LewanSould also has what appears to be a fairly well-built, if not particularly long reach, 6DOF arm for $119.99 with servos (not the digital bus servos above, though). This will definitely give the Lynxmotion Arms some serious competition.
Friday, November 03, 2017
New Affordable Bus Robotic Servos from LewanSoul
No idea as to the actual quality of these, but LewanSoul now has the LX-16A 17 kg·cm (236 oz·in) @7.4V robotic servo controlled over a bus (perhaps with the same command structure as a Dynamixel) with position/voltage/temperature feedback for only $15.99.
LewanSould also has what appears to be a fairly well-built, if not particularly long reach, 6DOF arm for $119.99 with servos (not the digital bus servos above, though). This will definitely give the Lynxmotion Arms some serious competition.
LewanSould also has what appears to be a fairly well-built, if not particularly long reach, 6DOF arm for $119.99 with servos (not the digital bus servos above, though). This will definitely give the Lynxmotion Arms some serious competition.
Sunday, May 29, 2016
Web controlled robotic arm with Pi
A web controlled robotic arm based on the Raspberry Pi.
There is a daemon running in the background called web_arm.py, which sets up a FIFO (arm.fifo) to receive commands. The cgi-bin is xy.py which receives x,y coordinates from a click on an image, then the cgi-bin sends the coordinates to the fifo. web_arm performs inverse kinematics and drives the servo controller over the serial port, and the servos drive the arm.
Full code on GitHub.
There is a daemon running in the background called web_arm.py, which sets up a FIFO (arm.fifo) to receive commands. The cgi-bin is xy.py which receives x,y coordinates from a click on an image, then the cgi-bin sends the coordinates to the fifo. web_arm performs inverse kinematics and drives the servo controller over the serial port, and the servos drive the arm.
Full code on GitHub.
Wednesday, January 20, 2016
Desktop GUI for AWS Ubuntu
This seemed harder to Google for than you'd think.
Imagine you want a desktop on Ubuntu Linux on an AWS EC2 instance, and you'd like to access it from OS X. The sanest thing is to follow these instructions on installing kubuntu-desktop and tightvnc on the Ubuntu instance [with more info here], and the "Screen Sharing" app built into OS X. Access the Ubuntu instance with "[ip address]:590[x]" where [x] is the X display number that tightvnc attaches to, generally 1 (so it would be "[ip address]:5901"). Don't forget to set up a security group to allow the VNC port(s).
Imagine you want a desktop on Ubuntu Linux on an AWS EC2 instance, and you'd like to access it from OS X. The sanest thing is to follow these instructions on installing kubuntu-desktop and tightvnc on the Ubuntu instance [with more info here], and the "Screen Sharing" app built into OS X. Access the Ubuntu instance with "[ip address]:590[x]" where [x] is the X display number that tightvnc attaches to, generally 1 (so it would be "[ip address]:5901"). Don't forget to set up a security group to allow the VNC port(s).
Friday, March 13, 2015
Netflix Speed Graph over Time
As we learn about the new "Net Neutrality" rules from the FCC, I thought it would be interesting to share the graph from the Netflix Speed Report of average Netflix access speed for various carriers over time. Of note, things got much better in Sept. 2014 when interconnection arrangements were improved for several major carriers.
Wednesday, March 11, 2015
3D Scan/Print Kiosks Review
I recently ran into two 3D scan/print kiosks in malls. The first is DOOB, which has kiosks at Santa Monica Place, CA (just off Third Street Promenade) and Chelsea Market NYC. The kiosk has a very large cylindrical room with a wall covered with cameras and lights. You walk in, get captured instantly by the cameras, and then they ship you the prints in about two weeks. Prints go up to 14 inches tall, but they are a little expensive, so I went for the cheapest one (4 inches tall). The result is cool, but looks a bit "out of focus":
DOOB does not give you the mesh file, which is sad.
Then I ran into Twindom, which has studio locations in San Mateo, CA, Houston, TX, and a kiosk at the Valley Fair Mall in Santa Clara, CA. They also had a cylindrical "Twinstant" room with 91 cameras. Unlike DOOB, "Twinstant" appears to use typical consumer DSLRs (you can hear them click), and it takes about a second for them all to snap a shot. The prints I saw at the kiosk were clearly inferior to DOOB, so I did not bother with a print, but they said I could get a scan and download the mesh file, so I went for it. Here is my mesh in mesh lab:
Overall the mesh looks OK, but a closeup of my face shows a lot of problems. I think it might be useful if they dedicated some zoomed in cameras for enhanced photogrametry in the face. As human beings, we are more likely to notice problems in the face than problems with your pants, etc.
My verdict: Given the huge number of cameras in these rooms, I was expecting better results. However I will admit that there is no simple way to quickly and affordably do an "instant" scan your entire body except for DOOB & Twindom. My experiments with structured light Kinect & Scanect where I had to spin around on a chair without moving much did not do much better with my face.
On the other hand, if you just want to capture a 3D mesh of your face or torso, you are better off freezing in one place and letting a friend take 10 or so good shots with a DSLR and then process them into a mesh with Autocad 123D Catch.
That said, my kids are happy that I can now hang out with Peppa Pig and her friends!
DOOB does not give you the mesh file, which is sad.
