9. Autonomous Lawnmower Project (ALP) Test Videos
Post date: Feb 19, 2014 6:36:27 AM
My part of the Autonomous Lawnmower Project (ALP) is near completion. As described in project's webpage, the major goal of ALP has been to develop a cheap approach of robotic localization, the system by which a robot can "known" where it is in real-time, and implement the approach into a small-scaled version of a lawnmower, designated as the "moving-object". The approach to robotic localization actually implemented utilizes the computer vision algorithm Continuously Adaptive Meanshift (CAMShift) at its base.
The videos below demonstrate the moving-object completing two modes of autonomous operation, the first of which is the moving-object traveling to a series of green-colored way-points and the second of which is the Lawnmower Path Algorithm. Another video will later be available, and it will describe in detail what is happening in the videos.
In addition to obstacle avoidance, a recent goal of mine was the implementation of GPS and computer vision for robotic localization. The idea behind combining two (or possibly more) forms of robotic localization is using redundancy as a method to avoid failure and/or improve performance. For instance, say the autonomous lawnmower travels behind a building large enough to prevent the lawnmower's GPS from its signals. Strategically placed camera could make computer-vision an excellent potential solution.
With those ideas said, due to my other commitments, such as research and academic courses (and several other projects), the goals of obstacle avoidance and combining various methods of localization I have decided to leave to the next potential team who will continue the project (though they may very well take the project in a different direction, if they see fit)! For now, my major priorities in regards to ALP include fixing bugs in the software and completing the project's documentation.
More information will come!