Model Predictive Control-based haptic steering assistance to enhance motor learning of a bicycling task: A pilot study

Abstract

Learning to ride a bicycle is challenging, and can be dangerous, as it involves acquiring several motor skills, including balance and coordination while interacting with a complex dynamical system. Haptic assistance could potentially help to enhance the learning of this especially complex task in a safe environment. We propose the use of a Model Predictive Controller (MPC) to provide steering assistance while training to learn a complex cycling task. We conducted a feasibility study with ten participants riding a steer-by-wire bicycle on a treadmill. The goal of the task was to collect laterally positioned virtual stars, shown on a display mounted in front of the treadmill. Participants trained under two conditions in random balanced order: with or without assistance from the MPC. Short-term learning was compared between conditions. We did not find evidence that training with MPC-based assistance improves the performance in steering the bicycle to collect the virtual stars after training compared to training without assistance. However, we found initial evidence that training with the MPC could be beneficial for less-skilled cyclists to learn the steering task. In conclusion, haptic steering assistance using MPCs may be a promising tool for enhancing bicycle steering skills, especially in initially less-skilled bicyclists.