Monthly Archives: August 2009

Improving robot interactions with the environment

augmented reality, computer vision, control, gripper, willow garage

I want to know if they tested the robot’s skull egg crushing gripper without the plastic bag first, or if that was a scientific development discovered during the research process. Also, do the bags have any effect on the performance of the gripper?

They also have some interesting information about using human turks to train robot vision classifiers. I am sure the Augmented Reality crowd can make some use of the cv-web-annotation-toolkit sometime in the next few months.

Willow Garage also has an RSS feed for their blog so now it is easier than ever to keep up with their research.

The future of reality

augmented reality, awesomeness, computer vision

Bruce Sterling is a man who understands the future.

His presentation on the augmented reality industry is insightful. He explains the opportunities and pitfalls awaiting this new industry and how AR companies need to prepare for them. I wish someone explained the future this clearly to the web 0.1 startups back in the day.

The basic idea is that AR is either going to turn out to be VRML 2.0 or “Bigger then the web”. I personally am betting that augmented reality will be bigger than the internet, if we can solve the problem of mobile data charges.

It may not be initially clear as to how this can be true or how this applies to the field of robotics, so let me use this graph to explain.

As you can see from the graph, awesomeness will increase exponentially as we head towards the future. Proof of concept and demo applications are just now starting to pop up showing how augmented reality can allow you to find your way to the nearest subway, play games, go shopping or look up the wikipedia entries on anything you (your camera) can see.

These applications give the first glimpse of what the future is going to look like, however the biggest impact will be after augmented reality applications have gone mainstream.

After the foundations made up of augmented reality databases, fiducial markers and 2D bar codes are in place, we will begin to see the development of a vast new range of practical civilian mobile robotics. When every food product has 2D barcodes linked to an internet accessible database allowing your iPhone to check the nutritional information of the items against your current diet and allergies in real time, this is useful technological progress. However this same technological infrastructure will also allow robots to assist people in ways that are much more complicated than simply vacuuming your floor.

In someways these applications will start with something trivial like having a robot perform beverage retrieval while you are watching a movie. However more complicated tasks can easily be envisioned such as using 2D barcodes embedded into the washing instructions of your clothing allowing your robot to not only add the correct amount of bleach for your white shirts while doing your laundry for you but also to take out the dry cleaning to get done.

Wide deployment of augmented reality is the step right before we start seeing robots walking down the sidewalks of New York City driven by tourists from Tokyo on virtual vacations. The worst case is that I expect to be able to drive a robot around the streets of Akiba from my apartment in New York City within three years. I hope to be aided by localization, navigation and automatic translation provided by QR Codes painted on the sidewalks and sides of buildings so I can read all the signs in English.

There are thousands of practical robotics applications that will be unlocked as the technological foundations for augmented reality are built. So, get ready for the future!

Dell Latitude 2100

ground station, laboratory supplies, laptop, review, telemetry

So the Mini 9 is gone going to be gone soon, so it is time to consider the alternatives for robot ground stations. The Dell Latitude 2100 is a netbook that seems to be designed for students whose destructive tendencies approaches that of angry primates. This also makes it useful when working on robots outdoors.


The outer surface of the case has a texture that is almost exactly the same as a dodge ball. The case itself seems well built and sturdy and the hinge for the display looks like one of the sturdiest designs I have seen in a laptop.


For some reason, Dell wanted me to agree to a EULA that they failed to include in a format I could actually read. However, I was able to install Ubuntu via USB flash drive without an issue. Make sure to update the operating system once you connect it to the network otherwise the Netbook Remix is broken in many random places.

All of the hardware works under Ubuntu 9.04 including the touchscreen and wifi. The bluetooth came up and I would assume it works, however I did not test it yet.

Trying to get Dell to sell you the hardware without windows seems to be impossible, as some of it is not officially supported under Ubuntu 8.04, perhaps Dell could sell misconfigured/unsupported systems for Linux users. On the upside, Microsoft probably allows the ram on the Latitude 2100 to be upgradable since you have to remove the keyboard to upgrade it. I blame Steve Ballmer.


The SD Card slot is fairly well designed allowing a normal memory card to protrude only 1-2mm outside of the case. It is as close to flush as it can be and still be easily removed. This allows those of us who like to keep our home directories on a memory card to keep it in at all times without risking damage. I would rather that it was flush but this is a reasonable design compromise for most users.


The Network Activity Light on the lid is semi functional under the Ubuntu Netbook Remix 9.04. If the network is switched on it is illuminated, if you disable wifi it turns off, it does not blink based on network traffic. The wifi switch appears to act like a USB hub connect/disconnect and the light itself appears at first glance to come up as a separate USB device. If the light can in fact be controlled via USB then it could be used as a visual alarm if you are using the laptop as a ground station.

Overall the Dell Latitude 2100 looks like it will make a great ground station for controlling robots. The only features missing are integrated GPS and cellular data, but based on the pricing for cellular data this is not a huge loss. In the future I’ll look into integrating the Canmore GPS when I try to install a solid state drive for extra ruggedness

School is out for the summer!

competition, micromouse, NXT, robotics, video

I strongly suggest against using medium density fiberboard for the walls of the maze; it does not work that well and the price for pine is only slightly higher. Drilling holes for the wood dowels tends to destroy the structural integrity of the MDF and there is no way to route the edges of the walls to join them. I would definitely paint the floor of the maze black next time and use white electrical tape for the line following portion of the challenge. Black lines on white worked well but the floor of the maze tends to get dirty.

Building the maze in accordance with the micromouse specifications allows the maze to be used for multiple contests. The Lego NXT robots seem to work best with about twice the width between the walls, so just leave out every other micromouse wall.

The NXT-G Programming language worked fairly well until some student’s algorithms become too complicated. At the end some groups were constantly running out of memory and having issues broken myblocks and unconnected wires and elements. However for the majority of students without a programming background the NXT-G programming language was much easier to use then anything else despite its limitations. The biggest problem with using the lego NXT sets for a robotics class is that the students consider them toys so they initially underestimate the challenges ahead of them.