• VISUAL MONTE CARLO LOCALIZATION
    WITH ROBOTICS AND iOS ODOMETRY
    Fully wireless robotic localization and navigation.
    Built using Python and OpenCV.
    GitHub Code
    Video Demos

Accessible Robotics

Completely untethered robot with WiFi capabilities.
Built on an easy to use coding platform.

Accessible Platform

In order to maximize accessibility, we deployed commercial grade robots that can be easily modified, and wrote all scripts in Python.

Fully Wireless

Using Raspberry Pi along with a WiFi adapter, the robot can sense and navigate from the commands of a remote computer.

Unified Power Source

All systems share a single power source with the Create iRobot, eliminating the complication of having multiple batteries.

Monocular Visual Monte Carlo Localization

Using monocular vision and a suite of image matching algorithms, our implementation of the Monte Carlo Localization algorithm can robustly and accurately localize a robot given a map of the surroundings and an image sequence.

  • Fully image-based localization.
  • Support for different image matching algorithms, such as SURF, SIFT, and ORB.
  • Added robustness by accounting for motion blur with the variance of the Laplacian operator.
  • Dynamically Optimized Retrieval decreases the span of the image search and improves speed.
  • Bag-of-Words implementation uses machine learning and NLP techniques to cluster features into visual words and greatly reduces runtime.

Autonomous Navigation

Using our Visual Monte Carlo Localization algorithm, as well as the external odometric data provided by the iOS device, our robot can navigate a map of its surroundings. First, it performs a predetermined set of movements and captures images. Through our Monte Carlo particle filter, the robot can localize itself and determine its initial position. Then, using external odometry and success localizations, the robot can navigate to a user-defined position in the map.

Python Robotics Platform

Scripts written entirely in Python allow for complete wireless control of Arduino robots or Create iRobots, as well as access to wireless cameras and iPad odometric data.

Github Repo

Visual Monte Carlo Localization

Complete image-matching suite that allows robust localization within a visual map, with support for different matching algorithms and optimization techniques.

Github Repo

Fully Autonomous Navigation

Using our robot and localization algorithm, the robot can autonomously navigate through its surroundings using only visual data from its camera.

Github Repo

Pictures

Videos