Some applications of wearable robots (e.g. exoskeleton robots) focus in assisting humans in physically exhausting tasks such as carrying a heavy tool or backpack, for example. In these cases it is desired that the robot does not impede or resist the users movements, but only provides extra force for execution of the task in question. In other words, ideally, the robot should be imperceptible (i.e. transparent) to the user. Making a wearable robot imperceptible to the user is an extremely hard and complex control goal to reach satisfactorily. The constant and intrinsic interaction between human and robot and varied user-dependent behaviors are the main difficulties in the design of such cooperative control. In this work, we propose a new approach to transparency control. Its a learning and movement prediction algorithm based on segmentation and grouping of healthy user experimental data. From this data, we wish to predict the dynamics of trajectory segments and also the sequence of such segments. The gait style we will focus on is walking.
News published in Agência FAPESP Newsletter about the scholarship: