SERVICE ROBOTS FROM EARTH TO SPACE... AND BACK

Transcription

1 SERVICE ROBOTS FROM EARTH TO SPACE... AND BACK Paolo Fiorini (1), Erwin Prassler (2) (1) Department of Informatics University of Verona Strada le Grazie 15, Verona, Italy (2) FAW University of Ulm Helmholtzstr. 16, D Ulm, Germany 1. INTRODUCTION Service applications of robotic technologies begin to appear in several aspects of daily life, motivated by a number of different reasons. One of the most significant is the over-aging and decreasing of the population in Europe that will cause dramatic social changes in all European Countries. The poor state of most of Health Care Systems in Europe is not promising the level of assistance and the resources needed by the aging population, and social as well as technological counter-measures are being considered. Short of indiscriminately opening the borders to a large number of immigrants to fill the expected needed positions, a solution which may create significant social tensions and goes against some popular favor, an ambitious alternative is to foster emerging new technologies, especially robotics, and direct them to the development of versatile mechatronic and robotic assistant systems. Service robots or robot assistants might be used to automate burdensome daily routine tasks such as cleaning, housekeeping, personal logistics. In a professional environment, they may support an elderly worker in tasks that demand physical endurance. It is foreseeable that in not too distant future, robotic devices assisting and serving humans in domestic as well as professional environment will become as common and easy to use and to program as today's VCRs or dishwashers. Unlike industrial robots, service robots are not designed for industrial automation. They can improve the standards of living for all members of our society significantly, for those with limited physical abilities, disabled or elderly people, as well as for those who do not have those limitations. They will increasingly free humans from burdensome daily routines and entertain their users through theirs ability to communicate at the same time. The ultimate service robot will be a robotic errand boy, a combination between a small collapsible indoor fork lift and a mobile manipulator, which can lift and carry heavy objects and can be programmed and/or controlled by a haptic interface. This ideal service device is also the target of some research work in the area of assistant robots for space operation and exploration. Although some of Earth service robots have already reached the commercial stage, space service robots are still being developed using ground-breaking technology, which hopefully will make its first appearance in space operations in a near future. The most advanced of such prototypes is NASA Robonaut, or Robotic Astronaut, which consists of a dual arm, single legged robot of human-like dimensions and dexterity capable of using tools and passages designed for humans. The evolution of Robonaut for planetary exploration is envisioned as the robot mounted on a wheeled base, capable of dexterous, dual arm mobile manipulation. Parallel to this development, an evolution of Mars rovers is also being proposed, equipped with an armed turret for astronaut support. However, planetary servicers do not need to have anthropomorphic features to support human exploration. In the last few years, NASA has proposed a number of specific adaptations of rovers to perform useful functions, in light of a future robotic colony, precursor to a human colony on Mars. Robots will have repair capabilities, will be able to find and retrieve damaged rovers, and will be able to do autonomous assembly and disassembly and perform true teamwork. One the main problems of human robot co-existence and cooperation in space is safety. Even though safety on Earth is also a critical aspect of service robots, in space safety must take the most prominent role, and be enforced actively in every robotic device involved in human cooperation. Active safety means that robots must be able to foresee dangerous situations and actively prevent them, even when caused by human actions. Robots must be aware of their surroundings and maintain themselves out of all possible situations that can harm humans. The need to develop robotic assistant for astronauts is motivated by two main concerns: 1

