This project is a platform that lets me explore my technical interests. I've 3D printed robot parts and soldered a Raspberry Pi robot hat. Mostly I'm reading Sebastian Thrun's "Probabilistic Robotics" and I'm implementing here what I learned in the book and in online lectures.

This project is an enhanced version of my original robot that required BLE connection to a Mac which ran the main algorithms. The robot hardware is in large parts identical. It is very much a work in progress.

Hardware

• Dagu Rover 5 robot with offroad wheels (for the looks)
• Neato XV-11 Lidar
• Teensy 2.0++ microcontroller as motor interface (recommended for the large number of I/O pins, especially interrupt-enabled pins),
• connected via USB to an up-board (quad-core Intel Atom x5-Z8350)

Software

• arduino/ contains the microcontroller code and the header with the data definitions shared between the microcontroller and the Raspberry

  • It uses the http://platformio.org toolset to build and deploy the microcontroller software directly from the Raspberry
  • Uses PID motor control library: https://github.com/br3ttb/Arduino-PID-Library

• raspberry/ contains the more "interesting" parts of the code

  • It uses CMake as a build toolset and requires boost >= 1.55 and OpenCV >= 3.0
  • It includes 'libicp', an interative closest point solver from http://www.cvlibs.net/software/libicp/ which can use OpenMP if available
  • Build and run rover --help to get information about command line arguments. Currently, two modes of operation are supported:

    1. ./rover --port /dev/ttyUSBPORT --lidar /dev/ttyLIDARPORT --manual --map map.png tries to connect to the microcontroller on USB serial port /dev/ttyUSBPORT and the Neato lidar on port /dev/ttyLIDARPORT, let's you control the robot using the ERT - DG - CVB keys and outputs the current map with the robot's pose to map.png.
       
       The robot can also be controlled with a gamepad. Use e.g. nginx to host raspberry/html/map.html and open the page in a modern browser that supports the gamepad API, e.g., the current version of Chrome. If you have a supported gamepad, the website will send control commands to the rover executable which is listening on port 8088 for control commands. Use the --map argument to overwrite the hosted map raspberry/html/map.png regularly. This way, you can control the robot via the browser and see the generated map in the browser.

    2. ./rover --input-file log.txt reads the sensor data, runs a SLAM algorithm on the data, and outputs log.txt.mov. Useful for evaluating algorithms offline without powering up the robot.

  • raspberry/test contains a sample log file and sample outputs of the algorithms implemented in deadreckoning.cpp, particle_slam.cpp and scanmatching.cpp respectively.

The following is very incomplete. Haven't tried this out on a new machine in a while.

PlatformIO

pip install -U platformio

sudo apt-get install cmake

Install boost (>=1.55)

sudo apt-get install libboost-all-dev

Build OpenCV

See http://docs.opencv.org/3.0-beta/doc/tutorials/introduction/linux_install/linux_install.html

sudo apt-get install libopencv-dev

Install libmicrohttpd

sudo apt-get install libmicrohttpd

cd arduino 
platformio run

cd ../raspberry
mkdir build
cmake ..
make