Balrog information


The Balrog project started out 2009 as a collaboration between Linköping University and Saab Dynamics with the long term goal to develop an autonomous minesweeper. Over the years student groups have made different subprojects on the minesweeper. This year the purpose was to make the minesweeper capable of positioning itself and navigate in an indoor environment. poster can be found here


The Positioning module uses dead reckoning to estimate the motion of the Balrog, and AprilTag detections to get direct position estimates. The dead reckoning estimates are fairly accurate with low noise, but they drift over time. The AprilTags do not drift, but there are certain uncertainties. Therefore, the Positioning module uses solely dead reckoning when no tag is visible. When there are tag detections, these are used to slowly compensate for drift that occurred since the last AprilTag detection. This was implemented using an Extended Kalman Filter.



The Mapping module uses the LIDAR to implement a probabilistic obstacle map. The obstacle map is divided into grids sized 5 by 5 centimeters. Each grid has a probability that it contains of obstacles. This probability is decreased if the LIDAR beam traverses the grid without detection, and increased if there is a LIDAR detection in this grid. Eventually, most grids in the map converge to values close to 1 or 0. The mapping module also records which grids the Balrog has driven over and are considered explored.



The control module in the Balrog consists of two P-controllers, one that controls radial error and one that controls angular error. The reference is sent from the navigation module and fed to the controller via a coordinator function. Furthermore, the controller is tuned to turn first and then move straight to the reference position to avoid drift in the dead reckoning algorithm.


Our Group Members