Development of a self-learning system for optimizing the driving rules of autonomous transport vehicles and their temporally and spatially coordinated activities

Managing internal logistics operations is becoming more and more demanding in increasingly dynamic modern production environments. The state-of-the-art solution in this area is the introduction of autonomous mobile robots (AMRs) which are a more advanced version of automatic guided vehicles (AGVs) and have advanced sensors that give them the ability to localize in space and to navigate around without a magnetic strip placed on the floor. The main goal of the project is the development of algorithms for an efficient and flexible multi-robot transport. Important novelties and advantages of the proposal according to the existing approaches in the industry will be three-fold. First, a better flexibility will be obtained by automatic building or adjusting the map configuration, which will enable a more efficient transport with AMRs (e.g. shorter transportation times, less congestion and fewer conflicts that would require solution adaptations when planning AMR routes). Second, an important novelty will be a self-learning module for tasks assignment to AMRs, which will adapt the rules to the actual situation (i.e. a current map, current statistics of the transportation orders, properties of the implemented route planning algorithm, etc.). It would therefore enable an improved performance over time through a more efficient planning and lower complexity (according to the combinatorial complexity of the simultaneous solving of task assignments and route planning). And third, there will also be advantages in a coordinated route planning for a group of AMRs by setting occupancy windows for resources in the map sections, considering transportation priorities with a minimum required coordination and without conflicts and collisions. The obtained transportation plans will facilitate a local control with fewer necessary corrections during transport. These algorithms will be tested, analyzed and demonstrated at several levels of fleet management. It will be shown that it is possible to achieve more efficient and robust solutions than the existing ones with the above-mentioned new approaches to the abstraction of the intralogistic problem, multi-robotic route planning and task assignment. The project is divided into six work packages which include activities related to the project management, appropriate configuration of the logistics system determination, coordinated multi-robot planning, self-learning task assignment, integration and establishment of a demonstration system, as well as the dissemination of the results. There is a high probability of the project goals being fully realized as they have been previously verified by the studies mentioned in the project description and because the participating company has all the necessary infrastructure in order to implement the application. The solutions will be validated on simulations and also by implementing the sample application on newly developed robots from Epilog d.o.o. Furthermore, the latter will serve for the purpose of performance demonstration and dissemination of the results. The results of the research will be efficiently and extensively used by Epilog d.o.o. and their partners in upgrading existing solutions.