There are some really great open-source Python libraries for UHF RFID tags (battery-free tags that can be read from 6+ meters away): http://www.ros.org/wiki/hrl_rfid

These form the basis of my PhD work. Turns out you can do some cool stuff when you couple it to an autonomous, mobile, manipulating robot: http://www.hsi.gatech.edu/hrl/pdf/pervasive_computing_2010.p... (PDF)

A couple of questions... How do you get rid of the signal rebound due metallic presence in the robot?

provides an object a unique ID with extremely small uncertainty, whose location is inferred by lowerlevel sensory processing

How precise is your method to locate the object? Are you using an array of antennas to calculate the position of every RFID?

Multipath due to the robot's body is certainly a design issue. Generally, you try to use highly directional antennas facing away from the robot so that the influence is minimized.

The particle filter implementations for localization use the motion of the robot or panning antennas to integrate the otherwise sparse signals into something more meaningful. This is analogous to using an antenna array. Still, with these implementations (under the best conditions), you're talking about an error of ~20-50cm on average with a pretty high variance -- meaning other sensors are generally required, though our near-field antennas allow you to hone in on the position even better using the same tag.

Meanwhile, local methods (eg. servoing from the difference between two antennas) works quite well. Publication(s) forthcoming. ;)

The Arduinos are great. I have two of them. One is a DIY clone called "DorkBoard". Check it out. Also have a look at the MSP430 based LaunchPad kit, its very cheap and available from Mouser.com.

Yeah, I have 4 MSP430's, but they're a little tougher to program

Advanced features, but easy to use.