A Machine Learning Approach to Movement Intention Estimation Using Rollator-user Interaction Force

Rollators have gained popularity as rehabilitation and assistive robots due to their simplicity and compact size. Equipped with sensors and actuators, robotic rollators can provide guidance, assistance, and rehabilitation training during walking. However, existing assistance and training primarily support straight walking, presenting a challenge in providing intuitive assistance during turns. This research introduces an approach for predicting user intention during walking solely using the interaction force between the rollator and the user. A Long Short-Term Memory (LSTM) network-based machine learning method is proposed for establishing a movement intention model, accurately estimating the intentions of “Straight Walk”, “Turn Left” and “Turn Right”. Experimental results with elderly participants demonstrated a high accuracy of 92.85% in estimating movement intentions.