Lionel Gayraud improved the robustness of target tracking in the dstl/Stone-Soup repository by refining the handling of survival probabilities within the PointProcessUpdater. He ensured that survival probabilities were applied exclusively to existing targets, not to newly born ones, and integrated this logic into the missed-detection process. This adjustment reduced bias in target weighting and enhanced state estimation accuracy, particularly in scenarios involving missed detections or spontaneous target births. Using Python and leveraging expertise in algorithm refinement and Kalman filters, Lionel also introduced regression tests to align the implementation with established pseudocode, increasing reliability and coverage for complex state estimation workflows.