Topic ideas for embedded design

[polarmystery]

Hey guys,

I started a new semester of grad school and in my class our only assignment is to come up with an embedded system design of our choosing. Problem is, I'm having difficulty coming up with something that can fit in the schedule timeline to completion. Here are a few ideas other students came up with in the past:

- Roomba vacuum which fetched a tennis ball after it was thrown
- automated drink pouring machine (based on RFID tags of patrons)
- automated sensor machine to control temperature difference between two rooms in a house

I'm pretty terrible with coming up with ideas, and I've google'd a few of them. I'm having trouble finding something that would fit the scope of our project length as well (It's an electrical engineering class, so any kind(s) of camera, micro-controller, sensors, servo-motors, etc. can/should be used). Any ideas guys? I'll gladly consider anything mostly.

Cheers!

P.S. I have no imagination

 

[Modelworks]

Something I have on my list to do but I haven't gotten around to it is a pet door.
A pet door that recognizes and unlocks the door based on video of the pet, not a collar worn device, or anything else. Involves cameras, servo motors, mcu, visual processing.

Keeping in the same animal theme, how about a water gun mounted outside that will target neighbors dogs or cats and squirt them to keep them out of the yard. Would require motion tracking, servos to aim and fire, and mcu.

 

[MagnusTheBrewer]

An improved elevator call system.

 

[SMOGZINN]

How about you find out what one of the other teams projects is, and then build a robot to make what ever they are making.
So, they make a Roomba that fetches sticks, you make a stick fetching roomba building robot.

 

[Shadowmage]

One of my projects in undergrad was a self-driving Barbie Jeep:

http://www.youtube.com/watch?v=QjqRSZDC86Q

 

[William Gaatjes]

How about this :

Take 2 camera's with motorized image focus mechanics, an ultrasonic ranging system and a laser ranging system.

Use the two camera's and the feedback information from the lens focus to determine depth information. For example, three objects are placed on a table. A,B, and C. One is placed behind the other and other. Just from focusing on the three objects, you will be able to retrieve information about which object is placed further away. But you will need a software control over the focusing system. For example, the object B might return a focus feedback value of 100. A returns 98 and C returns 110. Using your ranging systems, you can determine which object is closest to your camera setup.
And then you also automatically know which value relates to an object being closer to you relative from another object. Then you can just rely on the focusing information and use the ranging systems to calibrate or when it is difficult to determine the depth using the focus information (Think of an object with a uniform color compared with an object with a checker pattern).

The second part is that the two camera's are spaced apart and both have a slightly different image when focused on the same object. This will also give you depth information. When you combine all this information to build a 3d map of the surroundings, you have succeeded into the first stage of a future robotic system.

Try to build a rough wire frame map form that data. Then use a feedback software algorithm using the two ranging systems to fine tune the whole imaging system.

You could use for example two beagle boards for such a development or two raspberry pi boards.
One board for each camera. And use the ethernet connection to exchange data between the boards.
It has a 3d accelerator on board and a display out as well to check the results.

If you succeed, you will have developed a system that is able to retrieve 3D information from just two camera's. After that you can continue to do image processing and recognizing and storing objects.

You are going to need to work with people who are good in mathematics and the basics of creating 3d game engines.

EDIT :
One more thing if you succeed, make the concept open source under GPL. But keep the specific software algorithms for yourself as a patent.
This will stimulate the robotic development in general but since you already did the ground work, you and team are going to be first to market.

 

[William Gaatjes]

If you desire something less extreme, just start simple.

This website might give you also some ideas :
http://www.societyofrobots.com/programming_computer_vision_tutorial_pt2.shtml