**6. Conclusions**

In this chapter, a machine learning design has been presented and implemented to discover space object behaviors. Our GTEL methods using CNN models the circumstance with PE game rather than a control problem. The stochastic modeling/ propagation, RSO tracking, and data level fusion are utilized to predict the relations for future space by game reasoning. In order to generate the data for training, the Marko game approach is used with maneuvering strategies. The method provides a way to solve the SSA using unknown behaviors. Additionally, the unknown behaviors exist where a satellite employs the tracking and sensing way to corrupt the tracking estimates to perturb the sensors. On the other hand, the space sensors decrease the uncertainties during tracking process. Finally, the CNN-DNN is used to train the numerical results, where the accuracy is 98% for our classification RSO model with 143 labels.

In the future, a multi-player game theory with adversarial network with SSA will be employed to enhance the deep learning for sensor management, combined tracking, as well as secure communications. Methods for diffusion-based cooperative space object tracking [16] and block chain [17] are emerging as methods for game-theoretical methods.
