Résumé : In the advanced field of bio-inspired robotics, the emergence of cyborgs represents the successful integration of engineering and biological systems. Building on previous research that showed how electrical stimuli could initiate and speed up a jellyfish's movement, this study presents a groundbreaking approach that explores how the natural embodied intelligence of the animal can be harnessed to address pivotal challenges such as spontaneous exploration, navigation in various environments, control of whole-body motion, and real-time predictions of behavior. We have developed a comprehensive data acquisition system and a unique setup for stimulating jellyfish, allowing for a detailed study of their movements. Through careful analysis of both spontaneous behaviors and behaviors induced by targeted stimulation, we have identified subtle differences between natural and induced motion patterns. By using a machine learning method called physical reservoir computing, we have successfully shown that future behaviors can be accurately predicted by directly measuring the jellyfish's body shape when the stimuli align with the animal's natural dynamics. Our findings also reveal significant advancements in motion control and real-time prediction capabilities of jellyfish cyborgs. In summary, this research provides a comprehensive roadmap for optimizing the capabilities of jellyfish cyborgs, with potential implications in marine reconnaissance and sustainable ecological interventions.