So the Cellbots team were at Google’s Cloud Robotics shipping container, prepping a demo of an “Antlerbot” – an RC car controlled by a Nexus One connected to a Sparkfun IOIO board – when disaster struck: the Antlerbot ran up under one of the doors of the shipping container, which was just too high for the sensors – but just low enough to hit the upper chassis smack in the window that showed off the IOIO board … shorting out the car beneath.

So Chaitanya, Arshan, and Charles worked hard to get the bot running again, transplanting the brain from the shorted bot over to the larger Antlerbot which was going to be running on a tablet. After some surgery, they got the bot rewired, but the output was still bad: the IOIO was fried too.

Then they replaced the IOIO board, and still no luck: the new board wasn’t outputting the right signals. It turns out the board was on the wrong firmware version! Chaitanya, the Antlerbot developer, had to go, so Charles, the Cellbots code lead, took over. The source control server with the right code was down, but JP, a random Googler, swooped in to save the day with a laptop full of the right source code he’d downloaded just that morning.

That still didn’t do the trick, and it wasn’t until next morning that Arshan diagnosed the problem: the old brain on the new frame was having brownouts. Once that was fixed, the Antlerbot was free to roam the Maker Faire and delight children and adults alike:

So we had real making at the Maker Faire! Go Cellbots!

UPDATE: Charles pointed out we left out one of the coolest parts of the Sunday heroics. Charles discovered that he could get a split second of control after
connecting the robot before the controls would fail (AKA the “life twitch” moment).
The disconnection looked normal on the Android end, and Arshan correctly guessed that it was a brownout issue caused by the steering controller which was disconnecting the IOIO board.

However, there was still the big problem: “How do we do this without steering?” Arshan’s solution was to hack a USB cable so that we could use a spare portable USB charger to power the steering controller and take it off the powersource that the IOIO was running. So the Antlerbot was actually running off of two separate batteries with USB cable that had been modded right there. Hacky, yet very cool, and the type of resourcefulness sums up the Maker Faire spirit quite well.

Incidentally, this second power supply was the reason the bot began to fail to turn at the end of Sunday: Anthony wasn’t there for the second hotwiring, and therefore didn’t know he needed to charge that battery before the final run. Still, it was a great success and we look forward to next year!

The Perception Manager is a Java class that abstracts the raw Android Sensor API into higher-level ‘percepts,’ effectively translating hundreds of samples a second from the accelerometer and gyroscope into binary features like “shaking” or “upside down”.

Yes, you could write these yourself – but why should you have to? Furthermore, the PerceptionManager has support for higher-level sensors like “movement in space” or “vertical motion” which you can use to build up your own percepts, or math functions for Verlet integration and vector math you can use to develop your own sensory processing.

To test it, the PerceptionManager is embedded in a PerceptionTestbed application which you can run on an Android phone to see both the raw data from the sensors and the extracted percepts. If you don’t care about the percepts, you can customize the view to hide them, or change the sample rate. You can also turn the gyroscope on or off; on some phones the gyroscope cannot be restricted to a sampling rate and can crash the app, so be warned. Below I discuss the PerceptionTestbed app and how to use its interface.

The PerceptionTestbed is not available on the Market, nor is it fully integrated into Cellbots yet; it’s a 2.3 app and Cellbots is 2.2. Over the next month or so we’ll be working on a new release of Cellbots which will incorporate the PerceptionManager, enabling you to write code to have your Cellbots stop their motors if they flip over or get shaken. We’re still working on the documentation, and the code itself is likely to change rapidly as we integrate it into Cellbots, but in the meantime, you can still check out the source code for the PerceptionManager from:

http://code.google.com/p/cellbots/source/browse/trunk/android/java/perception/

If you’ve not worked with the Cellbots Java codebase before, you can use the instructions here and read more about the Cellbots Java app here. We hope this software is useful to you … happy cellbotting!

-Anthony

P.S. Sorry, we have not tried crushing the Hasbro toy with the Cellbots tank … we think they should be friends.

Hello, all … there’s been a lot going on behind the scenes at Cellbots (check out the recent changes in our source code repository, with more in the works) and this weekend you’ll get a chance to see what we’ve been working on.

At Maker Faire this weekend in San Mateo, several of the Cellbots volunteers from Google’s Cloud Robotics team will be at the Google booth demonstrating a variety of RC cars controlled by the Cellbots app.

We’re not releasing v2 of Cellbots yet as it is under active development, but you can see a preview of the future at the Maker Faire!  Please join us!

Merry Christmas from Mike and the rest of the Cellbots team!

The app also works with LEGO MINDSTORMS, VEX Pro, and the custom Arduino cellbots and is available free in the Android Market for any Android 2.2 (Froyo) and up phone. After watching the video, download the app from the Market, or read more about how it works here, and of course download the source code if you want to modify it with more features and robot platforms.

Squirt can either be in guard mode or phone control mode. In guard mode it searches for moving objects and squirts them. In phone control mode it is driven like a RC car. Steering can be by software joystick or by tilting the phone in the direction the robot should go.

The big red dot in the middle is a software joystick for driving. The red lines indicate distance senor values from the robot. If the green circle icon is pushed, the big red dot turns blue and becomes a pan/tilt joystick for the water gun. The orange circle with red dot is the water gun trigger.

