08/02/2012 Author: Daniel Greenwald Reviewers: Hugo Elias, Rich Walker, Jan Gries, Jackie Duggan 1/13
Table of Contents 1 Introduction...2 2 Experiments...2 2.1 Opening and Closing a Plastic Jar...3 2.2 Opening and Closing a Sample Tube...4 2.3 Opening and Closing a Vial...4 2.4 Operating a Micro-Pipette...5 2.5 Using a Syringe...5 3 Results...6 3.1 Plastic Jar...6 3.2 Sample Tube...7 3.2.1 Mounted in a vice...7 3.2.2 In-hand manipulation...8 3.3 Small Sample Vial...9 3.4 Using a Micro-Pipette...9 3.4.1 3.4.2 3.4.3 3.4.4 Mounting the tip...9 Setting the volume...9 Pipetting...10 Discarding the tip...10 3.5 Using a Syringe...11 4 Scenarios for Further Testing...12 4.1 Inoculating an Egg...12 4.2 Pipetting Material Between Sample Containers...12 5 EtherCAT and BioTAC Upgrade...13 5.1 EtherCAT...13 5.2 BioTAC Sensors...13 6 Conclusion...13 2/13
1 Introduction This document describes the initial feasibility tests relating to the use of the Shadow Robot Smart Motor Dexterous Hand for biochemical laboratory automation. The aim of these tests was to ascertain the physical possibilities and limitations of the Hand in relation to various tasks that might be undertaken in a biochemical laboratory, especially those that pose a particular risk of infection to a human operator. A selection of such tasks was chosen and attempted using the Hand built for this project. The methods and results of this investigation are presented below. 2 Experiments A selections of tasks was chosen to test different aspects of the Hand. This comprised the following: Removing and replacing the lids of a selection of vessels - a large vessels to test the strength of the Hand, and smaller ones to test delicate handling. Operating a micro-pipette. This tests the strength of various movements in the Hand and its ability to perform operations on an object while grasping it. Using a syringe. This tests the ability of the Hand to grasp and operate a small fiddly object. Since the aim of this investigation was to test the physical capabilities of the Smart Motor Hand, we did not concentrate on the specifics of the control strategies used. The tests were performed using various combinations of simple automation scripts, direct control of individual joint angles using the Joint Sliders program and operator control using a CyberGlove. Tactile or force based control was not used in these experiments. The Hand was mounted on a robotic arm for these tests. However, the arm was only used for positioning the Hand. For opening and closing vessels, the arm was used to remove and replace the lids once they were unscrewed, but not to help with unscrewing them. 2.1 Opening and Closing a Plastic Jar The jar used in this experiment is displayed in the figure 1. The task was defined as: Unscrewing the lid. Lifting and replacing the lid. Screwing the lid shut. Initially, the jar was mounted in a vice on a desk. To test the maximum grip of the Hand we used a disk power grasp. To provide rotation of this grasp around a stationary axis would have required coordination of the wrist and arm. A simpler way of doing this was to grasp the lid of the jar in the Hand then manually rotate the jar itself. Figure 1: Sample Jar with 15cm ruler for scale. 3/13
2.2 Opening and Closing a Sample Tube The tube used is shown in Figure 2. Similarly to 2.1, this task was defined as unscrewing the lid of the tube, removing it, replacing it and re-closing it. Initially the tube was mounted in a vice. The same task was then attempted in-hand. Figure 2: Sample Tube with 15cm ruler for scale. 2.3 Opening and Closing a Vial The vial is shown in Figure 3. Once again this task was tested with the object mounted in a vice. A brief attempt was made to perform the task holding the tube in-hand, but this proved futile. Figure 3: Sample vial with 15cm ruler for scale. 4/13
2.4 Operating a Micro-Pipette The micro-pipette used is shown in Figure 4. This task consisted of four stages: 1) Attaching a tip (D). This is done simply by pushing the pipette down into the tip which is held in a rack. 2) Setting the volume. This is done by rotating either the button (A) or the dial (C). 3) Pipetting. This is done by depressing the white button (A) twice, once to fill and once to empty. A hard push is required to completely empty the pipette. 4) Discarding the tip. This is done by pushing down on the lever marked B. 2.5 Using a Syringe This test was defined as: 1) Drawing back the plunger to fill the syringe. 2) Pushing the plunger to empty it again. The syringe used is shown in Figure 5. Figure 4 Pipette Figure 5: Syringe. 5/13
