Search. Plan. Learn.

The ability to search for and find solutions to non-trivial problems, to plan several steps ahead, and to learn (from others’ or one’s own experience) are properties traditionally attributed to intelligent systems (or, as is commonly said today, artificial intelligence). It is not surprising that these (sub)fields of AI have been actively developing since the very birth of this discipline and continue to develop to this day.

In this talk, we will consider a number of interesting and practically motivated problems that are being addressed by researchers using methods of heuristic search, automated planning, and machine learning, and which fall within the scope of my interests. For example: how can we plan a set of non-conflicting paths for n mobile agents operating in a shared environment? Or: how can we solve a complex logical puzzle such as a Rubik’s Cube? Or: how can we construct a map of the environment in which a robot moves based on its sensor data? Special attention will be given to open problems that currently have no solution, but which (in my opinion) are quite feasible to tackle within the scope of a coursework or thesis project and to contribute to science.