Robot Control and Lane Detection with Mobile Phones

One cannot imagine daily life today without mobile devices such as mobile phones or PDAs. They tend to become your mobile computer offering all features one might need on the way. As a result devices are less expensive and include a huge amount of high end technological components. Thus they also become attractive for scientific research. This work addresses the exploration of mobile phone usage for robot control and lane marking detection. Basis for the work was an already available controller for mobile robots, named EyeBot. I developed an application, which allows to operate the controller from the mobile phone via a Bluetooth link. The results are promising and facilitate the usage of the mobile as remote control for other tasks. However, I detected some restrictions in terms of communication as the mobile phone was not able to process a continuous video stream fast enough. On top of the communication between the robot controller and the mobile phone, there are several applications possible, which might take over robot control. In this context I developed a sophisticated algorithm for lane detection, which uses the input data from the mobile phone’s camera. To address the special aspects of the lane marking appearance they exhibit a constant orientation over different parts of the recorded image I used the theory of steerable filters to get an optimized edge detection. Further on I developed a line extraction and clustering algorithm, which separates the binary edge detected image in several tiles. Lines in these tiles can then be described in terms of orientation of the eigenvector by performing a principal component analysis. The application was implemented platform independent to run on any system supporting C++. The outcome was an application on the mobile phone, which lends itself to be further used for lane tracking and direct or remote robot control.