The topic of this article is an off-line movement system for the mobile robots. Its main goal is to ensure their movement from one point to another in a partially determined environment with static and mobile obstacles. Several systems are described. They are calculation of the impact of a potential force field, path assessment based on a neural network, avoiding collision system and a multi-agent system architecture for detection of obstacles and mobile robot control. The article presents mobile robot agents, which process data concerning the environment, communicate with each other and allow a robot to avoid the static and dynamic obstacles. Besides that, the most important task for a robot is movement in a partially determined environment, in which the obstacles' coordinates are unknown for the system. From a technical point of view, navigation in a dynamic context is the next set of actions: a robot is given a certain trajectory from point A to point В and then it will be able to deviate from a certain route. The task is difficult because a mobile robot has its own design and kinematic restrictions. In crowded places there are many people. Consequently, there numerous moving obstacles to be avoided, i.e. the robot movement involves maneuvering and even escort of persons to the point of destination (department of a shopping center or museum exhibit). From a technical point of view, in order to avoid collision of a mobile platform with the objects their force fields can be used. The experimental results of the developed system are presented below.