Autonomous navigation in a known dynamic environment

Autonomous navigation is nowadays a popular field of research all over the world due to the great diversity of applications for which it can provide a solution. Most of these involve a need to carry out manipulations in areas where human presence is impossible, highly dangerous or otherwise impractical. For example NASA’s Sojourner, the small explorer robot that was used by Mars Pathfinder in 1997, found traces of water on the surface of Mars (see http://mars.jpl.nasa.gov/MPF/), a fact of great importance which would have been practically impossible to learn without the use of an autonomous robot. There are, however, some less extreme jobs that could be performed by robots but today take up valuable time of humans, for example office delivery, certain maintenance tasks, etc. The navigation process in these cases is quite similar: there is a known, well structured environment, which is also highly dynamic: humans – or other robots – may appear at any time. This similarity makes it possible for us to consider the task of autonomous navigation within a building in general, regardless of the specific purpose for which the robot is used. The navigation system of an autonomous robot has to decide at every given moment where to move next, taking into consideration all the available information in an intelligent way. To this problem, several solutions have already been proposed, which can be categorized as global if the algorithm relies mostly on a priori information provided on the environment, or local if its decision is taken using mainly the current sensor data. Both approaches are appropriate for several tasks, but for our case, neither is sufficient in itself. The method of navigation proposed in this paper combines the advantages of these two approaches and thus uses both types of information in an intelligent way. This is achieved by using a hybrid navigation system composed of two modules, one of which uses the a priori information and determines roughly the optimal route towards the goal, whereas the other carries out the actual navigation decisions using a local approach. This latter module may also be used independently if part of the environment is unknown.