T2-[Rolland] - Überblick

A Safe Wheelchair with Driving Assistance

The wheelchair prototype platform set up by Otto Bock Healthcare GmbH for T2-[Rolland].

 This project aims at increasing mobility and autonomy of users who are presently unable to safely operate an electrical wheelchair. The Driving Assistant compensates manual impairment, e.g. by avoiding obstacles and passing through doorways. The Qualitative Driving Assistant supports users who cannot continuously control the wheelchair but can give intermittent discrete commands. Both assistants can be used with the wheelchair's joystick or - for paralyzed users - with a special head-joystick.

Throughout the whole project we intensively cooperate with Stiftung Friedehorst working with patients and senior citizen to get feedback about the assistants as to their everyday usability. The evaluation will be carried out both in the building Cartesium (housing the SFB/TR 8) in the Bremen Ambient Assisted Living Lab (BAALL), and on the premises of Stiftung Friedehorst, with expert advice and under the guidance of Stiftung Friedehorst.

 

Driving Assistant

Virtual Sensors, i.e. precomputed descriptions of imminent movements, are the key concept of the Driving Assistant.

 The Driving Assistant provides a wheelchair interface for patients or elderly persons who are able to operate a joystick, but have severe or occasional problems to do so due to motoric deficits such as tremor or spastic movement. Often such users do not use or are not allowed to use an electrical wheelchair at all because they might endanger themselves or other people. Furthermore it addresses patients who can use the head-joystick to control the wheelchair but have similar problems. For the head-joystick we expect this to be rather common because it is more difficult to control the wheelchair with head movements. Regardless of the input device the Driving Assistant prevents collisions by automatic braking or obstacle avoidance. It corrects the user's joystick command when passing through doorways or for automatic turning, etc.; no precise joystick control is required, it suffices that the user gives the general direction.

 

Qualitative Driving Assistant

The Qualitative Driving Assistant maps user's gestures, e.g. a dedicated pitch movement of the head, to a metrical target pose, and navigates to that goal by the application of an A* algorithm.

The Qualitative Driving Assistant provides a wheelchair interface for patients who cannot control the wheelchair in a closed loop fashion but are still able to give discrete commands intermittently. These commands could come from any interface, in our case either the wheelchairs joystick or the head joystick. We start our investigation with the command set “rotate left”, "left", "go ahead", "right", "rotate right" which already has proven to be effective in previous studies and will investigate extension towards an 8-direction star.

The first step is to detect these discrete commands from the continuous (head-) joystick readings, i.e. from single gestures performed by hand or head. For such a small set this is not difficult, so we apply a heuristic classifier. However, it requires careful fine-tuning because the quality of recognition greatly affects how conveniently patients can use the Qualitative Driving Assistant.

Algorithmically the critical part is to find a set of considered paths from laser scan data and to evaluate which fits best to the user’s command. We do so by evaluating Voronoi graphs derived from a local occupancy grid. The actual target pose is navigated by an A* graph search algorithm, working in the configuration space of the wheelchair.

 

Experimental Evaluation

With the intention to assess whether the Driving Assistant, and the Qualitative Driving Assistant improve the quality of life of the users, an exemplary pilot study, and a clinical study in cooperation with the Neurologisches Rehabilitationszentrum Stiftung Friedehorst were conducted in the first year of T2-[Rolland].

In the first usability test at Stiftung Friedehorst, four patients were asked to navigate Rolland on a wheelchair slalom course using either a standard joystick or the headjoystick, both supported by the Driving Assistant module. Prior to the planned everyday usability evaluation, this experiment monitored each subject’s capabilities to navigate a non-trivial course (standardised by FEDPC) including several 360°/180° turns, a figure of eight, and an L-shaped upwards/downwards ramp.