3 Results 3.1 Plastic Jar If the jar was sealed as tightly as possible (as one would hope if it contained deadly pathogens), the Hand was not able to grip the lid hard enough to unscrew it. Even in an ideal case, where the Hand was closed tightly onto the lid in a disc power grasp and the jar rotated, the lid simply slipped within the Hand. However, if the lid was slightly looser, the Hand was able to undo it. To fully open the jar the lid, initially we simply used one finger to rotate the lid. This obviously limits the strength that can be applied and also means that the jar needs to be fixed quite strongly to the desk to prevent it moving. Using the grasp shown in Figure 6 provided more dexterity, however coordinating the movements was difficult. Tracing the fingertips precisely around the lid simply by setting joint angles is obviously very difficult, particularly as this simple control is not as precise or repeatable as would be preferred. The fingertips tended either to touch the lid too lightly, or to push too hard on it and slip off. Figure 6: The Hand grasping the jar lid with a precision grasp. 6/13
Using the CyberGlove allows greater adaptability of control, but this presented a problem due to the way in which it measures abduction and adduction of the fingers. The glove only measures 3 angles which are used to control 4 joints. Independent control of these joints is therefore not possible. Screwing the jar shut showed similar problems as unscrewing it. It was difficult to coordinate the movement of the fingers to perform the task smoothly. Using a power grasp (Figure 7), it was possible to do the jar up tightly enough to seal, although perhaps not quite as tightly as a cautious human operator. Figure 7: Hand grasping jar with power grasp. 3.2 3.2.1 Sample Tube Mounted in a vice This task was possible to perform using a tripod grasp (Figure 8), although it unsurprisingly proved considerably more difficult than the jar. This was due more to difficulty with control than the physical capabilities of the Hand. 7/13
Figure 8:Turning the tube lid. With scripted movement, the main problem was producing synchronising control between the finger joints to produce smooth movement around the cap. With a small object such as this, the imprecision in fingertip position is particularly noticeable. 3.2.2 In-hand manipulation This was just about possible, although it pushes the range of movement of the Hand. Finger joints 3 and 4 as well as thumb joint 4 are at the limits of their ranges. The tube was held as in Figure 9, with the thumb and first finger being used to turn the lid. While we were able to unscrew the pre-loosened lid, it proved very difficult to coordinate the movement to lift it any distance away from the tube without dropping it. Figure 9: Manipulating the tube lid in-hand. 8/13
3.3 Small Sample Vial With the vial mounted in a vice and using thumb and first finger to grasp the lid (Figure 10), this was possible with much the same caveats as for the previous two tasks. The lid had to be pre-loosened, and the control was too basic. Also, the size of the lid made handling it reliably (and also picking it up once dropped) particularly difficult. Figure 10: Grasping the vial The task proved extremely fiddly when performed in-hand. Due to the small length of the vial, holding it stable while manipulating the lid was virtually impossible. 3.4 3.4.1 Using a Micro-Pipette Mounting the tip No manipulation is required for this task. The Hand was able to hold the pipette strongly enough to push it into the tip. 3.4.2 Setting the volume A lack of grip on the fingertip made this difficult, although the dial is so slippery and the mechanism so stiff that this operation is fairly difficult for a human to do. This may be due to the sterilisation process that had been applied to the apparatus prior to delivery. However, it was possible to adjust the volume in one direction by pressing the first fingertip hard into the dial then flexing joint 3 (Figure 12). There was not sufficient grip to adjust the other way. 9/13
It proved very difficult to grasp the pipette effectively in a position where it would have been possible to use the thumb on the dial. By gripping the top button in a tripod grasp and rotating the pipette manually, it was possible to adjust the volume in both directions. 3.4.3 Pipetting This was constrained by the strength of the thumb. It was possible to depress the plunger and fill the pipette by flexing thumb joint 1 (Figure 11). However, it was not possible make the final extra push to completely empty the tip. Figure 11: The thumb operating the pipette plunger. 3.4.4 Discarding the tip This was possible by pressing the thumb joint 2/3 hub into the lever (Figure 11). The result was a little imprecise although effective. Figure 11: Close up of pipette showing the finger position for adjusting the volume and also the thumb operating the tip eject lever. 10/13
3.5 Using a Syringe It was difficult to find a grasp for the syringe that was secure enough and also allowed the plunger to be operated. This was eventually accomplished by gripping the barrel of the syringe in between the fingertips and proximal phalanges of the first three fingers, with joint 3 of the middle finger offset from the other two (Figure 13). The syringe could then be operated using the tip of the thumb. Figure 13: Syringe grasping and operation. This has the obvious drawback that it is not possible to see the scale on the syringe. Releasing the grasp of the ring finger allows access to the scale but the grasp seems to rely on the triangle formed by the three fingers, and was then not secure enough to operate reliably. 11/13
