Institute for
Robotics and Process Control

Deutsch   English

MONAMOVE - a flexible transport system for manufacturing environments

1. Overview and short description

MONAMOVE: MOnitoring and NAvigation for MObile VEhicles

monamove-concept.gif

The flexible and automated flow of materials, e.g. between different work cells and a computer controlled warehouse, becomes more and more important in modern factory environments. To obtain such a flexibility it is obvious to use autonomous guided vehicles (AGV). Many of the autonomous guided vehicle concepts, which are known from literature, use highly specialized on-board sensor systems to navigate in the environment. In contrast to these concepts, the flexible transport system MONAMOVE proposed by us uses only simple, low-cost on-carrier sensors in combination with a global monitoring system and a global navigation system. This combination of global monitoring and global navigation enables the carriers to navigate without any fixed predefined paths.

2. Navigator

The navigator is responsible for the preplanning of the desired route from the start point to the goal point. Most of the known path planning algorithms published in the literature work with a pure geometric world model: First the obstacles are represented in the work or configuration space. Subsequently, a path is computed in the respective space, using some decomposition technique, some potential field method or some roadmap approach. Moving obstacles, which are not controlled by the planning algorithm, can be integrated by introducing an additional time dimension, by dividing the planning process into a path planning and a velocity planning problem or by an other approach. However, for such an integration of moving obstacles up to now the trajectories of the moving obstacles are assumed to be known exactly. We distinguish between static obstacles which stay always at the same position, partly dynamic obstacles at constant positions for certain time periods and completely dynamic obstacles which are moving around in unknown ways. This classification can be computed automatically using the global monitoring system described below. Preplanning can be realized with a pure geometrical world model and one of the known methods, taking only static and partly dynamic obstacles into account. The incorporation of the actual position of completely dynamic obstacles makes no sense during preplanning, because theses obstacles very likely will be at an other position when the execution of the path is started. After preplanning, the runtime planning module takes into account all obstacles in the local environment -- especially the completely dynamic obstacles -- to modify the preplanned path -- if necessary. This approach has several disadvantages: The preplanned paths don't depend on the movement properties of completely dynamic obstacles. Consequently, in general many potential collisions during the execution of the path have to be resolved. The pilot plans at runtime path modifications which don't coincide with the globally optimized paths of the preplanned solution. In contrast to this approach, we integrate additional rules into the preplanning step to improve the executability of preplanned paths in a dynamic world without knowing the exact movements of the individual completely dynamic obstacles. As the individual movements are unknown during planning time, we cannot use the approaches mentioned above. In our geometrical path planning approach, we use explicit rules, like right orientation (traffic rule), to reach this goal. Under the assumption, that all completely dynamic obstacles follow this right orientation rule, the preplanned paths are well integrated into the dynamic behavior of all obstacles. As theses rules have to be defined manually, we describe a second approach working with implicit rules. These rules can be computed automatically by means of sensor input; they correspond to statistical information representing the average behavior of dynamic obstacles. The statistical information comprises the occupancy probability with partly or completely dynamic obstacles and the statistical moving direction.

3. Global monitoring system

The monitoring system has to observe all controlled vehicles and the complete changes of the dynamic factory environment; this offers the possibility of early path modification to the navigator, even if the obstacles can't be seen from the actual positions of the carriers. Simultaneously the monitoring systems reports the actual carrier positions and orientations to the local navigators, which compensate the differences between the driven and the computed paths in a closed control loop.To obtain the desired data, the monitoring system takes into account as much as possible a priori information. This information consists of the factory layout, information about the carriers and the planned carrier routes inside the layout. The monitoring system allows the integration of different sensor types to obtain rich environment information. The current implementation exclusively uses cameras equipped with wide angle objectives, mounted at the ceiling of the factory-floor, but also motion detectors or light barriers could be used in addition. The visual areas of the cameras, called view tiles, are overlapped and cover the whole area drivable by the carriers. Consequently, a seamless observation is guaranteed. All cameras are connected via a computer controlled camera multiplexer to an image processing system with region of interest capability. The enormous amount of picture data, obtained from the cameras is reduced by the new view tile concept. The basic idea of this concept is the division of the single tiles into subregions according to their geometrical and logical structure. This partitioning includes, e.g., regions of static and dynamic obstacles, free space, carrier areas, view tile borders, docking stations and socalled entering areas, where priorily unknown objects can enter into the observed area. Using this partitioning the analysis of the picture data can be reduced drastically with simple rules. In contrast of dynamic obstacle regions, which have to be analyzed continuously, the regions of static obstacles have to be analyzed only temporarily. Generally, the observation of a view tile can be reduced to the regions of dynamic obstacles and the entering areas. The analysis of a view tile without any entering area depends on the presence of a moving obstacle. For any dynamic obstacle a history of position and velocity is build up, which allows the predetermination of the area of picture acquisition for the next moment. Thus, the regions which have to be analyzed to get the real obstacle positions are reduced.

It took 0.31s to generate this page.