Volume 61, Issue 4 -

This paper surveys research on fully actuated humanoid walking machines. After a general introduction we briefly review the history of humanoid robots. We present the most prominent examples of current biped robots. Following a brief introduction of basic aspects of the dynamics of biped robots we review popular walking control methods. A brief review of open research questions concludes the paper.

The use of teams of cooperating mobile robots is necessary, if a task is inherently distributed or cannot be solved by a single robot. In cases when a task can in principle be completed by a single robot, the use of multiple robots can result in advantages in execution speed, robustness, and scalability. Multiple resource allocation problems must be solved to make robots cooperate, like task allocation or the allocation of portions of the work space. This article discusses architectures for communication and control in multi-robot systems as well as path planning and multi-robot learning. The possible utility of cooperating mobile robots is illustrated using four application examples.

Significant progress in aerospace and computer sciences has enabled a rapid development of autonomous airborne systems. Light weight structures, efficient aerodynamics, new propulsion and energy storage systems as well as advanced aircraft systems are accelerating this trend. In combination with matured capabilities of autonomous situation assessment and flight guidance and control techniques new domains for applying unmanned aerial vehicles are being explored. This article provides a high level view on trends and technologies.

We discuss that “cognitive robotics” implies the extension of classical robotics methods — esp. planning, control and learning — to external degrees of freedom (DoFs). External refers to the articulated and manipulable DoFs of the environment and the objects therein. Coping with these DoFs requires to go beyond vector spaces and Gaussian distributions and instead address the complex structure of natural environments, which is hard to formalize and translate to appropriate representations and priors for efficient learning. With this discussion we aim to highlight the theoretical challenges behind the goal of robots that autonomously explore their environment and learn to manipulated external DoFs. The integration of logic, geometry and probabilities is one of these challenges. In this context we briefly sketch own work on relational reinforcement learning, exploration and symbol learning.

This paper deals with vision-based navigation of autonomous mobile robots. An approach to improve visual odometry algorithms is presented that picks up the idea of actively influencing the image data by actuating elements of an active camera system without using an active camera system effectively. It is shown that via virtual fixations the spatial optical flow pattern can be influenced in such a way that the camera motion computed based on this optical flow is getting more precise. For computation of the camera motion, the active 7-point algorithm is introduced. Based on statistical evaluations it is derived which fixation points are especially suited and a strategy how to choose such suited fixation points is described. A comparison between the presented virtual-active approach and a state of the art visual odometry implementation is given.

Modern 3D simulation technology exploits and promotes synergetic effects of complex development projects in mobile robotics. It provides new methods for the multi disciplinary development of mobile systems. To this end, algorithms for simulation, control and data processing must be integrated with system models and real systems into a holistic framework. This active knowledge base simplifies the validation of new methods under various conditions and their application to other domains in the course of an efficient technology transfer. New concepts like the Virtual Testbed approach provide the conceptual foundation.

While robots, mobile or not, do not yet play an important role in everyday life, countless tasks that demand high precision and speed have been conquered by them. But their application domains get larger and larger, be it in form of autonomous robots able to discover deep sea regions, driver-assistance-systems in cars or unmanned drones in the sky. With every new application new requirements must be met, not only in terms of hardware like sensors or actors, but also with respect to perception- and navigation algorithms. Especially when harsh environments aggravate perception, robust navigation methods are required, that can cope with uncertain sensor measurements and select those robot motions out of many, which yield a safe and goal oriented behaviour. This paper attempts to showcase different navigation paradigms and their realizations on real mobile robots, as well as the reasons behind choosing a certain realization. The paper closes with an outlook towards possible future developments in the area of mobile robot navigation.