The use of Solar Electric Propulsion (SEP) for missions within the Earth-Luna system (cislunar) has been planned, due to its low propellent usage compared to chemical propulsion. However, SEP introduces much higher transfer times compared to the chemical propulsion systems, making optimal trajectory design and planning more challenging. This research explores the use of a real time (online) Neuroadaptive state observer, i.e., Modified State Observer (MSO) to improve the dynamic modeling that goes into the trajectory optimization. The MSO utilizes a single layer neural network to capture the unmodelled perturbations in the spacecraft dynamics, which could be caused by Earth's polar flattening (J2), Sun's radiation, and other forces. These can be incorporated into a multi-stage, receding horizons optimal controller to improve the trajectory planning and optimization. This is especially important in situations where the true dynamics are not modelled - due to lack of enough information or unknown dynamics, or the full or high-fidelity model cannot be used due to limitations in on-board computing power. The real-time adaptations could eliminate the need for offline training. The improved on-board tracking provided by the MSO could also eliminate a ground-in-loop control optimization, which would have communication delays in long distance travel. The MSO application was studied to ensure viability in a simple two-body problem scenario of a spacecraft orbiting the Earth in a circular orbit of altitude 100 km. A high-fidelity two-body model that incorporates the Earth's J2 perturbations was compared to a low-fidelity model without the J2 perturbations. The MSO was then incorporated into the low-fidelity model to adapt to the unmodelled perturbations. The resulting tracking errors in the MSO states were compared to the high-fidelity states. Position tracking errors of approximately 15 km and velocity tracking errors of approximately 0.015km/s were observed in the low-fidelity model, and observed to be increasing with time. The MSO improved these to 0.15 km and 0.004 km/s, respectively, showing that the MSO can model the unknown dynamics reliably. The effect of certain MSO tuning parameters - the observer gain, adaptation rate and neural network weight damping were also studied. Similar studies are currently being conducted on the three-body problem of a spacecraft moving in the Earth-Luna system. This uses a realistic high-fidelity model from the NASA General Mission Analysis Tool (GMAT) and a low-fidelity circular-restricted case, for various trajectories in the cislunar space.