Skip to main content
main-content

Über dieses Buch

This textbook results from a series of lectures concerning autonomous mobile robots which have been held at the University of Kaiserslautern - tween 1999 and 2009. Methods and algorithms are introduced which can be used for developing complex autonomous land vehicles. Starting from hist- ical remarks and application areas of service robots, the vehicle kinematics modeling is introduced and examples of the drive kinematics of di?erent vehicles are given. Thereafter, typical sensors and sensor systems are - scribed which are used to determine the internal state of the machine and its operational environment. Localization, i. e the determination where the robot is, is still a di?cult problem. In the textbook, several methods are discussed which can be used under speci?c preconditions. Map building as well as navigation strategies complement the set of basic methods. The last two chapters deal with the questions of how to compile the above mentioned methods using powerful control architecture and what frameworks to use to support the development process. This textbook is written for beginners and advanced students from the ?elds of computer science, mechanical engineering, and electrical engine- ing, specializing in autonomous mobile systems. The book is also suited for engineers with a special interest in the development of wheel driven service robots. The writing of the manuscript was only possible with the assistance of several researchers of our Robotics Research Lab. Special thanks to Seb- tian Blank, Tim Braun, and Martin Proetzsch for proof-reading and editing.

Inhaltsverzeichnis

Frontmatter

1. Introduction

Abstract
Industrial robots, which are among the most important elements for industrial automation, are the biggest commercial market for robotics. Since the 1970s more than one million units are in use in fields like car manufacturing or the chemical industry. These robots are operating in highly structured environments. The tasks performed by this type of robots are monotone and restricted due to a low exibility. An economic growth of the robotics industry will only be achieved if the systems become mobile, adaptive to a dynamic environment and can be used for different tasks. The robots mentioned above belong to the class of service robots. A service robot can be defined as a system which operates semi- or fully autonomously to perform services useful to the well-being of humans and equipment, excluding manufacturing operations.
Karsten Berns, Ewald von Puttkamer

2. Kinematics

Abstract
In this chapter, the foundation for path planning and navigation of a wheeldrive robot is given. First the basic formulas, which allow to describe the motion of the vehicle in a 3D environment, are introduced. Then the solution to the kinematics problem considering different types of wheels fixed at specific positions on the vehicle platform is described. At the end of the chapter, geometric kinematics solutions for typical wheel-driven robotic platforms are presented.
Karsten Berns, Ewald von Puttkamer

3. Sensors

Abstract
This chapter introduces sensors often usedcome on autonomous mobile vehicles. In general, a sensor or a sensor system transforms different kinds of physical values (e. g. a force or a velocity) into electrical signals. One can distinguish sensors according to the integration level (see also figure 3.1):
Karsten Berns, Ewald von Puttkamer

4. Localization

Abstract
For autonomous navigation, a mobile robot needs to consider its position and orientation within a certain working coordinate frame. The so-called localization problem is usually solved by a mixture of different principles contributing to the state variables of position and orientation, called “pose” in the following. Generally two different approaches are being considered: absolute pose determination and relative pose determination. For each time step, the incremental observation of velocities, angular velocities, forces, or visual characteristics give an information update for the pose variables. Absolute information derived from global landmarks like GPS, however, directly provides position information. In the following, the important variants of pose measurement as well as the required principles of selected sensor systems are described. Also some exemplary implementations are explained in more detail. There are several possibilities to group the different techniques, such as regarding absolute or incremental information or position and orientation sensors. However, it seems most logical to present the measurement principles in an order they can be put together as a whole localization system, supplementing each other.
Karsten Berns, Ewald von Puttkamer

5. Mapping

Abstract
Whenever a mobile robot is required to navigate beyond its sensory horizon, it must either rely on potentially ineffective or misleading local search strategies (such as the ‘bug algorithms’ [Lum87]) or use some kind of world model to store cues for navigation. Such a world model is generally called a ‘map’ and can either be provided a priori or built online using a mapping algorithm. The mapping approaches can be separated into world-centric or robot-centric. World-centric systems represent the pose of all objects including the robot of the environment according to a fixed coordinate frame. In indoor scenarios, a corner of a room or a fixed position in the entrance area of an apartment is often used. To specify positions in the operational environment of the robot in outdoor applications, global coordinate systems like the latitude, longitude, and height system, the Earth Centered, Earth Fixed Cartesian coordinate system, the World Geographic Reference System or WGS 84 (GPS) are often used. World-centric mapping is mainly employed for tasks like navigation or path planning while robot-centric approaches are used for piloting tasks such as collision avoidance. Using matrix-based coordinate transformations, it is possible to convert between these different reference frames.
Karsten Berns, Ewald von Puttkamer

6. Simultaneous localization and mapping (SLAM)

Abstract
In chapter 4 the localization problem is introduced, which is the estimation of the position and orientation of the amr in its environment. It is shown that this problem can be solved with specific sensors or based on specific features of the environment. The selected features are those which could easily be detected by the robot sensor system. Additionally, chapter 5 describes different map generation techniques, for which a precise position and orientation of the robot is necessary.
Karsten Berns, Ewald von Puttkamer

7. Navigation

Abstract
The aim of navigation is to drive the vehicle through its environment. This task splits into three different subtasks: The global path planning deals with finding a suitable path from a starting point to a goal point using a given representation of the environment. The local path planning defines path points taking into account the vehicle dimensions and kinematic constraints. Path control describes the task of generating suitable steering commands for following a precomputed path represented by reference points.
Karsten Berns, Ewald von Puttkamer

8. Control Architectures

Abstract
The generation of designated robot behavior is one of the most difficult problems when designing the control system for robotic applications with many sensors and actuators. Due to the diversity of tasks an autonomous vehicle has to fulfill, the control has to be embedded into a convenient framework. The process of building up a control system should be supported by an adequate methodology to help overcoming difficulties common to complex robotic systems, e. g. ensuring secure operation, modularity, or handling a system of growing complexity. Therefore, different types of control architectures have appeared with contrary approaches for tackling the emerging problems.
Karsten Berns, Ewald von Puttkamer

9. Software frameworks

Abstract
Developing software for autonomous robots from scratch is a complex, timeconsuming and error-prone task. There are many issues that need to be dealt with, including hardware access, modeling of the environment, behavior synthesis as well as providing convenient debugging and teleoperation facilities. Especially in larger projects, the software needs to be clearly structured in order to stay maintainable. Ideally, software entities can be easily reused in other projects. Software efficiency and fault-tolerance are further critical aspects.
Karsten Berns, Ewald von Puttkamer

Backmatter

Weitere Informationen