機械手控制系統(tǒng)設計

機械手控制系統(tǒng)設計,機械手控制系統(tǒng)設計,機械手,控制系統(tǒng),設計 1 Mechanical manipulator and control system 1. Introduction In June 2001 in Karlsruhe, Germany launched a humanoid robot special study is to develop in a normal environment (such as the kitchen or living room) under the able and human cooperation and interaction of the robot system. Design of these robotic systems is to be in non-professional, non-industrial conditions (such as live among many things), help us to capture different size, shape and weight of the object. At the same time, they must be able to be grasped object manipulation. This extreme flexibility only through a highly adaptive robot grasping system to get the so-called multi-fingered robot hand or robot. The above-mentioned research project is to create a humanoid robot, the robot will be equipped with such a robot hand system. The novice will be manufacturing the two agencies, they are the University of Karlsruhe, IPR (Institute for Process Control and Robotics) and c (Institute of Computer Application). The two organizations have the relevant experience to build such systems, but slightly different point of view. IPR Karlsruhe Dexterous Hand-made (Figure 1), is an independent Gripper four fingers, we described in detail in this article. IAI-made hand (Figure 17) is a disabled limb. IPR Karlsruhe Dexterous Hand IAI developed fluid hand 2. The general structure of the robot hand A robot hand can be divided into two main subsystems: mechanical systems and control systems. Mechanical systems can be divided into structural design, drive systems and sensor systems, we will be further introduced in the third part. In the fourth section describes the control system at least by the control hardware and control software. We will issue these two subsystems some basic introduction, and then show you Karlsruhe dextrous hand. 2 3. Mechanical systems Mechanical systems will describe how this hand looks and what elements from the composition. It decided to structure design, the number of fingers and the use of materials. In addition, determine the drive (such as motors), sensors (such as the position encoder position). 3.1 Structure Design Structural design of flexible robot will be very useful, that is, what type of objects it can crawl and grasp objects can be what to do. When designing a robot hand, we must determine the three basic elements: the number of fingers, finger joints and finger size and the number of placement location. To be able to work in the context of robot crawling and operational security objects, at least three fingers. To be able to operate on the objects being arrested for 6 degrees of freedom (three translational and three rotational degrees of freedom), each finger should have three separate joints. This approach was first used over Karlsruhe Dexterous hands. However, in order to grasp a heavy object without releasing it first and then pick up the case, at least four fingers. To determine the location of finger size and placement, can be used two ways: anthropomorphic and non-anthropomorphic. Then depends on the object to be operated and the choice of the desired type of operation. To place a personification is easy to shift from manual to hand crawling robot intent. But each finger a different size and location of asymmetric placement will increase processing costs, and a control system becomes more complex, because each finger must be controlled separately. Symmetrical arrangement for the same finger, often non-anthropomorphic method. For processing and building only a single finger module, thus reducing processing costs, but also simplify the control system. 3.2 Drive System Knuckle drive flexibility opponents have a great impact, because it determines the potential strength, accuracy and speed of movement joints. Mechanical movement of the two aspects need to be considered: the source and the movement direction of movement. In this regard, the literature describes the several different ways, such as 3, said hydraulic cylinder or pneumatic cylinder may be in motion, or, as in most cases use the same motor. In most cases, the motion drivers (such as the motor) are too big and not directly with the corresponding knuckles together, therefore, the campaign must drive (usually located at the connection point of the final arm) transferred. There are several different ways to achieve this sport, such as the use of keys, belts, and activity axis. Using this indirect method of driving knuckles, more or less reduces the strength and precision of the entire system, but also to control the system complicated, because each finger of the finger joints often are mechanically connected together, but in control burst into the software system to control them separately. Due to these shortcomings, the movement of small knuckle drive with a direct integration becomes quite necessary. 3.3 Sensing System Robots sensor system can transmit the feedback information from the hardware 3 control software. Grasp objects on the finger or a closed-loop control is necessary. Use the machine in the hands of 3 types of sensors: 1. Gripper state sensor to determine the location of knuckles and fingers and finger force on the situation. Know the exact location of fingertip control will enable precision made possible. In addition, the role that finger was caught in the force object, you can capture the fragile object without breaking it. 2. Crawl status sensor provides an object caught between the fingers and the contact status information. This tactile information can be identified in the crawling process in a timely manner and location of the object point of first contact, and to avoid incorrect crawl, such as catch the edge and cutting-edge object. In addition to the already grasped object can detect whether the fall in order to avoid falling objects because of damage. 