This study presents an autonomous robotic pruning system for grapevines that integrates vine‑structure estimation, motion planning, and a vision‑based end‑effector position‑correction mechanism. Grapevine pruning requires robots to operate in environments characterized by dense branching, frequent occlusions, estimation errors, and continuous structural changes caused by branch removal. To address these challenges, the proposed system combines end‑effector pose determination based on estimated vine geometry, trajectory generation constrained by an obstacle model, dynamic relaxation of constraints through obstacle‑model updates after cutting, and a hybrid path‑planning strategy that switches between linear and sampling‑based trajectories. Experiments conducted in a commercial vineyard demonstrated that the system achieved a pruning success rate of up to 82% and completed pruning of a single vine in 92 seconds. The end‑effector position‑correction mechanism, which fuses global structural information from a vehicle‑mounted camera with high‑resolution local observations, improved the pruning success rate from 68% to 82% and achieved accurate bud‑count control in more than 85% of successful cuts. The system also showed robustness against misidentification caused by branches remaining after cutting. Remaining challenges include handling failed cuts, improving depth‑sensing accuracy, and optimizing task sequencing. Overall, the results indicate that combining adaptive motion planning with local visual correction provides a practical foundation for autonomous grapevine pruning and represents a significant step toward the automation of vineyard operations.