The ankle exoskeleton has the ability to reduce the metabolic cost of human walking. However, due to individual differences and dynamically changing environments, there is a lack of adaptive control strategies that can meet the specific needs of users and adapt to terrain variations. To address this challenge, we propose a controller that integrates walking dynamics, using wearable sensors to estimate ankle joint power and generate biomimetic assistance curves in real-time. This controller can adapt to the user’s specific movement needs, providing positive power assistance. We validated the controller on various terrains, including level ground and slopes, as well as at multiple speeds. Additionally, we measured the muscle activity of the soleus and gastrocnemius muscles at two different walking speeds. Experimental results show that the controller can adapt to changes in terrain and speed without the need for terrain classification. Compared to the unpowered condition, the activation levels of the soleus and gastrocnemius muscles decreased by 17.14% and 14.60%.