3. Object state or attitude sensor used to determine the shape of the object within the fingers, position and direction. If you grab objects such information before the case is not clear, this sensor is very necessary. If the sensor can act on the objects have been arrested, then it can control the objects attitude (position and orientation), in order to monitor whether the decline. Depending on the drive system, the geometric mean position of a joint information campaign can drive or directly measured in the joints out. For example, if the motor and that there is a rigid coupling between the joints, then you can use a motor shaft angle encoder (in gear before or after gear) to measure joint position. But if this is not enough or the coupling stiffness to obtain very high precision, then this method can not be used. 3.4 The Karlsruhe dextrous hand mechanical systems In order to get more complicated such as heavy grasping operation, Karlsruhe dextrous hand (KDH ) formed by the four fingers, and each finger from the three independent joint composition. The design of the hand to be able to apply in the industrial environment (Figure 3) and control boxes, jars and screw nuts objects. Therefore, we used the same four fingers, will they be symmetric, non-anthropomorphic configuration, and each finger can rotate 90 (Figure 4). In view of Karlsruhe dextrous hand from the first generation design of the experience, such as belt caused by mechanical problems and a larger control problems caused by friction factor, Karlsruhe Dexterous Hand with a number of different design decisions. Each finger joint 2 and joint 3 between the DC motor is integrated into the finger in front of the body (Figure 5). This arrangement can be used very hard ball shaft gear will be delivered to the finger movement of the joints. At the motor shaft angle encoder (in gear before) at this time as a state of high precision position sensor. 4 Figure 3.Industrial robot on the KDH 3. Control System Robot hand control system to determine which potential can be actually used dexterous skills, these skills are provided by the mechanical system. As mentioned earlier, the control system can be divided into control computer that is hardware and the software control algorithms. Control system must meet the following conditions: 1. Must have sufficient input and output ports. For example, a junior with nine degrees of freedom hand, at least 9 of its drive analog output port, and have nine road from the perspective of the encoder input port. In addition to the force of each finger on the sensor, tactile sensor and the object state of the sensor, then the port number will increase the number of times. 2. Need to have fast real-time response to external events capacity. For example, when the detected objects fall, they can immediately take appropriate measures. 3. Need to have a higher capacity to respond to a number of different computing tasks. If more than that and objects can be executed in parallel path planning, coordinate transformation and the closed-loop control tasks. 4. Control system is smaller in size so that they can be integrated directly into the operating system. 5. In the control system and driver and the sensor must be connected to electrical short. In particular, the sensor is, if not, a lot of jamming signal would interfere with sensor signal. 4.1 Control Hardware In response to system requirements, control hardware generally distributed in several specialized processors. Such as through a simple processing very low-end microcontroller input output interface (motors and sensors), so the controller size is small, can be easily integrated into control systems. But the high level of control port will require higher computing power, and need a flexible real-time operating system support. This can be easily resolved by PC machine. Therefore, controlling the hardware usually consists of a non-uniform distributed computer system components, and its end is the micro-controller, while the other end is a powerful processor. Different computing units are connected by a communication system, such as bus system. 4.2 Control Software Robot hand control software is quite complex. Must be parallel to the fingers in real time and control, but also plan your fingers and objects of the new track. Therefore, in order to reduce the complexity of the problem, it is necessary to bring the question into several sub-problems to deal with. On the other hand involves software. Robot Hand is a research project, its programming environment such as the user interface, programming tools and debugging facilities have to be very strong and flexible. These can only use a 5 standard operating system can be met. Among commonly used in mechanical hierarchical control system after pruning methods to meet the special control requirements manipulator. 4.3 The Karlsruhe dextrous hand control system As mentioned in section 4.1, the Karlsruhe Dexterous Hand , control hardware, using a distributed approach (Figure 7). A microcontroller to control a finger each driver and sensor, a microcontroller is used to control other objects state sensors (laser triangulation sensors). These microcontrollers (Figure 7 left and right outer box) directly on the hand, so it can guarantee and a short drive and the electrical connection between sensors. These microcontrollers are using a serial bus system and the host computer is linked. The host computer (Figure 7, Figure 8 in the gray box) is composed of six industry computer to a parallel computer. These computers are arranged in a two-dimensional plane. Adjacent to the computer module (a computer up to eight adjacent blocks) using a dual-port RAM for fast communications (Figure 7, dark gray box below). A computer used to control a finger. Another state of sensors used to control the object and calculate the position between objects. The remaining computers were mentioned in front of computer security around. The computer used to coordinate the entire control system. Reflected in the structure of control software to control hardware architecture. Shown in Figure 9. Figure 7. KDH II control hardware architecture Figure 8. Control KDH II parallel the main computer One on the hand control system of the three highest levels of online programs are being planned. Ideal object displacement command superior robot control system can be obtained and used as objects, the precise path planning. Have been generated according to the target path planning can be a viable capture behavior (finger on the possible role of the object crawling position). Now know that the object of the exercise program, you can draw from the finger path planning trajectory of each finger, and passed to the system part of the real-time capabilities. If an object is grabbed, then the movement path of their fingers on the object passed to the state controller. The object of the attitude controller, which consists of fingers and objects determined by the state sensors to get the object profile. If a finger is not with the objects in contact, its moving path will be directly passed to 6 the hand controller. The hand controller will pass the relevant location is expected to finger the finger to all controllers to coordinate the movement of all fingers. The finger sensor in turn, drive the help of finger drive. Figure 9. KDH hand control system 5. The experimental results To test the ability Karlsruhe dextrous hand, we have chosen two requests operational problems. One problem is the Internet on in the external objects caught under the influence of posture (position and direction) control. Another issue is to be grasped object must be able to rotate around any angle, this can only be achieved through re-arrested. This may reflect the Karlsruhe Dexterous Hand on the complex task of operating capacity. 