Then I ran into Twindom, which has studio locations in San Mateo, CA, Houston, TX, and a kiosk at the Valley Fair Mall in Santa Clara, CA. They also had a cylindrical "Twinstant" room with 91 cameras. Unlike DOOB, "Twinstant" appears to use typical consumer DSLRs (you can hear them click), and it takes about a second for them all to snap a shot. The prints I saw at the kiosk were clearly inferior to DOOB, so I did not bother with a print, but they said I could get a scan and download the mesh file, so I went for it. Here is my mesh in mesh lab:
Overall the mesh looks OK, but a closeup of my face shows a lot of problems. I think it might be useful if they dedicated some zoomed in cameras for enhanced photogrametry in the face. As human beings, we are more likely to notice problems in the face than problems with your pants, etc.
My verdict: Given the huge number of cameras in these rooms, I was expecting better results. However I will admit that there is no simple way to quickly and affordably do an "instant" scan your entire body except for DOOB & Twindom. My experiments with structured light Kinect & Scanect where I had to spin around on a chair without moving much did not do much better with my face.
On the other hand, if you just want to capture a 3D mesh of your face or torso, you are better off freezing in one place and letting a friend take 10 or so good shots with a DSLR and then process them into a mesh with Autocad 123D Catch.
That said, my kids are happy that I can now hang out with Peppa Pig and her friends!
Tuesday, March 10, 2015
Long Loops and Slow US Internet Access Speeds
Like many people, I've been wondering why the US seems to lag many other countries on Internet access speeds.
Here is a key issue: Our twisted-pair copper local loop infrastructure has longer local loops:
(source "Explaining International Broadband Leadership", The Information & Technology Innovation Foundation)
Because we have longer local loops, DSL speeds are always going to be slower than countries with shorter local loops. For example, loops over 3km are never going to see 10 Mbps DSL:
Why does the US have such long local loops? This is a question I don't have a good answer to. I do not believe it is just that the US is more rural, as I have heard complaints about 3 km local loops in fairly urban areas. Australia also has lots of rural areas, but it has shorter local loops on average.
It is possible that it is because the US adopted electronic telephone switching before other countries, and/or perhaps there were stronger forces driving Central Office consolidation. You can imagine that with 4 km local loops, one CO can serve the same area as 7 COs with 1.5 km local loops. Thus it is more efficient for telephony to have fewer central offices and longer local loops, but it turns out to be bad years later for DSL.
Because of this, the highest speed broadband in the US is going to be dominated by cable (DOCSIS), Fiber to the Node (FTTN), and Fiber to the Home (FTTH), in order of expense and speed. But a lot of people are cheap, and like DSL. Plus DSL runs on telephony infrastructure that is already run almost everywhere, while the other faster technologies need newer build-outs. So this structural difference will likely keep US average Internet speeds down even while higher speed services like 1 Gbps FTTH Google Fiber is being deployed.
There also is a competitive issue - the slow DSL reduces the pressure on alternative connectivity mechanisms to deliver faster service.
That said, Americans are getting faster average Internet speeds over time, now doubling every three years. Here is a graph based on data from The Akamai State of the Internet Reports:
Here is a key issue: Our twisted-pair copper local loop infrastructure has longer local loops:
Because we have longer local loops, DSL speeds are always going to be slower than countries with shorter local loops. For example, loops over 3km are never going to see 10 Mbps DSL:
Why does the US have such long local loops? This is a question I don't have a good answer to. I do not believe it is just that the US is more rural, as I have heard complaints about 3 km local loops in fairly urban areas. Australia also has lots of rural areas, but it has shorter local loops on average.
It is possible that it is because the US adopted electronic telephone switching before other countries, and/or perhaps there were stronger forces driving Central Office consolidation. You can imagine that with 4 km local loops, one CO can serve the same area as 7 COs with 1.5 km local loops. Thus it is more efficient for telephony to have fewer central offices and longer local loops, but it turns out to be bad years later for DSL.
Because of this, the highest speed broadband in the US is going to be dominated by cable (DOCSIS), Fiber to the Node (FTTN), and Fiber to the Home (FTTH), in order of expense and speed. But a lot of people are cheap, and like DSL. Plus DSL runs on telephony infrastructure that is already run almost everywhere, while the other faster technologies need newer build-outs. So this structural difference will likely keep US average Internet speeds down even while higher speed services like 1 Gbps FTTH Google Fiber is being deployed.
There also is a competitive issue - the slow DSL reduces the pressure on alternative connectivity mechanisms to deliver faster service.
That said, Americans are getting faster average Internet speeds over time, now doubling every three years. Here is a graph based on data from The Akamai State of the Internet Reports:
Wednesday, December 03, 2014
First run with Kinect Xbox & Skanect
So I went ahead and picked up a Kinect for Xbox and Skanect (for Mac). By the way, the Kinect came with a power supply, so I didn't need to purchase a separate power supply.
Skanect was pretty easy to use. I'm not totally happy with the mesh from this first run result, so I'll have to work on refining either my scanning technique or learning how to smooth meshes in Blender.
Skanect was pretty easy to use. I'm not totally happy with the mesh from this first run result, so I'll have to work on refining either my scanning technique or learning how to smooth meshes in Blender.
Subscribe to:
Posts (Atom)








