In this paper, we consider some approaches to the task-level robot learning from demonstration for the target defense by a team of unmanned water surface vehicles. We assume that the problem of target defense by a team of unmanned water surface vehicles represented as a problem of learning of rhythmic motor primitives. We consider for the problem neural networks as oscillators to learn rhythmic motor tasks. Also, we use the approximate period problem and introduce the approximate period problem

more »

... for a set of strings with a set of restrictions. In this paper, we present experimental results for different synthetic test data set for stationary and moving targets. Keywords: unmanned water surface vehicle, task-level robot learning from demonstration, rhythmic motor primitive, approximate period problem Unmanned water surface vehicles (USV) have been extensively studied over the last thirty years [1] . It is clear that USVs can be used to solve many practical tasks. In particular, we can mention sea-bed mapping and ocean sampling tasks [2], environmental monitoring [3], cooperative surveillance [4], search and rescue [5] , and patrolling and protecting different areas [6] . It should be noted that USVs can significantly increase the capability of other surface or