Skip to content

Update 4/18/13

April 18, 2013

We’re continuing to try and push forward to complete as much as we can before open house next Wednesday. Lucas has made some progress on characterizing the carpet of the third floor:


I’ve been working on cross-calibrating the Sonar array scan and the binary image scan in order to make each more reliable:


Image taken with a large white board standing vertically approximately a meter away from the front of the robot


The tick marks on the axis indicate meters. Dark blue: uncalibrated sonar scan; light blue: calibrated sonar scan; green: scaled image scan; red: calibrated image scan

Note the warping of the sonar scan above: even though the robot is facing a flat surface, the sonar array would have us believe the surface was curved around the front of the robot. This is likely a result of some kind of interference between adjacent sensors.

Other various news:

  • Things on the hardware list are slowly being ticked off
  • We will likely be receiving a product donation from Trimble in the form of a GPS sensor which is compatible with OmniSTAR! I got a friendly call from one of their representatives a couple of days ago requesting our shipping address.
  • The Quadro 6000 GPU that NVidia promised to donate us should be coming in either this week or the next.
  • One of our twelve sonar sensors was confirmed dead late last night. This cause of death is as yet unclear. It was possibly linked to our lack of a protective bumper in front of the robot, allowing sensors to occasionally smash into things while testing. A pool noodle/PCB enforced bumper is the highest-priority item on the hardware to-do list. It could also have been linked to a surge problem we have been observing when the remote kill switch is hit. Frank believes that the solution to this surge issue is to decouple the PSoC from the computer, using an Ethernet connection rather than USB.
  • Andrew has made excellent progress on getting V-REP up and running. We expect to have it integrated into our setup by the end of next week, at the latest. We hope to use it to run the robot through the basic and advanced courses.

Separate from running autonomously on the 3rd floor, Sagar has been working on getting white lane detection working outside. His current strategy is to blur the image, grayscale it, run Canny edge detection, and then send it through a Hough transform. He was able to get all of this done on our GPU, except for the Hough transform. One innovation he made is to run an HSV threshold on the original image to detect whiteness, dilate those pixels in the image that passed the threshold, and then do an AND operation between the thresholded binary image and the gray scaled image that was processed with Canny edge detection, just before running it through the Hough transform. The reasoning for this is that it gets rid of edges that are not apart of the more relevant, white areas of the image. Shown below are his initial results from data we recorded with white planks of wood on the ground. (All of our previously bagged data with actual white painted lines was recently lost due to file corruption on the external harddrive we were using to store bags. We felt it wouldn’t be wise to paint the lawn beside ENS without permission.)




Lines drawn from the Hough transform are shown in blue

Note that the image above was not taken with actual white painted lines, and it was a cloudy day, so the whiteness of the lanes is probably stronger than in reality and the lighting of the image is generally favorable. This is particularly relevant because of the notorious “barrel” problem cited in many an old IGVC design doc. It seems that the color on the reflective strips of construction barrels is indistinguishable from the color of the lanes. We hope to get some bags at the intramural fields either today or Friday in order to test with more realistic data.


From → Uncategorized

Leave a Comment

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: