Powered wheelchairs play a vital role in bringing independence to people who suffer from a wide range of mobility impairments. However, some wheelchair users struggle to drive safely and eﬀectively. To address these issues, a few fully autonomous robotic solutions have been developed, which limit the level of control the user can exert on the system. However, wheelchair users typically want to be active drivers. Therefore, to bridge this gap, we propose a collaborative control mechanism. Our system aims to assist wheelchair users to drive safely, without inhibiting their inherent capabilities. Using a multiple–hypotheses method, the collaborative controller continuously predicts the short–term goals of the user and calculates an associated conﬁdence of the prediction. If the system determines that the user cannot successfully achieve this target state, it will pro–actively provide assistance, based upon the aﬀordances of the surroundings. In contrast, when it determines that the user is capable of manoeuvring safely without any assistance, it continues to update its hypotheses, but does not intervene. In a series of experiments, we show that our collaborative controller enables wheelchair drivers to manoeuvre safely and eﬀectively. It allows high precision manoeuvres to be performed without the need for precise input signals from the user. Conversely, we demonstrate that without the assistance, performance was suboptimal; many participants exhibited erratic joystick movements and experienced multiple collisions. Moreover, experimental results suggest that our approach can even help proﬁcient wheelchair users under certain conditions, i.e. when they are under a heightened workload. Our approach to collaborative control can empower wheelchair drivers of diverse proﬁciencies to manoeuvre eﬀectively and safely, especially when their input is imprecise or their attention is limited.