Tactile sensor provided new function for the robot

  • Time:
  • Click:57
  • source:DALLMAN CNC Machining
8 years ago, researcher rolled out a kind of new-style sensor skill that calls GelSight, it uses the very detailed three-dimensional map that with the physics of the object the contact offers its appearance. Now, go up through installing GelSight sensor the clamping apparatus in robot arm, two groups give a robot taller sensitivity and address degree. The GelSight sensor that receives robot clamping apparatus to go up repeatedly makes the robot can decide its hold the position of a small screwdriver accurately, its from insert a mouth and insert in inserting groove, although clamping apparatus screwdriver of the screen on the camera from the robot. 8 years ago, science of computer of Masschusetts Institute of Technology and artificial intelligence lab (CSAIL) the new-style sensor technology that studies group Ted Adelson rolled out a kind to call GelSight, it and object are contacted, offerred a plan of very detailed 3-D curved surface. Now, go up through installing GelSight sensor the clamping apparatus in machine arm, two MIT group gives a robot greater sensitivity and address degree. Researcher published two articles on international robot and automation congress. In a paper, adelson group uses the data of GelSight sensor, make the robot can judge the hardness of the surface that its contact, a crucial ability, if family expenses robot handles every day object. On the other hand, the Robot Locomotion Group of the Russ Tedrake of CSTE uses GelSight sensor, make the robot can operate smaller than before object. In some way, gelSight sensor is the solution that a low technology solves difficult problem. It is comprised by a transparent balata, of its name " gel " , one of face besmear have metallic lacquer. When lacquer face is pressed on the object, it accords with the appearance of the object. Metallic lacquer makes the surface of the object is reflexed, because the geometrical appearance of algorithm of this computer vision becomes more easy. Installing what go up in as opposite as face of lacquer of oak blob of viscose sensor is trichromatic the lamp and single watch for an opportunity. Have the lampion of different point of view, have this kind to glance next material, through examining color, the computer can find out the three-dimensional appearance of this thing. In two groups of experiments, install GelSight sensor a side in robot clamp, this device is similar to the head of forceps a bit, but the clamp face that has evenness is not most advanced. One, perception of contact dot touch still can help a robot distinguish see the object that is like likeness. In the job previously, the robot tries to come up to evaluate the hardness of the object gently through putting them in flat surface, and they see gently stand sth on end how much did they give. But this is not the main way that the mankind measures hardness. Contrary, our judgement is the rate that the osculatory area between the finger that is based on object and us changes as our pressure it seems that. Softer object often can become evener, increase interface to accumulate. The researcher of experimental process Masschusetts Institute of Technology uses same method: Use candied mould to create object of 400 groups of silica gel, every groups 16 objects. In every groups in, the object has same figure, but hardness is different, yuan uses standard industry dimensions to measure. Next, the hand is moved press the GelSight sensor of every object, recorded the case that osculatory mode changes along with time, basically produce a short film for every object. Hold the bulk that data can manage to standardize data format, she extracted 5 frame from inside every film, time is even, described pressed boy or girl friend be out of shape. Finally, provide data nerve network, search the dependency between the change that osculatory mode and hardness measure automatically. Gets system regards an input as video frame, produce hardness to notch with very tall precision. 2, block up view normally, own robot will use system of vision of some kind of computer to coach its are opposite of the object in the environment operate. Such system can provide the very reliable information of the position about the object, pick up fetch body up till the robot. Especially if the object is very small, a lot of things will be clipped by the clincher of the robot, make positional estimation more difficult. Accordingly, in the robot accurately need knows the place of the position of the object, its estimation becomes fluky. GelSight itself is based on namely photograph those who resemble a head, the data that so its data output compares other touch sensor more undertake with visual data easily compositive. In the experiment, the robot of the clamping apparatus of GelSight must capture a small screwdriver, cover its from the skin in be taken out and its go back. Of course, the data of GelSight sensor did not describe whole screwdriver, it is one fraction only. But, want visual system to go to a few centimeters to the closeness in estimation of the initiative position of screwdriver only, algorithm can conclude which one part of the screwdriver that gives GelSight sensor bring into contact with, decide screwdriver thereby the position in robot hand. CNC Milling CNC Machining