Building a Pi Camera Mini-Tankbot with a Raspberry Pi, Arduino, and ROS

Building toy robots is always fun. I had a Raspberry Pi and a Pi camera lying around, the small form factor, power and weight with a linux core was just begging to be roboticized.

This miniature tankbot takes advantage of ROS and rosbridge to get communication up quick and allow tele-operation over the internet using either a web GUI, android app, or a google glass app. The camera also has pan-tilt control using 2 servos.
Mini Robotic Tank toy
We could do everything with the raspberry pi, however I had 2 micrometal gear motors that run on 5V and comes with a Zumo kit designed for an Arduino. Arduino with the kit it is, used as the embedded real-time controller.

Architecture
The 2 motors have optical encoders soldered on. The Arduino keeps track of the encoder counts for wheel velocity as well as runs the motors. It also controls two servo motors that holds the raspberry pi camera, and 1 LED as a flashlight.

The Arduino is connected to the Raspberry pi over USB and controlled through the serial port using rosbridge. This is just a simple hack, since we could have wired over SDA/SCL pins instead (providing for 3.3V-5V interfacing between the two boards), or even had the motors and encoders controlled by the Pi through a 3.3-5V interface.

With rosbridge we can create nice little publishers for the wheel velocity data.


// Publish left and right wheel encoder ticks
void publishWheelVelocity() {
  // updateWheelPositionVelocity();
  // encoder velocities are in rev/sec, convert to m/s
  wheel_vel.data = encoder_left_velocity * METERS_PER_REV;
  lwheel_pub.publish(&wheel_vel);
  wheel_vel.data = encoder_right_velocity * METERS_PER_REV;
  rwheel_pub.publish(&wheel_vel);
}

We also publish the battery voltage level, and subscribe to commands to update our velocity.

// Command Velocity Subscriber
ros::Subscriber<geometry_msgs::Twist> cmd_vel_topic("cmd_vel", &cmd_vel_callback);

The Raspberry Pi is connected to a raspberry pi camera and the arduino. We use mjpeg-streamer to view the camera stream over the internet, we provide a javascript interface to also control the robot remotely. The pi communicates over a web socket with a main computer or laptop using ROS via rosbridge, which has a Javascript interface as well.


 
// Publishing a Topic
var setPos = new ROSLIB.Topic({
  ros : ros,
  name : '/pantilt/set_position',
  // messageType : 'pantilt/pan_tilt'
  messageType : 'rosserial_arduino/pan_tilt'
});

var setLEDs = new ROSLIB.Topic({
  ros : ros,
  name : '/set_leds',
  messageType : 'std_msgs/UInt8MultiArray'
});



Stuff used (top to bottom):
  • Raspberry Pi + Camera
  • 15000 mAh Anker Battery
  • 3D Printed Chassis
  • Arduino Uno
  • Pololu Zumobot Chassis/Motors + Encoders
  • Hot Glue/Zip-ties/Optimism

Source code on Github

Popular posts from this blog

Visualizing TLE Orbital Elements with Python and matplotlib

Image
A Two-Line Element Set (TLE) is a way of encoding orbital elements of Earth-orbiting objects at a given time. This is the format NORAD provides data on many satellites in space right now. It'd be nice to see that as a 3d visualization. We use python to parse text files pulled from NORAD, and plot with matplotlib . Military satellite orbits For example the TLE for SpaceX's Dragon spacecraft during the CRS-2 mission was dragon = """DRAGON CRS-2 1 39115U 13010A 13062.62492353 .00008823 00000-0 14845-3 0 188 2 39115 51.6441 272.5899 0012056 334.2535 68.5574 15.52501943 306""" orbit.py  contains the code to parse the TLE into components as well as call graphics to draw it. print "Satellite Name = %s " % title print "Satellite number = %g ( %s )" % (satellite_number, "Unclassified" if classification =
Read more

Strawberry DNA extraction and viewing with a cheap USB Microscope

Image
I had recently bought a cheap USB microscope  (10/60/200x magnification) and thought a fun use for it was to try and see DNA! DNA is really small though, so we can't expect to see the components of it with a cheap microscope, however it is very long, and it turns out strawberry DNA is the easiest to see and extract. Microscope with DNA on a slide I followed this DNA extraction lab by Steve Spangler , which was very easy and fun to understand. I took an album of videos and photos during the process. We can clearly see strands  10x Magnification - DNA strands clearly visible within slide As we zoom in further, we can see the clumps of DNA strands start to show up. 60x Magnification - we can see some foreign debris in the slide, the strands look like spiderwebs. One thing to note, the microscope puts out color images and default camera settings, which with a backlight makes it pretty hard to see anything. I added a light source to the side as well as pushed
Read more

Relearning web development by writing a clicking game using React, Npm and Browserify

Image
I had some free time this vacation, and came across this interesting little sandbox game called  Pokemon or BigData by pixelastic . The aesthetics of the code are clean and simple. It uses React , which seems to be used for real-time web UIs. It also uses  npm  as a package manager. My previous forays into async calls in web dev were not nearly as pretty, so this seems like a good opportunity to learn something new. I decided a fun first start would be a simple game, and the simplest game I could think of was clicking a button to make a count go up. To give it some real-time feedback, I added some upgrade buttons that automatically click for you, a common incremental game technique. The buttons don't show up until there are enough clicks, and are grayed out if there's not enough to further upgrade. I set it to run at 50Hz to feel responsive. The heart of the code is basically a state machine that contains the current number of clicks and the number of upgrades
Read more