Method of Determination of the Shortest Path of Mobile Robotic Platform in the Conditions of Limited Resources

Abstract

With the transition to Industry 5.0, mobile robotic platforms (MRPs) have become an important element of industrial automation, replacing outdated conveyors. They optimize the transportation of materials at industrial enterprises, integrate with control systems and adapt to changing conditions. The authors focus on their navigation in dynamic environments, avoiding obstacles when interacting with people in conditions of limited resources. The paper analyzes the limitations of existing approaches to route planning for MRPs, in particular the A*, D*, DLite, M algorithms in dynamic environments. Special attention is paid to the optimization of algorithms using federated learning, introducing artificial intelligence to increase productivity.

The authors proposed a method for determining the shortest route for AGV, which is based on classical algorithms for finding the optimal route in conditions of static and dynamic obstacles, taking into account limited resources, in particular the remaining battery charge and the time to complete the task. Stationary obstacles include walls and fixed workplaces of personnel, and dynamic obstacles are classified as living (industrial personnel) and inanimate (other MRPs or unpredictable objects such as boxes). Federated learning is used to predict the voltage drop of MRP batteries taking into account the individual characteristics of the platforms. The developed method involves dividing the route map into a uniform grid, constructing obstacle matrices, predicting battery voltage and calculating the optimal route using the A*, D*, DLite, M algorithms. The effectiveness of the method is evaluated by the following parameters: route length, number of cells passed, execution time, remaining battery charge. The results of the method are presented on the example of the AGV Formica 1 route, AIUT, Gliwice, Poland.

The D* algorithm underlying the developed method is the most efficient in terms of execution time, number of cells passed, and battery conservation, which makes it optimal for dynamic conditions. The DLite and M algorithms also show good performance in static conditions, with lower resource consumption. A*, although it finds the optimal path, is the slowest and less efficient under limited resources. Thus, for dynamic environments, D* is the best choice, while DLite and M are good options for stable conditions.