Squirt can also be controlled by SMS text message from any cell phone. This is accomplished by using the Android phone to relay the message to the robot. For instance, if you send my Droid phone a message where the first character is a period, it will command the robot to fire the squirt gun.

The robot has a timid personality. It verbally comments on external events and on the human’s driving skills. In general, Squirt is wary of things sneaking up behind it.

The water gun is a slightly modified SIG SAUER STR70 Saturator electric water gun. It shoots 4 “bullets of water” every second; range is 6 meters. The amount of water is small and unlikely to damage even the most fragile of plants. Interestingly it features “realistic sound”. To me it doesn’t sound so much like a machine gun but more much like a powerful electric water gun, which it is. The noise is loud, which hopefully can scare raccoons away from tomato plants. How Squirt Works: Squirt uses 3 onboard processors plus a processor in each servo. The master processor is an Arduino Mega which is used for AI, navigation and as a master controller for the sub systems. The second processor is an ARM 9 board for controlling the high speed communications to the servos, last is a graphic processor for the OLED Display.

The Arduino Mega sends and receives data from the Android phone via blue tooth. The Arduino is constantly sending sensor and AI information to the phone. The robot receives driving, servo controlling and mode setting data from the phone. Personality functions are carried out both on the phone and the robot.

The water gun fires by using a relay to bypass the trigger circuit. A 3 color 1 watt LED has been mounted on the barrel to display status of the mood of the gun. Red, of course, means its upset and likely to fire shortly. Summary: Squirt demonstrates that a smart phone and a robot can work together to accomplish tasks. The next step is to design an simple irrigation system based on the technology, get philanthropic funding and hopefully help grow food.

Check out more on the construction overview and videos page for Squirt.

Fun Video:

Contact info for Michael Winter: robotmike@comcast.net

We have some video streaming from the phones to the big screen TV and people can also use the remote to trigger still pictures to be snapped. The demo shows off the use of Android Nexus One phones as remote controls using the accelerometer to drive the robots. That sends XMPP commands over Google Chat to another Nexus One on the robot, which act as the main brain. That phone then sends commands over Bluetooth to the robot hardware.

We hope the excitement continues and we see more kids making Cellbots this summer. Our parts lists for the Android robots and the source code project site can help you get started. Those short on time can also check out our robot kits page where we can put together a package for you. Then be sure to jump in on the discussion list to ask questions and share news of your projects!

Getting this demo ready was a big team effort that took a lot of work to go from short five minute demos to having four robots run all day for a week. Damon Kohler came through with awesome upgrades to last week’s Android Scripting Environment. Charles Chen was extremely helpful in getting the video and still image streaming setup, and used some code from Darrell Taylor’s work. Jason Holt, maker of the Boxbot, helped with the remote control mathematics and getting the demo table ready. Tim Heath upgraded the kits page for those of you interested in using one of our designs, and Glen Arrowsmith’s recent Arduino code that saved calibration data to the EEPROM was critical. Many thanks to you all!

If you are in the San Jose area this week, be sure to stop by the convention center on Thursday May 13, when the International Science and Engineering Fair is open to the public from 9am to 9pm.

The only tricky part was getting a thin 30 gauge wire to run into the back of the phone and wrap around the positive battery terminal. We tried some thicker wire at first but it prevented the phone from maintaining a connection and it would shut off accidentally. The thinner wire sits there nicely and we might try making a permanent connector on the outside of the phone so it can fit back in a pocket without a wire sticking out.

The top picture is a working version with the red 30 gauge wires above is an early test with a thicker green wire that had a flaky connection. You can see that we’re still using the HTC breakout board wiring for TX, RX, and ground, and we’re not yet connecting the fourth wire that you see here. That would be used for charing the phone but doesn’t provide enough power out to run the robot.

This video overview shows you the whole thing moving around to prove it works. No word yet on the effects of doing this on your battery so please do so at your own risk.

If you get something similar working, share it in the Cellbots support & discussion group. Bonus points if you hook your robot up to a wireless charging station!

The latest version of the code supports two range sensors, where one is intended to move forward and another looking down at the ground. These are both optional but should help prevent your robot from collisions and dangerous tumbles. It is especially handy if your streaming video and controlling the robot with some high latency.

We’ve tested with this Maxbotix and Parallax ultrasonic sensors, and we’ve also ordered some infrared proximity sensors and very small reflectance sensors. Our goal is to have the robot not bump into things or fall off ledges and we’ll likely need somewhere between four and six total sensors to do that. We want to keep the price down so we’re trying different options on our Cellbots and will share the results with you here.

We did this by turning the phone on its side since Bambuser works with the phone in landscape mode, while the Truckbot was designed for it to be in portrait. Then we created a simple HTML page that includes an embedded Bambuser video player, and a Google Talk gadget. We can now type in commands via XMPP chat while watching video streamed back from the robot.

The latency is kind of high at around one second, which makes it a challenge to drive in hazardous conditions like table tops or crowded places. We’re interested in finding alternative video streaming methods  since Qik and Ustream both have about the same delay. Nrover is using a modifcation of Movino, which we’d love to see compiled for Android in addition to what we’re doing on Symbian.

We’ve posted the source code to the remote control page to save you a few minutes from making your own. It really is easy to get started by running the Cellbots Python app on Android, and then kicking off Bambuser. We’re working on scripting this as well so the robot can initiate video streaming from a remote command. Enjoy chasing your cat!