Abstract:

The creation of a stochastic model for collaborative robot interaction in a production setting with POMDP seeks to examine the influence of uncertainty on decision-making during robot collaboration. The model considers the restricted observations, the stochastic characteristics of state transitions, and the necessity to optimize actions based on the reward function. The simulations enabled the evaluation of the relationship between total and instantaneous rewards, model parameters, and interference levels, mirroring actual production settings. The results collected validate that the implementation of POMDP offers flexibility and adaptability to the system in response to varying scenarios and conditions. The study of the data revealed the impact of weight coefficients in the reward function on the equilibrium between productivity and safety, enabling the formulation of various interaction techniques. Simulations revealed the stability of average outcomes and the model's suitability for optimizing robotic behavior in a complex environment. The proposed method can be integrated with neural network techniques and planning algorithms that utilize the states of the production environment to enhance decision-making accuracy and efficiency. The proposed model serves as a foundation for subsequent research focused on multi-level scenarios and human-robot interaction within the context of the Industry 5.0 idea.