When spacecraft are in the outer solar system, communication delays with Earth can exceed hours, and fixing trajectory errors could waste critical fuel or even doom a mission, making autonomous real-time decision-making essential. I'm developing a deep reinforcement learning system that can autonomously plan fuel-efficient trajectories for spacecraft navigating multi-planetary missions. The algorithm will need to optimize complex maneuvers like gravity-assist flybys while adapting in real-time to the uncertainties that actual missions face. Using Proximal Policy Optimization (PPO) algorithms integrated with an astrodynamics simulation framework, I'll train the spacecraft to make intelligent trajectory decisions that balance fuel efficiency with mission constraints. I'll end up with a complete RL-based trajectory optimization system benchmarked against real mission scenarios and validated through Monte Carlo simulations. Especially where fuel costs escalate dramatically for deep space missions, even modest improvements could enable missions that are currently financially infeasible.