Michael Gerke

GA-based path planning for mobile robot systems employing an active search algorithm

Abstract A genetic algorithm (GA)-based path planning software for mobile robot systems focusing on energy consumption is presented. One special feature of this software is the consideration of changing textures—and therefore changing energy requirements when moving—within given environments. Problem specific genetic operators are introduced, and especially the handling of exceptional situations is described in detail. After that, an active search algorithm is introduced which allows to overcome the drawbacks of an earlier version of that software. This algorithm is executed within two different phases of the optimisation process. For each obstacle within the environment regarded a path circumventing it is computed in a preparation phase. Then, in the execution phase of the GA itself, the results of the preparation phase are used to find optimum paths. The mode of action within both phases is described in detail. Furthermore, the suitability of the approach is substantiated by an example.

Genetic path planning for mobile robots

Optimal path planning is an important feature of autonomous mobile robot guidance and control during outdoor missions. Offline path planning of any manoeuvres in cluttered environments depends on pre-mission knowledge about topographical features of the landscape between the robots start position and goal position. A topographical landscape consists of regions with different traversal costs, and it is not necessarily binary in nature. For certain extraterrestrial or cross-country missions, where a lack of vehicular resources has to be handled, there is a strong need to calculate an optimal path according to a given optimization criterion. The paper presents a genetic approach to the problem. Evolutionary computation is used to carry out path optimization.

Planning of optimal paths for autonomous agents moving in inhomogeneous environments

Long-range route planning is an important feature of autonomous agents guidance and control during out-door missions. Offline path planning for any manoeuvres in cluttered environments depends on the pre-mission knowledge about the topographical features of the landscape between the agents start position and goal position. A topographical landscape consists of regions with different traversal costs, and it is not binary in nature (in terms of passable/impassable). For certain extraterrestrial or cross-country missions, where a lack of vehicular resources has to be handled, there is a strong need to calculate an optimal path according to a given optimization criterion. This paper presents an optical approach to find optimal paths in inhomogeneous environments, which follows from FERMATs principle. It is derived from the beam propagation method, which is commonly used in optical research areas.

Realtime collision avoidance for industrial robots

Abstract A Realtime solution to the collision avoidance problem for industrial robots (e.g. PUMA 562) in a multirobot workcell is developed. Reducing the problem to more convenient kinematics and evaluating their actual and the most critical configurations leads to a decision between the preplanned and a modified trajectory which is based on security criteria. Components of the definitive trajectory are computed within robot controller cycle time and transformed back towards the 6 DOF-robot under consideration.

A virtual laboratory for an inverted pendulum and crane control

Abstract A virtual laboratory allows students to gather experiences with practical on-line experiments at remote sites. Moreover, it becomes possible to share expensive equipment among several universities or education centres. Several German universities were engaged last three years in the ‘Leam Net’ project, in order to develop a network of remotely accessible virtual laboratories and server with various courses. One of our contributions to this project is a laboratory experiment for inverted pendulum and gantry crane control. Several new components for real-time control and Internet access have been developed and implemented.

Fuzzy Backpropagation Training of Neural Networks

Adaption of parameters during neural network training according to the actual shape of the error surface is supposed to be a powerful instrument to enforce convergence and to decrease time consumption of neural network training.

Visual inspection of power lines by U.A.S.

Unmanned aerial systems (UAS) can be used nowadays for automatic visual inspection of power lines. Two challenging problems need to be solved: The detection of the power lines and the tracking of the lines. This paper presents a method for vision systems to extract power lines from cluttered natural backgrounds e.g. trees, cloudy skies. By means of range filters, some morphological operations, Canny operator and Hough transformation this separation is carried out, which results in binary images containing the detected lines only. These images can be further processed for inspection of power lines and pylons.

Lighter-Than-Air UAVs for Surveillance and Environmental Monitoring

This paper reports on the development of a computer controlled airship for applications in surveillance and environmental monitoring. This covers the areas of crisis management and security situational awareness. Small airships are combining many features of planes and rotorcrafts. They allow measurements very close to the scene, they are able to carry even heavier payloads than comparable flight systems, and they can move very slowly as well as close above ground. As no dynamic force for lifting is necessary, on-board energy resources can be completely applied to actuation, computer systems, and sensor equipment. Also, an airship possesses several inherent security features such as usually small velocity, low weight, and a soft hull.

Airship aerodynamics - Simple control of the elementary flight parameters

This article describes a simple control of elementary flight variables for an unmanned airship such like airship flight altitude and airship azimuth or flight direction. For this purpose were all controllers designed on the basis of nonlinear mathematical model of airship flight mechanics which was obtained by PEM (Projection Equivalent Method) [13]. All controllers were designed for “silent” airship flight mode. “Silent” flight mode means, that during flight on-board airship controllers compensate oscillations about all three airship axes. By means of these controllers, it is possible to achieve required controller precision and thereby to guarantee “silent” airship flight mode which is necessary for photo or video recording or other airborne measurement tasks. This paper not only shows simulation results, but also provides measured data from real experimental airship flights which were realized by applying these controllers.

Modelling of Airship Flight Mechanics by the Projection Equivalent Method

This article describes the projection equivalent method (PEM) as a specific and relatively simple approach for the modelling of aircraft dynamics. By the PEM it is possible to obtain a mathematic a...