With the increasing popularity of human–robot interaction (HRI), control systems are becoming more and more interesting and complicated because they need to coordinate continuous human movements and discrete robot decision-making in dynamic environments. The current strategies tend to be less flexible when working collaboratively, take longer to react and become less stable in interaction. This study presents a hybrid nonlinear state-space model, hybrid automata, Model Predictive Control (MPC), Extended Kalman Filter (EKF) and Adaptive Hybrid Interaction Algorithm (AHIA) based Hybrid Dynamical Human–Robot Interaction (HD-HRI) framework for real-time human intent estimation and robot motion planning. The framework is assessed based on the accuracy of interaction, prediction of intent, trajectory tracking error, response latency, collision avoidance rate and task completion success. Experimental results confirm that 95.8% of the interactions are accurate, 94.9% of intents are correctly predicted, 21.6% of trajectory tracking errors is reduced, 18.4% of response latency is reduced, 97.3% of collisions are avoided, and 95.6% of the tasks are completed. The results prove that the proposed hybrid dynamical framework can be effectively implemented in practical applications, which can be reliable, stable, and efficient for human–robot collaboration.