Purpose. To develop a simulation model of the motion of a group of objects in three-dimensional space to generate datasets and individual samples for further use in training neural networks and developing a mutual positioning system. Methodology. Analytical review, the automated control theory, imitation modelling. Results. Using Python, a 3D model has been developed. Digital and graphical representations of the linear trajectory of movement for a group of 4 drones based on input data, including coordinates of initial and final points and an error range, have been received. Trajectories with compensated and uncompensated deviations have been compared. Approximation and simplifications have been justified by the goal of enabling the use of the model by those without fundamental knowledge of control theory, system modelling or advanced mathematics. Code testing and benchmarking have also been conducted, proving the financial feasibility of the solution. The model will be used by the authors for further research, including assessing the likelihood of obtaining biassed data from sensors of some drones in a swarm and finding ways to correct it. The source code of the developed model is provided for open access. Scientific novelty. An imitation model of interaction and movement in three-dimensional space for a group of autonomous objects has been proposed. These objects, capable of collective decision-making based on swarm intelligence, aim to compensate for errors or interference. Practical significance. Testing in the AWS cloud environment has demonstrated that even with minimal computational power the proposed imitation model yields adequate results, suitable for neural network training.
Read full abstract