Meeting Minutes: Multi-scale Manipulation and Automation
October 2015
October 6th
Lego Parts:
In the process of re-evaluating the part design two approaches to overcoming the adhesion problem have been selected. First, the parts will be moved to the plastic substrate of the petri dish to determine if surface forces are still too great to allow proper manipulation. Second, the design will be re-worked to include "nubs" that will reduce surface area contact on the sliding faces.
Testbed Status:
The secondary camera has been successfully added but cannot be used with the petri dishes currently used to store samples. Square glass slides will be ordered to allow clear view from the side and the full integration of the side camera in the process.
Micro-part Production:
Two companies, Newport and Teem Photonics, have been identified as potential suppliers of micro 3d printing systems at a much lower cost than the nanoscribe system. A conference call will be set-up for Tuesday October 13th to discuss options with their sales representative.
September 2015
September 29th
Lego parts:
The parts from UIC are great and the dimensions as exact as we could hope for. They may actually be too perfect as the adhesion on the glass is enough to interfere with the cage and release the part prematurely. Multiple solutions will be pursued including new manipulators that can reduce compliance and play and a re-design of the parts to reduce surface area in contact with the glass.
Testbed status:
The testbed is working well but integration of the side camera is needed. Due to the small focal length of the lens initially used another lens with lower magnification will be used.
March 2015
March 31st
Testbed Status
The new optics appear to be ideal for adding the second camera to monitor Z height of the manipulators. Unfortunately the Flea3 camera that is intended to be used does not appear to be working. A support ticket has been opened with Point Grey to resolve the issue.
Tinker-Toy case study
The BLS algorithm has been shown to produce feasible solutions with the square plan approximately 87% of the time in <12s but the cube plan is much more challenging requiring >20mins to process. Based on this a new strategy will be explored which will reduce the number of sub assemblies that are considered either through randomization or a faster initial evaluation.
February 2015
February 24th
Testbed Status
No significant change since last week. The optical and soldering parts have been ordered and the upgrade will start at the next opportunity once they are in.
Tinker-Toy case study
Further work with Dr. Pan's group has hopefully led to successful parts. If not a redesign will be needed to address the problem of part removal from the substrate which up until now has resulted in part failure for the correctly dimensioned parts. The BLS algorithm developed by Ghandi et al. is planned to be used to generate an optimized assembly strategy (see paper here: http://www.sciencedirect.com/science/article/pii/S0952197614003017). Hopefully this will lead to some novel results that can be tested before the March 15th deadline for Service Robotics.
February 17th
Testbed Status
The testbed is in full working order with the current components. The next steps will be to integrate the soldering and side camera. In this interest a list of the parts needed including replacement probe tips will be made.
Tinker-Toy case study
The case study is currently in a holding pattern while the problems with part production are being worked out. In the mean time a planning algorithm will be developed and tested to determine the proper steps to create arbitrary configurations of the cubes and rods. This will be based on the matrix decomposition technique previously proposed.
February 10th
Testbed Status
The testbed has continued to be improved with video recording and more advanced controls. Some accumulated error due to the calculations used in the kinematic transforms was noted but has now been corrected. Further improvements will be implemented as time allows.
Tinker-Toy case study
At this time there are still major issues with sticking in the manipulation of the Tinker-Toy pieces as well as challenges with the interaction of the probes and the holes in the cube pieces. Different caging grasps will be tested and evaluated for their effectiveness at mitigating these issues. The parts provided by Dr. Pan's group still exhibit a slight asymmetry in the Z dimension which I believe will only allow the rods to be inserted into 4 of the 6 holes of the cubes. This will limit the early work to 2D structures. Anurup has been informed of the asymmetry and will be working to correct it.
Assembly planning
An initial idea for assembly planning is being developed based on the tinker-toy case study. In this technique, a matrix will be generated representing the final structure to be made and matrix operations representing different movement primitives will be used to "deconstruct" the assembly. These steps should then be able to be used in reverse order to create the design. Optimization will also be integrated in the form of weights applied to the primitives based on the estimated time and failure rate of each primitive.
January 2015
January 13th
Testbed Status
All basic features of the GUI are implemented. From now on the code will be expanded to aid in the manipulation tasks. First and foremost the manipulator positions will be converted into Cartesian coordinates of the end point and a simple calibration algorithm developed.
Tinker-Toy case study
The "tinker-toy" pieces will be used as a case study for ASME IDETC. The main focus will be on creating a bar bell structure, square, and cube as time allows. Further parts will be coordinated with Bixing including some design changes such as a potential taper to the rod pieces to aid in assembly.
January 7th
Testbed Status
The code for the GUI is in a final stage of debugging for basic features. Further expansions such as the ability to define multiple cages will be considered and implemented on an ongoing basis.
