Software Reading List

RAS Region V Robotics Software Reading List

Languages

• Python 3, for vision and control
  • A few (many actually) Python tips - watch this for some tips on how to be more 'pythonic'
  • Take a look at threading and multiprocessing
• C++ / Arduino
  • May also be used to implement computations used in vision or decision making algorithms, and called as Python modules
• Rust?

Frameworks

• OpenCV (See below)
• Numpy (with emphasis on linear algebra)
• Possible web server run using flask or django; this server would be used for controlling multiple robots or providing a debug output.

OpenCV

OpenCV will be used to identify obstacles, cubes, the mothership, etc.

• Good general tutorials on OpenCV for Python

• Relevant algorithms:

  • SIFT: Scale Invariant Feature Transform

    • a feature detector specializing in edges
    • old-ish (2004) but still quite effective
    • good tutorial here

  • SURF: Speeded Up Robust Features

    • feature detector specializing in regions of colors (a blob detector)
    • faster than SIFT, and better in many circumstances
    • tutorial here

  • NOTE: in order to use SIFT / SURF, you will need to build OpenCV with the opencv_contrib package; also in the tutorial, instead of creating a SIFT/SURF detector with cv2.SIFT() or cv2.SURF, these detectors have been moved to xfeatures2d:

   sift = cv2.xfeatures2d.SIFT_create()	# Returns SIFT object
   surf = cv2.xfeatures2d.SURF_create()	# Returns SURF object

  • Hough line transform / circle transform

    • Detect lines and circles in an image
    • Tutorial for lines
    • Tutorial for circles

  • QR Code scanning with ZBar (an external library)

    • Tutorial
    • If we use QR codes as identification, version 1 QR codes or Micro QR codes will be used

General Vision techniques

• HSV and RGB color space (NOTE: the default color space for OpenCV is BGR which is literally RGB backwards, because reasons)
• Dynamic threshholding and K-means

Miscallaneous Algorithms

• PID controller
• Markov chains and the (more useful for robotics) Markov decision process
• Monte Carlo Tree Search

Control Algorithms

• SLAM (Simultaneous Localization and Mapping)
• Combinatorial Filters
• Application of Kalman Filters to SLAM / Extended Kalman Filters for SLAM
• Range-only SLAM with indistinguishable landmarks
• Bearing-only SLAM with indistinguishable landmarks