2 Astronaut time is a valuable resource. Any assistance that can offload routine or repetitive tasks will free up the astronauts to focus their expertise where it is most needed. Robots can perform simple inspection and maintenance tasks, as well as scout terrain, build maps, and gather field samples. The space environment is extremely hazardous. The spacesuits that enable people to work in and explore this environment also impose severe constraints on mobility, dexterity, communication, visibility, and strength. Robots can enhance astronaut capabilities in these areas, creating a human/robot team that can accomplish tasks that neither one could do alone. This paper discusses some of the current advances in service robotics and compares and contrasts them with the research and development of assistant robots for space operation. In particular, next Section will summarize some of the main systems developed for Earth use, and service robots proposed for space applications. Section 3 will briefly describe some prototypes of space service robot. Section 4 will discuss some of the key technical aspects that are specific of space servicers. Finally Section 5 will present some conclusions and hopes for future development in this area. 2. RELATION TO PRIOR WORK In the past, the interaction of Humans and robots has been studied primarily in the context of Earth applications, such as medical and assistant robotics. Early medical applications have been based on the teleoperation of a robotic device by an expert surgeon to improve quality or performance. In [1] a master-slave robotic system for eye surgery is described. Its objective was to provide order-of-magnitude improvement of the surgeon precision during retinal procedures. In this system, however, the robot is optimized for a specific procedure, and the operating Human interacts with the master system, i.e. the low power component of the system, while the slave operates on the anesthetized patient. Applications closer to service robots for space are also in orthopedic surgery. In this case, the surgeon and the power manipulator work in close vicinity. However, in the case of hip surgery [2] the motion of the robot is decided in advance during preoperative planning and after extensive calibration and preparation the robot executes the planned motion. A closer cooperation is during robot-assisted spinal biopsy [4], where the surgeon positions by hand the robot end-effector consisting of a surgical needle. In this case, however, the robot simply moves pushed or pulled by the surgeon. An application where robots are used as strength enhancement is described in [3]. Here, however, robots are special purpose devices, often with an exoskeleton structure, designed to amplify the force output as a direct extension of the user s limbs. A passive approach to Human-robot cooperation is described in [9] and [10] where devices called Cobots are used to guide human interaction with machines. Cobots provide power assist and virtual guiding surfaces to let humans safely perform heavy or hazardous tasks. Human-robot interaction is also being studied in the context of automated assistance to the elderly and disabled. One approach is to engineer the environment to make it compatible with the robot. The robot is then programmed to execute a set of basic motions, which can be possibly combined by the user [5]. When robot and user actions cannot be precisely predicted, the robot arm is made soft and compliant, in particular using pneumatic actuators, such as those described in [6]. The integration of both methods leads to the development of a robotic room [7] where a compliant see-through manipulator is used to feed bed-ridden patients. In this case, safety is achieved by passive means, i.e. by ensuring that the robot will not damage anything, even in case of failure, since the patient is not expected to hit the robot inadvertently Reasoning and task understanding have been extensively studied in the past especially with reference to automatic task and path planning. However, the specific aspects of human task monitoring have received little attention. In [8] a neural network was trained to recognize the phases of a simple peg-in-hole task, with the objective of providing performance and safety monitoring. However, task representation was not considered in a generic form amenable to instantiation in different contexts. The control aspects of robot interaction have been studied mostly in the context of control theory, by developing algorithms to control the force with which a manipulator interacts with its environment [15], or with another manipulator [14]. However, most of the work done in this area is not relevant to this task, because it does not involve a human as one of the interacting parties. More recently algorithms for robot-human manipulation of a common object have been developed [11], [12], [13]. These schemes use a form of impedance control and are limited to the simple task of carrying a common object in the x-y plane. 2

3 Figure 1: An early concept of Mars service Robot. This brief description of related work points out that the problem of Human-Robot cooperation is only starting to be addressed by current robotics research. It must focus on integrating sensing, control and task knowledge into new paradigms, as opposed to previous work where each area was studied in isolation. We call this new paradigm awareness to indicate the robot ability to discriminate between acceptable and dangerous situations, together with its capability to react in a suitable, safe way. This capability is achievable not by developing new sensing, control, or reasoning methods, but by integrating state of the art algorithms into a new and efficient paradigm. 3. PROTOTYPES OF SERVICE ROBOTS Service robots need not be of anthropomorphic form, although they are those receiving the largest attention and press coverage, Especially in space applications. In the next two sections some of these devices wil be briefly described 3.1 Space Service An earlier proposal developed at NASA-JPL is shown in Figure 1, representing a wheeled rover, with a turret with dual arm, for astronaut support during exploration. The same idea is now pursued by NASA-JSC as the Extra-Vehicular Activity Robotic Assistant (ERA), shown in Figure 2. ERA is a wheeled robot used as a test bed for research into the requirements for successful collaboration between suited astronauts and autonomous robots. Field tests are carried out with suited astronauts and the results are used to improve the robot and/or spacesuit, enabling the team to accomplish more complex scenarios. Figure 2: The NASA ERA system. 3

4 The goal of this project is not to develop a robot for an actual manned mission, but rather to develop and maintain a functional test bed that enables scientists to establish the requirements for such a robot. Much of what is learned about human/robot interaction by the use of this test bed will be applicable to the design of autonomous robotic assistants for low earth orbit and zero-g applications as well as planetary surfaces. The EVA Robotic Assistant is equipped with many sensors, including GPS, stereo vision for obstacle avoidance and path planning/navigation, stereo vision for astronaut tracking, laser scanning range-finder, 6-axis accelerometer, compass, and inclinometers. It also has many ways to interact with its environment: 6-degree-of-freedom manipulator, pan-tilt-verge active camera platform, and drive platform that can handle rugged terrain while carrying tools, samples, and equipment and pulling a trailer. The processing is all done on-board on four PC/104 Pentium II computers using the Linux operating system, and using the CORBA standard for inter-process communication The best example of anthropomorphic robot proposed for space applications is NASA-JSC Robonaut, shown in Figure 3. The Robonaut project seeks to develop and demonstrate a robotic system that can function as an EVA astronaut equivalent. Robonaut keeps the human operator in the control loop through its telepresence control system. Robonaut is designed to be used for "EVA" tasks, i.e., those which were not specifically designed for robots. The challenge then is to build machines that can help humans work and explore in space. Central to that effort is the dexterous manipulation capability which should exceed that of a suited astronaut. A humanoid shape is used to meet NASA's increasing requirements for Extravehicular Activity (EVA), without having to forgo the work done in the past five decades, space flight hardware has been designed for human servicing. Space walks are planned for most of the assembly missions for the International Space Station, and they are a key contingency for resolving on-orbit failures. Combined with the substantial investment in EVA tools, this accumulation of equipment requiring a humanoid shape and an assumed level of human performance presents a unique opportunity for a humanoid system. While the depth and breadth of human performance is beyond the current state of the art in robotics, Robonaut's design goals targets the reduced dexterity and performance of a suited astronaut as ranges of motion, strength and endurance capabilities of space walking humans. 3.1 Earth Service Service robots for Earth applications have a much broader spectrum of applications, and tend to be more specific in their domain. The web page gives a comprehensive list of robotic devices that are used to perform a service task. These tasks include rehabilitation, house cleaning, humanitarian demining, gas refilling, lawn mowing, search and rescue, fire fighting, agricultural, and entertainment robots. Among those, there are anthropomorphic devices, since it is foreseen that the market for humanoid dedicated to people assistance will be equivalent to the automotive market in not too distant a future. Some of the goals of Earth service robots are similar to those of Space devices, although the latter will not become commercial products, and therefore it is worth examining what are the important open issues affecting the development of service robots. Figure 3: Front view of Robonaut 4

