This paper proposes a feedback-based conformal prediction framework for trajectory optimization that dynamically adjusts prediction regions using realized trajectories, improving decision performance while preserving coverage guarantees.