most of the recent research contributions are basic and come from the areas of Cognitive Science and Artificial Intelligence. The classical domain of machine learning is related to vision and image analysis and image description. Since about five years there has been encreased activity to apply and to extend these general learning methods to really complex systems such as autonomous robots. The purpose of applied machine learning methods is to facilitate and to aid automatic construction and generation of knowledge based systems and to increase the system performance with time. Autonomous robot systems will operate in an unknown real environment and are based on the interaction of the basic system components such as action planner, plan executive and action monitor which need knowledge about the real world and knowledge about how to behave to reach a given goal. A robot system can learn if it is able to make changes and extensions in its behavioural control structure that enables it to better perform a given task. This includes acquisition of declarative procedural knowledge, the developement of manipulative and cognitive skills through instruction and practice as well as the organisation of knowledge into efficient representations. The purpose of automatic self-configurating or self-modifying representations is to make experience based knowledge available for possible future use in the context of increasing robot system performance. This paper reviews some general methods of machine learning application and some first results achieved to date.
The Karlsruhe Hand is a non-anthropomorphic three finger dextrous gripper with 9 degrees of freedom, which is being developed at the institute for real-Time computercontrol Systems and robotics of the University of K...
详细信息
The Karlsruhe Hand is a non-anthropomorphic three finger dextrous gripper with 9 degrees of freedom, which is being developed at the institute for real-Time computercontrol Systems and robotics of the University of Karlsruhe.
This paper presents ROSI-Layout, a computer aided planning system for the generation of suitable workcell layouts for robot assembly applications. The idea is to derive first the spatial transformations for the produc...
详细信息
This paper presents ROSI-Layout, a computer aided planning system for the generation of suitable workcell layouts for robot assembly applications. The idea is to derive first the spatial transformations for the product parts, which lead to the assembly of the final product. A global database provides the product data and the precedence graph. This information comes from product design and assembly planning. The design of spatial transformations is derived from constraints given by the geometric shape of the product parts and from rules that define how assembly with robots should be accomplished. This is followed by the generation of a suitable layout. A layout can be developed that is specially designed for the product to be assembled. During the entire planning session, the database manages all planning data and ensures consisteny. ROSI-Layout is an extension of the RObot Simulation system ROSI.
This paper describes a method to simulate vision for use in robot applications. Our approach shows the physical modeling of the camera, the light sources and the objects within the scene based on a geometric model. Ma...
详细信息
This paper describes a method to simulate vision for use in robot applications. Our approach shows the physical modeling of the camera, the light sources and the objects within the scene based on a geometric model. Many parameters of the physical model may be adjusted to obtain simulation results close to reality. For processing of the picture data the SPIDER program is added to the simulation of the camera. Though a complete vision system is created. An example shows the simulation of a CCD-camera and a scene with workpieces taken from the ‘European Benchmark’ [COL85].
The problem of a polygonal object moving in an uncertain polygonal environment, i.e. the positions and orientations of the objects relative to each other are not exactly known, is addressed. Furthermore, the moving ob...
详细信息
The problem of a polygonal object moving in an uncertain polygonal environment, i.e. the positions and orientations of the objects relative to each other are not exactly known, is addressed. Furthermore, the moving object does not move exactly as it is commanded to move. Due to these uncertainties, even a motion generated by a correct motion planner may fail, i.e. for instance, an unexpected contact may occur. A situation analyzer that enables the moving object to acquire information on a contact situation is presented. This analyzer uses the nominal world model including information on the magnitude of the uncertainties in order to get hypotheses on the geometric entities involved in the contact. The discrimination among the hypotheses is based on test motions, i.e. motions whose feasibility depends on the validity of a certain subset of the hypotheses.< >
Deals with the problem of uncertainties in the nominal world model of a robot. In particular, the problem of analyzing unexpected situations in a 2D polygonal world is attacked. Such a situation analysis can be used t...
详细信息
Deals with the problem of uncertainties in the nominal world model of a robot. In particular, the problem of analyzing unexpected situations in a 2D polygonal world is attacked. Such a situation analysis can be used to acquire information on a contact situation encountered during the execution of a motion. The analysis is based on the concept of so-called test motions, i.e. motions, the feasibility of which depends on the actual contact situation present. Using a probabilistic uncertainty model, the focus is on the computation of motion constraints due to contact and the resulting selection of an optimal test motion.< >
F/sub A/TE, an online task-level planning system for an advanced assembly robot consisting of two manipulators and a set of different sensors such as an overhead camera and force-torque sensors, is presented. The plan...
详细信息
F/sub A/TE, an online task-level planning system for an advanced assembly robot consisting of two manipulators and a set of different sensors such as an overhead camera and force-torque sensors, is presented. The planning system takes an implicit description of the assembly task, plans a sequence of explicit robot commands and monitors the execution by the real-time robot control system. Because it runs completely online, the planning process is highly reactive using sensor information about the robot's present environment. This planning integrates as a key feature a dynamic mapping of assembly subtasks ready for execution onto the manipulators available at the moment. The system was implemented in Prolog on a SUN 4/75 SPARCstation running Unix.< >
As an example of an autonomous assembly robot, the mobile two-arm robot system KAMRO, is described. The system is capable of navigating within an assembly cell, of docking at a work table, and of executing an assembly...
详细信息
As an example of an autonomous assembly robot, the mobile two-arm robot system KAMRO, is described. The system is capable of navigating within an assembly cell, of docking at a work table, and of executing an assembly according to a given plan. A series of successful experiments has demonstrated the completely autonomous operation of the system.< >
This paper deals with the two fundamental problems that occur when objects are manipulated with multi-finger robot hands: the determination of the joint motions to perform a manipulation according to a given object tr...
详细信息
This paper deals with the two fundamental problems that occur when objects are manipulated with multi-finger robot hands: the determination of the joint motions to perform a manipulation according to a given object trajectory, and the optimization of the joint torques needed to ensure a stable and secure grip. The consideration of the effect of rolling and slipping of the fingertips at the contact points on the object surface leads to a set of linear differential equations for the joint angles and to a partly nonlinear optimization problem for the joint torques solved by the Hooke-Jeeves algorithm. The removal of redundant information reduces the computational effort to about 40% of the operations required for the standard procedure. Especially, the resulting object motions are demonstrated at an example: the rotation of an ellipsoid object with the fingers of the Karlsruhe dextrous hand.< >
暂无评论