The main problems of modern robotics is the development of cognitive robots who can independently acquire higher cognitive categories and on their basis to communicate with people and other robots. Sophisticated senzoric and motoric skills are a prerequisite for the development of such robotic systems. Therefore, humanoid robots are particularly suitable as an experimental base for studying cognitive processes in artificial systems. The assumption is that humanoid robots may soon acquire higher cognitive abilities, if they are actively involved with a human aor teacher. This talk will present the latest research, which have been linked to various ways and algorithms that offer humanoid robots to carry out complex tasks in field of manipulation and grasping objects in working with people in natural, human intended environments. The lecture will be primarily dedicated to senzomotoric operations, their generalizations and adapting to new situations and contexts.