A fundamental element of a centralized autonomous vehicle control process is the representation of the transport environment. The road surface has to be partitioned to be able to ensure the main safety requirements of the transport process: a vehicle can only be in one place at a time, and only one vehicle can be in one place at a time. The occupancy grid concept is widely used for this purpose; however, previous concepts mainly utilize square grid representations, considered as a static component of the traffic models. Our research focuses on discovering and introducing the impact of different grid representations by evaluating their efficiency, computational complexity, and safety characteristics. This result provides the basis for real-time optimization of self-organizing road transport networks, as our research gives clear information on the impact of different grid structures. It has been pointed out that applying other network structures than the traditional square grid structure can achieve better results. Accordingly, some of the newly introduced structures (such as the hexagon-based, triangle-based or combined structures) are proved to have significant advantages compared to the classical grid representation.