The objective of the proposed research is to investigate machine learning control and its impact on the discharge plasma of a Hall thruster. Hall thrusters are electric propulsion devices that are used to propel spacecraft that are used in deep space missions and satellite applications. As Hall thrusters are developed for higher power levels, oscillations become a growing concern, affecting the stability of the device. These oscillations can be high in amplitude and can damage the power processing unit, can erode the channel of the thruster, and produces EMI. Thus, developing control approaches to mitigate these oscillations has gained increasing attention. Previous approaches have investigated PID control and iterative learning control, but no work has investigated real-time physics-informed machine learning control of a Hall thruster plasma. Real-time control allows for more accurate shaping of the discharge voltage and current oscillations temporally, and physics-informed control also improves accuracy by using a physical model rather than data alone to determine better voltage setpoints. We start by characterizing the Hall thruster using signal processing techniques, then train neural networks on Hall thruster voltage perturbation data, allowing us to construct a better model of the load. We then use model predictive control to either increase thrust by phase synchronizing voltage and current, reduce oscillations, or modulate oscillations using FM/AM/PM to be used for communications purposes. Extended Kalman filters are used to estimate various plasma states of the discharge, allowing one to observe how control affects the plasma in real-time. The control work will involve simulations and experiment to prove the effectiveness of the approach.