5.1 Attitude Control objects The purpose of the object attitude controller is good to be grasped object in order to determine the location and orientation to fit a given trajectory. This task must be in terms of online access to real-time, despite the internal changes and external disturbances exist. Changes such as moving the object within the process, the spherical object on the finger was caught in the rolling. The situation shown in Figure 10, as shown in Figure 11. This will lead to unnecessary additional objects move and tilt. These errors is difficult to estimate in advance the position of objects. Therefore, the object sensor input state must modify these errors. For the Karlsruhe Dexterous Hand , its the three laser triangulation sensor is used to correct such errors. Figure 12 shows quantitatively in the absence of objects in Figure 9 under the tilt attitude control the situation. The following figure shows the X direction in the expected trajectory over time, while the figure shows the actual objects in the rotation (tilt) results of the situation. State control because the object is enabled, Figure 13 objects in the tilt has been greatly reduced. Objects on the map rotation to keep almost constant, which, and expect the same. Figure 10. By rolling the extra displacement Figure 12. There is no state control of 7 the object tilt Figure 11. Because of spherical objects rolling fingertips in Figure 13 arise. Objects to reduce object under state control Do not expect additional cases of inclined slope Object state interference compensation controller is also very necessary. For example, the robot (arm, hand or finger) or grasping objects, collision with the outside world may lead to fall of objects. This is more likely to result in the loss to be grasped object, which can not happen. In order to avoid an object in this case the loss, we must detect the object slide and quickly take action to stabilize the object state. In order to verify the Karlsruhe Dexterous Hand control system processing power of this interference, we did the following experiment: After the object was captured, the finger contact force constant decrease until the object began to fall. In the laser triangulation sensors detect slip, the object state controller to take measures to control the object back to the desired location. Figure 14 and Figure 15 shows an example of such experiments. In particular, Figure 14, it shows an object fall to start very suddenly and quite fast. However, the state controller object can be detected quickly enough and compensation fall, so the location of objects (here: in particular, X direction is the direction of decline) and orientation of objects able to start with the most consistent expectations quickly. 5.2 re-arrested Although the Karlsruhe Dexterous Hand is flexible, but it can not be the first operation can be an ideal object for each control. This stems from the fact that: finger relative to the normal industrial robot is very small, so the scope of work have also very limited. If the object is fingers caught it the first time all the fingers can only be manipulated within the remaining space. Feasible operating conditions are all located in the contact points must be long-term joint refers to the phase scope of work. This greatly limits the feasibility of operation. In order to overcome such limitations, an operation called a re-grasp on the need to implement. That is reached when a similar joint contact point refers to the restricted areas, the finger must be off from objects, and move to a new contact position. This must be more than three fingers of the hand to enable reliable operation. Cyclical movement of the fingers, so that any operation 8 can become feasible. An example of such operations on that was caught in the wide-angle rotating object, then re-grasp action is necessary. Figure 16 shows the Karlsruhe dextrous hand in a nut-like objects rotating series of pictures. This object is rotating around its vertical axis of. In a to c graph of all the fingers are related to objects in contact, and only four fingers coordinated campaign to object rotation. Figure d to Figure f shows a heavy finger grasping movements. In d the graph of this finger has been movement to limit the scope of their work location, then the coordination of all the finger movement is terminated. The fingers of the left front and separate from the object to another point of contact. In the finger in Figure f re-engagement with the objects, the other fingers can be repositioned at this time (not shown). After all the finger re-positioning, coordinate rotation to continue. As the case may be, Karlsruhe dexterous hand can simultaneously re-arrested a few finger movements. This can speed up the process of re-arrested, but can only be grasped object in contact with the outside world is possible under the conditions. For example, the nut screw or nail holes in a hook. Figure 17 shows the Karlsruhe dextrous hand to a stake from the base of a square hole pull out of a series of pictures. Figure a Figure b shows wooden posts to be pulled out half, then left and right means that from the object at the same time and re-positioning (Figure c to graph e). After that, in front of and behind the fingers re-positioning (Figure f). After that, the wooden pillars were pulled out, thereby further operation (not shown). Figure 16. Use of heavy rotating nut-like object grasping Figure 17. Use of re-grasp pull out from the hole in the wooden pillars 6. Conclusion In order to complete the flexible manipulator precise operation, a suitable mechanical system and control system is required. The introduction of standards is necessary to be considered, as the paper said. Karlsruhe dextrous hand very successful performance. This robot can grab a wide range of different shapes, size and weight of the object. Attitude object was captured can be controlled reliably, even in the case of external disturbances. In addition, this system, a complex fine manipulation (such as re-arrested) can be achieved. Peoples Bank of robots in the field of special study, based on a concept called the fluid of different (Figure 2), based on a small robot also 9 has to be people and machines. This concept is the IAI by the Karlsruhe Research Cen