This paper describes the continuation of a series of research in which human behavior, in the form of models and algorithms, endeavor to be mapped into the field of mobile robotics, all with the aim of more efficient path prediction with collision detection in a dynamic obstacle environment. In our latest research [1] we proposed a new approach called Sliding Holt (SH) algorithm which is used for calculation of future spatiotemporal state of dynamic obstacle. Besides SH algorithm we also proposed method for (non)collaborative collision detection with dynamic obstacles in 2D and 3D space which ensures full de-centralization of the collision detection process. Research described in this paper focuses on the application of Computer Vision for detecting the position and movement direction of mobile robots in order to create assumptions for collision detection and path planning with within the production environment. We propose a new approach for detecting mobile robot direction based on triangular shape and use of Cost of Action in order to allocate tasks within robotic swarm.