This paper considers how to create an environment model suitable for use in research on the problem of spatial sensor distribution. A software solution was developed that recognizes the boundaries of interiors, furniture, and other obstacles in a three-dimensional point cloud and records their floor plan as a set of polygons. The paper briefly explains the theoretical concepts that were the starting point for the research – the art gallery problem, stereoscopic cameras, point cloud registration, RANSAC segmentation, and others. The process of recording the space using the TurtleBot3 mobile robot and depth cameras in a simulation environment is described, followed by the details of the creation of an environment model. The effectiveness of the algorithm for creating an abstract environment model from a given point cloud was verified by comparing the obtained results with the ground truth, where a similarity of 98 – 99% was calculated for simple environments, and 92 – 96% for complex environments. The obtained example of a more complex environment was used for the analysis of the sensor placement problem, which tries to find the optimal number and placement of sensors needed to achieve complete coverage of the closed environment. This example is significant because it can provide an insight into the effectiveness of the sensor placement model in real environments that we encounter in everyday life. An experiment was conducted using six metaheuristic optimization algorithms and up to 50 sensors. Based on the ANOVA statistical analysis of the data, conclusions were drawn about the influence of the selection of the optimization algorithm and the number of sensors on the achieved relative coverage of space and execution time. It was observed that the artificial bee colony algorithm achieves the best results, and the particle swarm optimization algorithm also gives good results.