DRIVERLESS WHEELCHAIR IN HOSPITAL TRANSPORTATION
In a hospital environment there are many services where external consultations, examinations, analysis and surgeries are carried out that involve the displacement of patients that go to these services, which deserves some attention especially in situations of people with reduced mobility, such as citizens with disabilities. Nowadays, the transportation of patients in hospitals is done by either patient transporters, non-specialized personnel, which is not very pleasant; or nurses who are specialized but lack of time. It could often be done using a wheelchair, and that would be more pleasant and done in an autonomous way. As a result, it would avoid delays and failures due to the lack of personnel to carry out the transportation. A driverless wheelchair system has been investigated and developed by Portuguese researchers in collaboration of local Health Ministry. The prototype of this system this prototype is according to the concept of an intelligent wheelchair, capable of autonomous navigation with safety and avoiding obstacles, flexible and robust interaction with the patients and also communicating with the hospital system, other devices and other wheelchairs. To achieve the intelligence on the wheelchair, the functions of Multiple forms of interaction with the user, autonomous navigation in dynamic environments and communication with other devices (i.e. elevators, service robots, other wheelchairs) have to be fulfilled. The system is retrofitted with the necessary sensors, actuators, power systems and computing units to perform the autonomous functions safely and comfortably. Navigation refers to the robot’s ability to represent the environment, determine its own pose and execute a trajectory towards its goal. Mapping is performed with GMapping, a Rao-Blackwellized particle filter to create grid maps from LiDAR data. In its current implementation mapping consists of two steps performed one time for each desired environment. First the wheelchair is driven manually around the environment while recording the laser scan and odometry data of the trip. Next, GMapping processes the raw sensor data and creates a 2D occupancy grid map. The map is then made available to the localization system through a map server. The implemented localization mainly consists of a map-matching algorithm referred as Perfect Match (PM). To estimate the position of the robot, PM minimizes the matching error between the data acquired by the LiDAR and the 2D environment map stored in the robot’s database. A Kalman Filter is used to merge odometric data provided by wheel’s encoders with those from LiDAR sensors, providing a more robust pose estimation. The trajectory controller receives information about the trajectory from the mission control algorithm and outputs linear and angular references to the motor controller. Decision consists on the wheelchair’s ability to determine a trajectory it has to perform to transport a given patient to the operating room. Predetermined paths are stored in a directed graph, allowing a high degree of motion predictability and repeatability. In the present scenario those are important features considering that the wheelchair will share the environment with humans. Graphs are constituted by vertexes (possible stopping poses) and edges (parametric curves that connect two vertexes). Each edge has its own linear velocity associated, which is constant when the robot is moving along that specific path. On the other hand, each vertex may have some actions associated with it, such as exchange map or interact with elevator. The specific trajectory performed by the robot for each objective is calculated and optimized by a modified routing A* algorithm that takes into account the speed of the obstacles. The system interface allows information flow between the wheelchair and external systems. The Humam-Machine Interface displays information to identify the user it is transporting, its origin and destination. It also displays virtual buttons to allow the health care team start, cancel and finish the trip. The other two interfaces (LIGHt interface and Elevator interface) allow communication with the hospital information system and the elevators, and are further detailed in the next Subsections.