DASL Smithy Introduction Video
March 20th, 2009
This video (big file, download only!) is a compilation the introduces the Smithy lathe, its operation, basic features, and quirks.
Dance Improvements
February 14th, 2009
I’ve imported all of the old gestures from the previous dance code. The implementation of Alex’s serial code allows us to control the motor positions at up to 50Hz (a MATLAB limitation). As shown in the video below, the robonova can now:
- Vary the speed and extent of motions
- “Crossfade” between gestures to create fluidity
- Follow an arbitrary path and return to the starting position, all while maintaining rhythm
Check out the Video to see the improvements.
Computer Infrastructure Upgrade Schedule
February 9th, 2009
Our lab’s workstations are now set to be replaced ona 3 year timeline. Purchase dates & upgrades are listed in this spreadsheet, which will be updated as computers are built and purposed. Email me if you’d like access to edit this sheet.
POSTECH Visit - Feb. 16th
February 5th, 2009
On Monday Feb. 16 at 10:00 AM, students from POSTECH will visit Drexel; a key stop on their tour is DASL. In March of 2008, Dr. Oh, myself, and 4 of our undergraduate students visited POSTECH as part of a week-long introduction to Korea. The Professor that hosted our visit is bringing 20 undergraduate students, so we are returning the favor.
Lead Coordinator: Rob
Translator: Youngbum Jun
Demo Schedule (10 minute blocks)
- Slide presentation (Overall Lab activities): Keith
- Virtual UAV Training with 50″ LCD display: James
- UAV work (e.g. Videos of Blackhawk), SISTR and HO-scale fly-by: Keith
- Ground Vehicles (e.g. Videos of DIAS), and MGC Vehicle: Noah
- Robonova Dancing: Dave Grunberg
- Hubo Choreography: Clayton McNeil & Bob Sherbert
I’d like each presenter to write down a 300-500 word description of their part, so that Youngbum has source material to base his translations on. A rehearsal date is TBA, but will be late next week (week of 2/9/2009).
Midweek Status
February 4th, 2009
Humanoid Dance Project
Thanks to Dave’s hard work over the past 2 weeks, the Robonova can now dance to any streaming audio. Our latest demo video shows it dancing in realtime to all sorts of music off of an iPod ( AVI ) ( MOV ).
A critical bottleneck we’ve been stuck with for the past 6 months or so with the Robonova has been it’s processing power. It simply can’t scan for gesture commands more quickly than about 3-10Hz. At these slow rates, timing error and loss of gestures can be a real problem for fast tempos. The result is dance moves that don’t quite line up with the tempo, even though they are generally well-timed. In the worst case, the Robonova will ignore a leg return command, and take a step assuming it is ready (then falling!)
Alex Kushleyev at UPenn had developed replacement code for the microcontroller that allows direct control of the servos at speeds of up to 100Hz (more on this later). I’ve finally made my own control code in MATLAB that can send compatible commands to the modified Robonova, which will give us almost real-time, dynamically updating servo control.
To truly take advantage of this new scheme, however, we need to mix the beat tracker code with the serial output. A simple way to do this might be:
- Process audio for x milliseconds, calculate command position
- Output serial update to robot
- Repeat
The problem with writing code this way is that the calculation time for audio and command processing is unpredictable, or at least inconsistent. A better way is to use interrupts, which pause the main program at specified intervals to perform a specific task, then resume when it’s done. Luckily, MATLAB has a well supported feature for this: the timer. As of today, I’ve incorporated this timer feature to make a simple interrupt-driven serial output. This also allows us to define the positions we want for joints at many times, and simply have the interrupt function choose the current position from memory. In other words, the open loop trajectories don’t need to be generated on-the-fly.
Next steps
- Choose a data structure that can concisely describe the motion of a gesture.
- Describe current dance motions in this new way
- Port dance motions (upper body) to Hubo using Clayton’s code
- Compare results in terms of timing, accuracy, interface similarity, and output appearance
Treadmill Drive
January 30th, 2009
RJ has confirmed that the Hubo lab’s treadmill uses a 400W, 220VAC servo drive. Due to the expense and difficulty of finding Samsung servo drives for sale in the US, I’ve compared other options. The results are listed in a trade study here.
New Website
January 29th, 2009
My new graduate student website is now up and running thanks to Bob Sherbert’s invaluable help. It will be much easier to keep this site up to date due to the software backbone and tools in WordPress. The main blog area will be for news and highlights, and the pages across the title bar describe my activities. Be sure to check back for updates; I’ll be adding new pages, media, videos, and tutorials in the next few weeks.