Markov Chain Monte Carlo-Based Parameter Estimation for Entry Descend and Landing Applications

In light of recent events, the race in space exploration is getting heated every day. Space agencies around the world are trying to reach out to other planets to collect data samples or even to establish a permanent human residence, if possible. These missions often require a safe landing of a payload on the surface of a body with an atmosphere. Given the aerodynamic and thermodynamic requirements, the shape of payloads in these missions is usually blunt bodies which tend to be dynamically unstable geometries. During the descent, the unstable nature of the body leads to an increase in the angle of attack oscillations of the body around mid to low-supersonic speeds. Unless the oscillations are addressed during the design stage, they can lead to catastrophic results which are directly related to the success of the mission. To address this issue, the dynamic stability characteristics of the vehicle must be identified. However, identifying the dynamic stability coefficients along with the uncertainties in the predictions is a challenging task due to the complex physics environment of the problem. In this study, we propose a deep neural network (DNN) based Markov Chain Monte Carlo (MCMC) framework to predict the dynamic stability coefficients and identify the uncertainties in the predictions. Since the Bayesian-based approaches require a prior distribution, the DNN is utilized to generate the priories required for the MCMC method. Then, the MCMC algorithm predicts the dynamic stability coefficients with the uncertainties by using the priories obtained from the DNN.