Exploration and mapping with mobile robots

摘要

Models of the environment are needed for a wide range of robotic
<br>applications, from search and rescue to automated vacuum cleaning.
<br>Learning maps has therefore been a major research focus in the
<br>robotics community over the last decades. Robots that are able to
<br>acquire an accurate model of their environment on their own are
<br>regarded as fulfilling a major precondition of truly autonomous
<br>agents. In order to solve the map learning problem, a robot has to
<br>address mapping, localization, and path planning at the same time. In
<br>general, these three tasks cannot be decoupled and solved
<br>independently and map learning is thus referred to as the simultaneous
<br>planning, localization, and mapping problem. Because of the coupling
<br>between these tasks, this problem is very complex. It can become even
<br>more complex when there are dynamic changes in the environment or
<br>several robots are being used together to solve the problem.
<br>
<br>The contributions of this thesis are solutions to various aspects of
<br>the autonomous map learning problem. We first present approaches to
<br>exploration that take into account the uncertainty in the world model
<br>of the robot. We then describe how to achieve good collaboration
<br>among a team of robots so that they efficiently solve an exploration
<br>task. Our approach distributes the robots over the environment and in
<br>this way avoids redundant work and reduces the risk of interference
<br>between the individual team members. We furthermore provide a
<br>technique to make use of background knowledge about typical
<br>spacial structures when distributing the robots over the
<br>environment. As a result, the overall time needed to complete the
<br>exploration mission is reduced.
<br>
<br>
<br>To deal with the uncertainty in the pose of a robot, we present a
<br>solution to the simultaneous localization and mapping problem. The
<br>difficulty in this context is to build up a map while at the same time
<br>localizing the robot in this map. Our approach maintains a joint
<br>posterior about the trajectory of the robot and the model of the
<br>environment. It produces highly accurate maps in an efficient and
<br>robust way.
<br>
<br>In this thesis, we address step-by-step the different problems in the
<br>context of map learning and integrate our techniques into a single
<br>system. We provide an integrated approach that simultaneously deals
<br>with mapping, localization, and path planning. It seeks to minimize
<br>the uncertainty in the map and in the trajectory estimate based on the
<br>expected information gain of future actions. It takes into account
<br>potential observation sequences to estimate the uncertainty reduction
<br>in the world model when carrying out a specific action. Additionally,
<br>we focus on mapping and localization in non-static environments. Our
<br>approach allows a robot to consider different spatial configurations
<br>of the environment and in this way makes the pose estimate more robust
<br>and accurate in non-static worlds.