« Using random video input for music generation | Main | ICM Midterm Dialogue »

Midterm project completed - field report & lessons learned

Shinyoung (Rona), Younghyun & myself took the "intelligent restaurant lamp" to a restaurant. We've documented our experience, and some lessons learned, after the link. Please also look at Rona's blog and Younghyun's blog.

Click here for the demo movie.

Interestingly, the waiter told us that two years ago a similar product had been tested in the restaurant, a button-like object that lit up red or green, according to whether the customer wanted to order or ask for the check.

He did not think his restaurant was big enough to justify such a product, and thought the staff was quite alert. However, customers at the next table disagreed, and thought the lamp would be a good idea!

We learned a lot carrying out this project:

1) IR sensors are very delicate, and sensitive to physical positioning!

2) IR sensors can be triggered by tungsten lights in the room. This has the unforeseen effect of changing the function in the lamp according to room lighting. When it is dark, the lamp is triggered by detecting IR light reflected by the user's hand. However, when there is bright IR-emitting lighting, the lamp should be triggered by detecting an IR shadow cast by the hand. We've tested this by manually making the switch, but it should be possible to automate by averaging readings from across all three sensors, and looking for the "odd man out". Alternatively, choose a better proximity sensor.I now understand why IR sensors like this are popular for faucets - hiding them under the tap shields them from room light.

3) The gesture control mechanism is very appealing, possibly even more than the real utility of the lamp. People really enjoyed playing with it. The way the LEDs light up progressively gives very crisp clear feedback. I noticed many people would naturally try to reverse the gesture, but by then the lamp was locked into "order" or "check" mode. I would consider making a simple room lamp with the same hardware, only using gestures to make the lamp brighter or dimmer.

4) There is no problem with misrecognized gestures. The sensors are very precisely focused, and the lamp will only trigger on the exact gesture. In fact, it is almost too precise.

5) As it stands, the lamp is too big for a restaurant like Cozy's. It needs to be smaller (flatter or narrower, height is OK), or stacked over the condiments.

6) The finite state machine as the core of the program worked very well, it was very easy to modify or add behaviours.

7) It is important to plan ahead when doing the wiring. In hindsight, I would have run a power and ground bus to all levels of the lamp. This would have meant 15 less wires running down to the breadboard, since 3 LEDs on each level could share a common ground, and the IR emmiter/sensor could share a power lead.

8) Given how easy it was to build and program, I would not hesitate to add additional functions, perhaps a solenoid-activated bell, and a few more control gestures.

TrackBack

TrackBack URL for this entry:
http://itp.nyu.edu/~gpv206/cgi-bin/mt/mt-tb.cgi/36

Post a comment

(If you haven't left a comment here before, you may need to be approved by the site owner before your comment will appear. Until then, it won't appear on the entry. Thanks for waiting.)