MEng Project: Vision-Led SLAM
I know Ive missed this competition but I am about to start a big project and wanted to share my ideas with some like-minded individuals who may have some suggestions.
I am about to embark on a 4th year university project involving doing something clever with an iRobot Create. I am looking towards approaching the problem of location recognition in mobile robots by attaching a command module and a webcam to the robot and using computer vision (CV) methods to identify landmarks.
I thought an interesting goal would be to get the robot to map a floor of my department building. After that it might attempt to classify distinct rooms. There is a white board on the wall of a corridor that the robot could potentially locate and possibly draw on with the use of a robotic arm. The ultimate goal of this project could then be to get the robot to draw a map of the floor on the white board (perhaps a bit excessive).
I am aware this project would involve implementing SLAM, integrating the map data with feature information from CV-processed real-time images and developing some kind of recognition determination model (i.e. neural networks or something).
After that I clueless. I have never even looked at a robot before, but Im really excited about finally having a go.
I would really appreciate it if anyone has any thoughts or relevant experiences or warnings :) then please let me know! Also, any other ideas for what I could do with the iRobot, a camera and 6 months ;)
PS Ive never seen this site before and it is awesome! I want to read everything :D