ECE497 Project Beaglebone Blue Robotics
Embedded Linux Class by Mark A. Yoder
I'm using the following template to grade. Each slot is 10 points. 0 = Missing, 5=OK, 10=Wow!
08 Executive Summary - Good, but seems dated 07 Installation Instructions - Wiring instructions are needed. A picture would be very helpful. 09 User Instructions - Clear instructions. Nice screen shot 08 Highlights - Nice videos. I would have liked to see more of the web interface in action. 09 Theory of Operation 10 Work Breakdown 10 Future Work 10 Conclusions 08 Demo - I wish more of the web interface was demo'ed. 10 Not Late Comments: You got a lot done on this project. Score: 89/100
(Seems out of date.)
The stated goal for this project was to "do something interesting with robotics" using the BeagleBone-Blue as part of a final project for ECE497. We interfaced the Blue with an existing robotics platform and added a remote web interface. We developed a proportional-control object following program and had the robot follow a red ball within a small area using a pixy camera for image detection and continuous rotation servos for drive. We also developed a web interface for reading from the other sensors and driving the robot using a keyboard. A pixel grid on the web page shows the detected position of the object the robot is following in real-time*.
We ran into a problem using the motors built into the platform. They were designed for a 5v controller and the built in controllers on the blue use 12v. For a number of reasons we decided to use continuous rotation servos instead. Some mechanical modifications were required to convert the existing wheels on the platform to work with continuous rotation servos. This saved us from having to order an external h-bridge or motors.
To conclude: We had a total of 3 weeks to work on the project and the first half of that time was spent dealing with the quirks of the Beagle Bone-Blue. Due to the nature of working with an alpha version of the board there were a number of hurdles and workarounds to figure out just to get the board to connected to the internet and successfully communicating with the sensors and motors. The basic project goals were met but there is yet more to be done.
∗Not really, by "real-time" we mean about 5 times a second :)
- Experimental schematics for a beaglebone blue
- spark fun robot base
- Pixy Camera
- 3 IR distance sensors
- 2 Servos
- 2 continuous rotation servos
- Wires and connectors
- Follow git hub instructions to configure blue for robotics
- Follow the instructions for installing the pixycam libraries on the BeagleBone
- Follow the instructions for connecting with a Raspberry Pi over wifi, the process was the same for the Blue, your network may be different
- It's a good idea to set passwords for the bone at this point but its not critical
- Clone our github project to the blue
- Running setup.sh should install any needed libraries
- refer to our readme if any other problems are encountered
(How do I wire it?)
Once everything is installed
cd into the root directory of the project
Will start the wifi program in a tmux session called wifi, and will output the ip address for the blue once it connects
Will start the server program in a tmux session called server, be default it should be on port 8090
With any browser go to yourIP:8090, from there the robot can be controlled with the keyboard keys, and sensor reads can be done with the buttons. The ball following program can be started and stopped from the web interface.
The blue is able to read position data of objects from a pixy camera and display the coordinates of the object on a grid in a socket.io based web interface.
The coordinate position data also controls the speed and direction of the robot to track a brightly colored object.
Integrated a variety of devices and sensors into a web based application on the bone.
Theory of Operation
The theory for the project was rather simple, start a web server and have it make function calls based on input from the user of the website. Most of the work for the project was debugging libraries and hardware. The python server program makes calls to the socket.io library for communicating with the client computer and the robotics cape library for controlling hardware
Routine for image tracking:
1. The object to be tracked is programmed into the pixy camera using the built in detection button on the camera
2. There is a python script that interfaces with the libpixyusb library to communicate with the pixy camera over a USB interface.
3. The python script uses the object position to proportionally control 2 continuous rotation servos(via the robotics cape library).
4. The object position is written to a JSON file every 50 frames (roughly one second)
5. The webpage has a 127x127 gird. The webpage polls the JSON file every second through the socket.io interface and colors the appropriate grid representing the position.
List the major tasks in your project and who did what.
Also list here what doesn't work yet and when you think it will be finished and who is finishing it.
- Alvin's Tasks
- Build web interface
- Update Wiki page
- Interfaced connectors
- Integrate IR sensors, bump sensors, and gpio with the bone.
- Integrate servos
- Integrate IR sensors with the web interface.
- Created instructions and install files for the web server
- Made the server webpage look prettier
- Sam's Tasks
- Find and order connectors
- Install pixycam libraries on the blue and use position data to control servos.
- send and visualize object position data on the web interface
- Mount the blue to the robot
- retrofit continuous rotation servos onto the wheels we were given with help from Gary and Jack
- integrate servos
- Created instructions and scripts for setting up the pixy cam
- Setup a socket.io function call from the webpage to start the pixycamera script automatically.
- Made the server webpage look prettier (it's ugly and jumbled at the moment)
- incomplete tasks
- The motor drivers were not used because our motors operate on 5v rather than the supplied 12v. We decided that retrofitting continuous rotation servos was a better idea.
- PID control for the object follow script (currently only proportional control is implemented) would make the robots movement less erratic.
- Add a screenshot button and object recognition button to the web interface so that new objects could be programmed in remotely without being forced to use the hardware button on the pixy.
- implement a web programming interface like blockly to make the Blue accessible to younger students.
The BeagleBone Blue has some very good hardware built into it, one of the flaws with this project was that we just used what was really available to us for interfacing with it. It would have been a lot more fun to pick out parts and build a platform for it to fully utilize all its functions. The board still being in an experimental state resulted in some interesting hardware issues, the ones we encountered where that some times the 12 Volt jack is bumped and shuts down the entire board. If the blue boots from just USB power, and then the 12 volt power or battery is plugged in then a few pwm signal pins will stay at 3.3 volts, making pwm impossible on those pins, rebooting the blue solves this. Finally whenever the voltage rail for the pwm is started, nothing can actually be attached to rail, or It will fail to reach 6 volts, the solution to this is to start with everything unplugged and then plug them back in. Aside from this the blue worked as expected.
Embedded Linux Class by Mark A. Yoder