 As part of the NASA tele-robotics program, the Jet Propulsion Laboratory is pursuing advanced technology in remote manipulation in two discipline areas, tele-operation and supervised autonomy. In tele-operation, an operator uses hand controllers to command the robot's motions directly. In supervised autonomy, an operator interacts with the system at a much higher level, while relying on the robot's information of the task and environment to generate its own motion commands. The results of this research will help to develop the technologies required to fulfill NASA's potential future needs for space robotics. Since its beginning in 1985, the broad objectives of the JPL program have been to develop the robotic technologies that would be used in space operations and to integrate them into systems. These technologies would be used to increase the productivity of astronauts, conduct operations too dangerous for humans in space, or perform boring, repetitive tasks. To date, the integration of these new technologies has resulted in the development of a single robotic system that can operate in tele-operation or supervised autonomy. In the future, the system will be able to operate in hybrid modes that combine features of both called shared control. This tele-robotic test facility provides an environment for research and development of key robotic technologies. Several features will be highlighted. Tele-operation with force reflection, autonomous compliance strategies, and operator designate to support the operator's use of supervised autonomy. This demonstration will simulate the removal of an orbital replacement unit, an ORU, from an Earth-orbiting satellite. An arm will remove the ORU and will then hand it to another manipulator arm which will insert it into a storage rack. This sequence represents a portion of an in-flight maintenance operation similar to those which will be performed on the space station. This demonstration utilizes a combination of tele-operation and supervised autonomy. These two manipulators, left and right arms, are used for grasping and manipulating objects. This third manipulator is the camera arm and is used to position and orient four cameras. These four cameras, combined with three wing cameras, provide the operator with multiple views of the workspace. Two of these cameras are used to provide the operator with a color 3D stereo display. This truss structure simulates the body of a satellite or the space station and supports a mock-up of an ORU. To simulate the uncertainties that may exist in the space environment, the initial location of the truss is not known to the robot beforehand. The module represents an ORU and has two grapple lugs. The robot carries a picture of this module in its database. However, one of the grapple lugs, the left one, has been intentionally misaligned to simulate a modeling error. The experiment will show how the robot deals with this. In addition, a cylinder has been placed around the grapple lug to simulate occluded viewing conditions that might be expected in space. The two arms used for grasping have wrist sensors, which detect forces and torques. The information received from these sensors is used in force reflection and compliance. Because the position of the truss is not known to the robot at the beginning, the operator will first need to determine where the actual ORU is located. The operator begins the task by using a procedure called operator designate to determine the position of the module body. When the module body is located, the position of the grapple lug can be determined from the picture of the module carried in the robot's database, even though the grapple lug cannot be seen. On two monitors, each showing a slightly different view, the operator overlays a line drawing of the module. The operator then specifies the true locations of several corners of the module body by indicating which corner of the overlay corresponds to which corner of the video image. Because the camera images have been carefully calibrated, this information can be used to update the robot's database. With the module's accurate position stored in its database, the robot can then locate the grapple lug. With this accomplished, the system then moves the left arm toward the left grapple lug for the grasping operation. This is the lug that was intentionally misaligned. Using active compliance, the left arm gently and autonomously moves in several directions to accommodate the misalignment and allow the grasping action to be completed. Since October 1989, the speed with which these operations can be performed has been increased by a factor of three. This is because we have introduced the capability to plan actions at the operator's site while simultaneously executing other commands at the remote site. In tele-operation, the operator uses the force-reflecting hand controllers and looks at the 3D display to view the workspace while he removes and transports the module. Dr. Henry Stone. A force-reflection is a mode of tele-operation in which the forces that are sensed at the end effector of the arm are reflected back to one of these hand controllers here, and that provides the operator with a sensation of contact with the environment. Should the geometry prevent the left arm from storing the module, the module would have to be handed off to the right arm. In this case, the left arm holds the module and the right arm performs a grasp sequence identical to that just executed. We have not placed a cylinder around the right grapple lug so we can provide a better view of the upcoming autonomous error recovery sequence. Because the left grapple lug was misaligned, the actual position of the right grapple lug is significantly different than the robot's model would predict. This causes the automated grasp to fail. The operator is informed of the failure and decides to check the database by displaying the image of the module on the video monitors. The error is apparent. Because the right grapple lug line image on the video monitor and the module do not coincide, the operator can see that the cause of the failure is due to the mispositioning of this grapple lug. To recover from this failure, the operator again chooses object designation to correct and update the database model. With this new information in the database, the operator executes the autonomous sequence successfully and grasps the module. The operator uses tele-operation to release the left side of the module and move the left arm away. The operator also uses tele-operation to move the right arm, inserting the module into its storage location and completing the ORU removal task. In addition to tele-operation and supervised autonomy, this system is also capable of combining the two modes into one called shared control. At this time, it has been partially integrated into the system. During shared control, the operator is using the hand controller to basically guide the position of the end effector, whereas the autonomous system at the same time is controlling forces and torques to maintain a particular profile. Shared control will now be used to simulate an optical cleaning task. This dome represents an optical surface. The operator sits at the console and commands the arm to grasp the cleaning pad autonomously. The operator commands the movements of the pad using a hand controller. The autonomous system ensures that the pad follows the contour of the dome and controls the pressure exerted on it. During this operation, only the planar motions of the hand controller that I'm making are being mapped into motions of the pad about the dome. The other degrees of freedom are being controlled by the autonomous system in order to maintain certain forces and torques that are required for the actual polishing to occur. The demonstrations you have seen represent technologies being advanced to meet the needs of NASA's space program. These demonstrations at the JPL tele-robotic test facility showed a state-of-the-art integrated robotic system capable of performing complex sequences of operations using a mixture of autonomy and tele-operation, where the operator has the ability to select the mode which is best suited for the task at hand. In the ORU exchange task, the system coped with uncertainties and inaccuracies in its database by using the features of active compliance and operator designate. Operator designate was used to update the position of the ORU so that the occluded grapple lug could be found. Active compliance then permitted the grapple lug to be grasped, even though it was misaligned. Later, operator designate was used again to overcome the error in the assumed position of the right grapple lug caused by the misalignment of the left one. With the operator-controlled polishing of the dome, we showed one example of where supervised autonomy was combined with tele-operation to tailor specific capabilities to specific task requirements. This system and the robotic technologies that are integrated within it as direct application to a variety of operations that will be required in space in the future. In particular, it can be used for satellite servicing and repair and maintenance of the space station freedom. Flexibility and adaptability will be essential capabilities needed for tele-robots used on future space missions. As shown by these demonstrations, these qualities are important goals of the JPL program.