Mobility assistance for wheelchair navigation is typically based on the shared control paradigm. Traditionally, control swaps from user to machine depending either on a trigger mechanism or on a explicit user request. Alternatively, in collaborative control approaches both user and robot contribute to control at the same time. However, in this case it is necessary to decide how much impact the user has in the emergent command. User weight has been estimated based on his/her command efficiency or on the environment complexity. However, the user’s command efficiency may change abruptly, whereas the environment complexity depends on the user’s skills. In this work we propose a collaborative control approach where this weight is determined by the user’s ability to cope with the situation at hand with respect to an average person. This estimation relies on an standard navigation skill profile extracted from a large number of traces from real users. This approach has two major advantages: i) the user receives more assistance only when needed according to his/her own skills; and ii) we avoid an excess of assistance to prevent loss of residual skills. The proposed system has been tested with a group of people with disabilities. Tests prove that resulting efficiencies are similar to other collaborative control approaches although the amount of assistance is reduced.
