The Multi-Agent Pathfinding (MAPF) problem consists of finding a set of conflict-free paths on a graph. On the one hand, it is known that even under many simplifying assumptions, obtaining optimal solutions to this problem belongs to the NP-Hard class. On the other hand, efficient algorithms for solving MAPF are highly valuable in practice, for example, in the development of automated warehouses. This is precisely why the field of MAPF is actively developing, and the corresponding approaches are constantly being improved, both through the algorithmic enhancements and through the application of modern machine learning methods and models. Despite the progress achieved, many interesting and unsolved problems remain in the field, and researchers at the Markov Lab are working on some of them.