Abstract: To enhance the intelligence of wheeled assistive walkers in supporting individuals with limited mobility, we investigate user intention recognition and personalized gait adaptation based on a wheeled mobile robot (WMR) platform. Firstly, we design a variable admittance controller that integrates both horizontal and vertical force components to detect the user’s walking intention in real time. Secondly, we propose a learning method based on cooperative dynamic movement primitives to capture the user’s motion characteristics within a complete gait cycle, enabling rapid adaptation to individualized gait patterns. In addition, we integrate an emergency response strategy based on abnormal vertical force detection to improve user safety. Finally, we conduct experiments on a fully constrained four-wheeled mobile robot platform to validate the effectiveness of the proposed approach. Experiment results show that the system successfully achieves intention recognition and gait adaptation, reducing user energy expenditure by up to 62% at the same walking speed, and achieves an emergency response time of less than 150 ms.