He is now at an Assistant Professor at Rice University at the Multi Robot Systems Lab working on the r-one robot (see figure). The target price is about $200. That is cheaper by a factor of 10 compared to the robots he worked with at iRobot. It also makes them inexpensive enough for each college student to have one. The big difference between the r-one and most other robots on the market is cooperation. The r-one is designed to work with its sibling, lots of its siblings.
The r-one design is the culmination of a lot of research. It is rugged, expandable and cooperative. It is about the diameter of a DVD and fits nicely inside a Tupperware container. There are two circuit boards. The one on top has 20 LEDs, five around each of three push buttons. These include red, true green and blue LEDs. There are 4 IR LEDs in the center that can be used in conjunction with room cameras that can be used to detect the position of many robots at one time. Four LEDs provide orientation and sequencing them provides a way to uniquely identify a robot.
The LED selection and identification are just the kind of details that experience brings. An LCD or OLED display would actually be less effective when multiple robots are in the mix. A display is handy for close work with a single robot but difficult, if not impossible, to see at a distance. LEDs, especially blinking and colorful ones, make things easier to debug. The robots also have a sound chip providing another dimension.
Need more? There are two circuit boards in the system held together by four screws. In the final version, two will be slightly taller because they can be connected to charge the 2000mAh battery. By the way, the battery is an easily removable cell phone battery.
The top circuit board has the buttons and LEDs on top along with a JTAG port and an expansion port. The USB interface can also be used for charging. Underneath the top board is an array of IR LEDs and sensors. These can be used for distance and obstacle detection as well as communication and identification of other robots in the vicinity. There are light sensors as well so they can follow a flashlight. That is one of the first experiments for students.
The robots have 3D gyro and 3D accelerometers. The plastic skirt actually floats atop the bottom circuit board that has switches that detect contact with an object.
The IR communication via the sensors under the top circuit board is for nearby neighbors. The 802.15.4 wireless interface provides longer range, mesh networking support. It is possible to use ZigBee but that is probably overkill for most applications.
The heart of the robot is a 50MHz Texas Instruments Stellaris microcontroller. This Arm Cortex-M3 micro provides plenty of processing power for the built-in systems but add-ons such as cameras may require additional processors. For example, some students are working on interfacing the r-one to a Gumstix board (see Stick Gets A New Bus). There is a Texas Instruments MSP430 to handle power management and control chores.
Software is still in a state of flux but there is a Python interpreter running on robots. Developers are using Code Sourcery's Eclipse-based IDE. Code Sourcery is now part of Mentor Graphics. There has been some work using FreeRTOS but an r-one OS is in the works. Software libraries for features like IR communication and 802.15.4 communication is available but not supported. Like I mentioned, this is just the beginning and it doesn't pay to lock down APIs while development is still in a state of flux.
Texas Instruments provides StellarisWare in ROM on the Stellaris microcontrollers. Developers have access to this support. James notes, "It's the first software I have not been compelled to rewrite."
The r-one project is still a work in progress. Right now they are tackling ESD issues. These robots need to operate in a range of environments and need to be rugged. They are expected to last more than a semester and students are not always kind to poor robots.
The r-one was the winner of Proto Labs Cool Idea Award Program. James and company have been taking advantage of Proto Labs CNC and injection molding process. They were able to prototype the bump sensor with less than a week turn around.
Rice University's r-one press release provides more details and the following video.
Swarm robots were always a lot of fun to watch and the research is very interesting. The challenge has always been making it available to more people. Now it will be practical for students. A common platform offers the opportunity to exchange research. This is something that is a challenge with most platforms currently available.
I'll keep you up to date on the advances of the r-one. I know I am looking forward to seeing how it progresses. Swarm research is already being done for a range of application areas including the military. The r-one makes it practical to teach in school.