Self-Landmarking for Robotics Applications [chapter]

Yanfei Liu, Carlos Pomalaza-Raez
2011 Advances in Mechatronics  
Introduction This chapter discusses the use of self-landmarking with autonomous mobile robots. Of particular interest are outdoor applications where a group of robots can only rely on themselves for purposes of self-localization and camera calibration, e.g. planetary exploration missions. Recently we have proposed a method of active self-landmarking which takes full advantage of the technology that is expected to be available in current and future autonomous robots, e.g. cameras, wireless
more » ... eivers, and inertial navigation systems (Liu, & Pomalaza-Ráez, 2010a) . Mobile robots' navigation in an unknown workspace can be divided into the following tasks; obstacle avoidance, path planning, map building and self-localization. Selflocalization is a problem which refers to the estimation of a robot's current position. It is important to investigate technologies that can work in a variety of indoor and outdoor scenarios and that do not necessarily rely on a network of satellites or a fixed infrastructure of wireless access points. In this chapter we present and discuss the use of active selflandmarking for the case of a network of mobile robots. These robots have radio transceivers for communicating with each other and with a control node. They also have cameras and, at the minimum, a conventional inertial navigation system based on accelerometers, gyroscopes, etc. We present a methodology by which robots can use the landmarking information in conjunction with the navigation information, and in some cases, the strength of the signals of the wireless links to achieve high accuracy camera calibration tasks. Once a camera is properly calibrated, conventional image registration and image based techniques can be used to address the self-localization problem. The fast calibration model described in this chapter shares some characteristics with the model described in (Zhang, 2004) where closed-form solutions are presented for a method that uses 1D objects. In (Zhang, 2004) numerous (hundreds) observations of a 1D object are used to compute the camera calibration parameters. The 1D object is a set of 3 collinear well defined points. The distances between the points are known. The observations are taken while one of the end points remains fixed as the 1D object moves. Whereas this method is proven to work well in a well structured scenario it has several disadvantages it is to be used in an unstructured outdoors scenario. Depending on the nature of the outdoor scenario, e.g. planetary exploration, having a moving long 1D object might not be cost effective or even feasible. The method described in this chapter uses a network of mobile robots that can communicate with each other and can be implemented in a variety of outdoor environments.
doi:10.5772/21625 fatcat:uh6tqzlkzzdvjduy5nbwiymugi