Wearable robots particularly lightweight, comfortable soft exosuits are increasingly transitioning from laboratory prototypes to everyday assistive devices. Recently, exosuits employing wire tendon actuation without force sensors have been introduced to reduce complexity and cost. In this study, we propose a controller capable of accurately delivering assistive profiles of varying magnitudes, shapes, and timings through model predictive position control. Our method used models of thigh motion and human-suit interaction stiffness. The latter estimated as an actuator property that varies with individual body stiffness, walking speed, and assistance intensity. This estimation enables precise force tracking and cable slack management, despite the absence of force sensors. To validate our approach, we conducted treadmill experiments with three subjects walking at 1.25 m/s, applying assistive force profiles with different magnitudes, shapes, and timings. Results demonstrated reliable force delivery across all conditions, with a magnitude error under 8.06 N, RMS error under 10.55 N, and timing error under 1.23 %. These findings suggest that our sensorless force control strategy can be widely adopted across various tendon-driven robots and other human-robot interaction domains. The main limitation is that this model-based control dependes on actuator robustness. Future work will focus on predicting actuator properties before operation to ensure robust model prediction performance