4 Scenarios for Further Testing In order to test the applicability of the hand to automated performance of real biochemical research operations, the project partners have agreed on the following two more realistic scenarios to attempt with automation: 4.1 Inoculating an Egg Eggs are used in biochemical research to incubate viral pathogens. An egg has a hole carefully drilled in its shell. A syringe is then used to inject inoculum into the air sac between the shell and the membrane. Depending on the pathogen involved, all handling of the virus must be done in containment. The initial drilling of the shell is done outside of the handling cupboard. Starting with a pre-drilled egg, a tube full of liquid and a syringe with needle mounted on it, the procedure for carrying out this scenario will be: 1. Open the tube. 2. Grasp the syringe. 3. Fill the syringe from the tube. 4. Position the needle of the syringe through the hole in the egg's shell. 5. Inject the material into the egg. 6. Discard the syringe. 7. Close the tube. 4.2 Pipetting Material Between Sample Containers While this scenario may seem simple, it is an operation that is carried during many different procedures in biochemical labs. It also in fact involves many steps. Starting with two tubes (one full, one empty), a pipette, a box of pipette tips and a container filled with disinfectant, the steps are as follows: 1. Open the tubes. 2. Grasp the pipette. 3. Fit a tip. 4. Set the volume to be transferred. 5. Fill the pipette from the first tube. 6. Empty the pipette into the second tube. 7. Fill the pipette with disinfectant. 8. Eject the tip into the disinfectant container. 9. Put down the pipette. 10. Close the tubes. 12/13
5 EtherCAT and BioTAC Upgrade The tests described in this document were carried out using a C6M2 Hand. As agreed, the hand will be upgraded to C6M2E. It will also be fitted with Syntouch BioTAC sensors. Once this work has been carried out, the tasks described here will be repeated. 5.1 EtherCAT The EtherCAT bus allows a much greater bandwidth for communication between the Hand and the host computer. In turn, this allows finer position control, as well as effective torque or velocity control and active compliance. This will bring improvements for the tasks tested here. Grasping and turning the lids of containers securely and accurately would be a lot easier using compliant control. The greater precision and stability of position control will also it easer to handle small objects. The EtherCAT system closes the control loops on the host (with the exception of torque control). This will allow more complex control strategies to be employed than were possible for these tests. The tests described in this document were carried out using position control only. This means that the maximum possible strength of the Hand was not available. As the control of the EtherCAT Hand can be adjusted dynamically, greater force could be applied when needed. This could have particular relevance to the opening of tight containers, and also to the operation of the micro-pipette plunger. 5.2 BioTAC Sensors The addition of the BioTAC sensors will allow fine touch-based control. This should further ease fine manipulation tasks. The BioTACs are softer than the standard fingertips on the Hand, and the gripping surface also extends further around the fingers. This again should make it easier to grip, particularly while the object and grasp are moving. The addition of the BioTAC sensors will change the kinematic structure of the Hand. As a BioTAC sensor and its adaptor replace both the proximal and middle phalanges of the digit, the distal joints of the fingers and thumbs are no longer flexible but set at 18. The ranges of movement of the tips of the fingers and thumb are therefore reduced. The smallest diameter that can be held in a wrap type grasp is also increased. In most situations, this should not be a problem. Indeed, in some cases, this may make manipulation easier to control as the non-deterministic coupling between the middle and distal phalanges is removed. However, to operate a syringe in the way we did for these tests will longer be possible. Another method of grasping the syringe will have to be developed. 6 Conclusion In general, the Hand was able to manage the tasks attempted, although often not as precisely or repeatably as would have been preferred. The use of higher level control strategies should make these tasks easier and more reliable to perform. However, these tests have highlighted some issues with the Hand itself. We are optimistic that many of these will be addressed by the switch to EtherCAT and the addition of BioTACs. Other problems will require more work on the part of Shadow. We look to our partners to help develop the high level control which will make these tasks possible to perform repeatedly and automatically, and also to further inform our work improving future versions of the Hand. 13/13