An interactive robotic workstation for applications in rehabilitation

A novel approach to controlling robots has been developed which forms the basis of an interactive robotic workstation for use in educational and vocational settings by severely physically disabled people. It is felt that the inability of the severely physically disabled to participate in educational and vocational activities is among the most damaging consequences of physical disability. The aim of this research has been to increase the level of physical interaction that a disabled person has with the environment through the use of an intelligent robot which is under the ultimate control of the person. This approach is in sharp contrast with mainstream robotics research and practice where the emphasis is on removing the human involvement in manual activities. The robot workstation comprises a UMI RTX robot, a purpose-built vision system to identify coded markers attached to the objects in the robot's environment, and a flexible user interface which supports commands ranging from simple switch selections to natural language instructions spoken into a voice recogniser. A robot programming language with facilities for specifying information about the robot's environment has been developed to enable the disabled person to use English-like commands to instruct the robot and to install and access the information in the data-base. Objects need not be uniquely defined but may be referenced using abstract descriptions which are resolved using the descriptive and physical information stored in the data-base. Software has been developed which uses the information stored in the data-base in order to decompose a high-level command into the series of primitive robot operations necessary to carry out the desired task. A practical on-line motion planning algorithm which interrogates the data-base and plans collision-free robot motions based on known obstacles, such as walls, shelves and other objects, has been developed. Motion between two robot configurations is done using a simple goalpost motion strategy comprising three phases of motion - a departure phase, a transport phase, and an arrival phase. The departure and arrival phases of motion are planned using a generate-and-test algorithm, while the transport phase is planned by searching for a path through an Obstacle-Free Space Map (OFSM). The OFSM, which is a quantised representation of the robot's configuration space, is generated by identifying the collision sites and therefore the collision interval for each of the robot's limbs and payload. Since the OFSM is a dynamic structure which varies over the course of an activity as objects are picked up and set down, a search algorith which directly processes the OFSM, as opposed to one which processes an alternative representation of it, was favoured. Motion planning algorithms based on fundamental search techniques have been developed to find a path between the current and desired robot configurations. During the course of this research, extensive file testing of the robot system has been carried out, and the experience gained from these tests has influenced subsequent technical developments. Although many of the applications are educational, the workstation and the principles of interactive robotics may be extended to vocational, domestic, and recreational activities.