Micro Solder Options
Two main companies have been identified as potential suppliers of soldering equipment:
Hakko: http://www.hakko.com/
JBC: http://www.jbctools.com
Most likely the auto-feed system from Hakko (374) will be used and modified to allow computer controlled soldering.
Another option discussed was treating pieces of solder in the workspace as work pieces to be placed by the system and simply heated by a soldering iron. The main obstacle at this time is the available tip radius of soldering irons seeming to bottom out at 0.1mm. Further options will be explored including using a current through a probe tip to join pieces.
ASME IDETC
There is the potential to provide a paper or technical briefing involving the case of the "tinker toy" system created with the sterolithography technique. This will be explored with the current available pieces and the feasibility of having something ready by the end of the month evaluated.
December 2014
December 17th
Testbed Status
No significant progress since the 10th. Plan to look at existing code for micro robot test bed in order to avoid duplicate work.
Application Ideas
Pan: micro batteries that will need to be placed onto antenna.
Wood: placement and soldering of sensors and components onto robobee.
To support these applications as well as the plane folding will need to look into a computer controlled soldering iron. This would be moved with the Sutter system and would be used to solder components or potentially achieve folding by heating. This would be achieved by the expansion of dissimilar metal layers. Ideally there would be temperature control through the program in order to realize different folding angles.
December 10th
Testbed Status
All code converted to VS2013. After contact with technical support all Point Grey SDK components working. Camera feed will be integrated into existing control code and a GUI developed. Goal to have system up and running manually by 12/17.
Origami ideas
At this time the focus will be on a simple example of origami folding, the paper airplane. This will require only a few simple folds and will allow facets to be fixed permanently after folding.
One issue will be holding facets in place while they are joined with epoxy or some other solution. A possibility to solve this is to have fixtures in the environment that can be used as tools by the manipulators to hold pieces in place or define fold angles. Magnetically controlled fixtures are one possibility that would use mobile fixtures and then hold them in place with a magnetic field. The electromagnet used would need to surround the objective which may become problematic based on space restrictions around the workspace. Another possibility is the presence of immobile fences or guides that could be used for the same purposes.
A few possibilities for fixing the facets include epoxy, photo-curable substances, solder, and mechanical forces. The epoxy would be the simplest to implement but may not be compatible with all substances or folds. The photo-curable substance could be dispensed by one manipulator which also integrates a laser for instant curing something like welding at the micro scale. Solder could be used if metal tabs are designed at contact points but should be investigated for size restrictions. Mechanical possibilities could include a slot/tab design much like that investigated by Rob Wood's group with actuated surfaces.
Ideally a dispenser compatible with epoxy and photo-curable substances should be found. Two possibilities are Kleindiek and Eppendorf which should be compared for their suitability to this application. In general injection pipettes and syringe pumps should also be investigated.
November 2014
November 19th
Test Bet Status
Nikon stage controlled through portable c library.
Point Grey camera SDK not fully compatible with Visual Studio 2013. Technical support will be contacted for solution.
Microscope focus non-functional. No response from service technician. Follow up needed.
Test Case
Potential test case in assembling micro force sensing robot for Wuming. Would involve picking up rectangular magnet, dipping in epoxy, and applying to force sensor. Should be fully automated with part discovery.
Challenges:
Clearing height of dishes containing parts and epoxy.
Z calibration for height of epoxy (detect contrast change during dip?)
Ensuring good contact with part (not glued to dish or glue rubbed off)
November 5
Test Bed Status
Sutter manipulators fully controlled through portable c library.
SDK packages found for Nikon and PointGrey systems.
Goal: have system running manually before end of semester.
Test Case
Need to have a test case worked out by March 1st next year for IROS deadline. Could either be comparison of folding techniques for origami-like tasks or robo-bee assembly-like tasks.
Should show a clear advantage of such a system and identify any areas for improvement of the control of the system.
After this work on automating the task (potentially for IRCA?).
October 2014
October 8
New computer has been ordered, monitor is in
New control program has been started - initial work on controlling manipulators currently underway
Investigation of desktop manipulators and comparison metrics to be started this week
September 2014
September 24
Current Test-bed
Get current system up and running in C-based program
Order new computer for test-bed
Order micro-grippers for system
Desk-Top System
Investigate ideal desktop manipulators:
Kleindiek manipulators
miBot manipulators
Sutter manipulators
Dan Popa's group modular manipulators
Investigate setup for stereo vision system (see Brad Nelson, ETH Zurich - I think they did something like this).
uROS
Looking to modularity and structure of ROS to build new uROS
Drivers for common manipulators/stages
Leverage current simulation packages (Gazeebo, V-REP, etc.) and then plug in micro-world physics engine once available
Strategy
Short-term:
Work on generic, flexible, software system for current set up and implement
CAD-based microassembly
Case-studies
Identify ideal desk-top system components
Long-term:
Hardware design/specification/procurement for desk-top system and implementation
Open-source uROS package