5 4. IMPORTANT ISSUES A requirement of long duration Human space flights will be the efficient allocation of time and resources, and similarly during Earth service tasks Humans, often unskilled, will share the same workspace of robotic devices. However, robotic technology has not yet devised ways to let robots and Humans share the same workspace and cooperate on the same tasks. An important concept that is still not investigated is to give robots a sense of awareness of their surroundings by integrating spatial perception, map building, and task context into a unified model, and by using the model to produce expectations about task evolution. Expectations will then be used to identify situations that are potentially dangerous, to adapt control parameters to specific situations, and to trigger corrective actions. Awareness is instrumental to, and will be used for the improvement of, safety and contact control during Human-Robot cooperation. This feature will also be instrumental in the development of automatic sequences for the execution of repetitive tasks, both on Earth and in space, Examples of these tasks are the bolting and unbolting of panels in the Space Station, and suturing in robotic surgery. These simple, by Human standards, tasks should and could be made automatic to simplify and reduce Human intervention. Human-robot safety is currently addressed only with passive means. Robot surfaces are made compliant to avoid harsh contacts, no sharp edges are allowed, robots are prevented from moving at fast speeds, and emergency halt buttons are suitably located. Awareness will enable an active approach to safety. A robot equipped with awareness will be able to recognize when a human is too close and risking collision, or is in an unexpected position, and it will actively correct the perceived hazard. Spatial sensing techniques alone are not sufficient to provide awareness, since spatial information must be related to task execution. Therefore, task context must be instantiated in the spatial representation provided by sensing. By fusing spatial perception and task knowledge into awareness, the Human position relative to the robot can be monitored and the robot can react appropriately. Similarly to safety, contact between Humans and robots should also be awareness based. Force or impedance control methods should be enhanced by awareness to modulate the control gains according to task different phases and Human proximity. Furthermore, an added benefit of awareness is that it will simplify Human-robot communication, by providing a common semantics for task elements independent of physical variations during the identification of objects and their locations. Main technology elements contributing to the development of this new feature are: Context understanding. Consists of an appropriate framework for the development of robotic awareness, and in particular for the development of a new method for task representation that can support different environments and situations. This representation will constitute the task context component of robotic awareness. It should be characterized by geometrical, sensing and timing parameters, each with appropriate ranges, to adapt the context to each specific situation. The result of this learning will be a map of the task parameters to the current environment. By representing each element in the task context with the range of its relevant parameters, the robot will be able to recognize anomalies, generate inquiries and warning, and take corrective actions. Task-driven contact control. The service robot must have motion and manipulation capabilities, which are adaptable to different instantiations of the same task and to different task evolutions. This technology will give assistant robots the ability to change their control parameters according to different situations, thus adjusting their stiffness property to the most suitable for the current context. The robot control mode should be a function of the context in which it operates. Therefore the robot should be able to identify contexts of operation by using virtual sensors for detecting transitions of contexts. Examples of changes in context are an astronaut entering the robot workspace, or the release of a control joystick after a teleoperated task segment. Transitions could also occur upon voice or other input commands. Robot awareness. Finally, the integration of perception and action should give the robot awareness, i.e. the contextdependent interaction of perception and action. The robot should be capable of mapping a task description into a specific physical and temporal scenario, i.e. specific object distribution and action timing, and of moving and manipulating the required objects according to the current context. System performance should be quantified in terms of the differences between the generic task description and its instantiations in terms of percent variation of its parameters. 5. CONCLUSIONS In this paper we briefly review some of the developments and of the issues related to space and Earth development and application of service robots. After a short review of past work in this area, the main space service robots currently being developed, ERA and Robonaut, have been described, and then some reference has been made to Earth service robots. A common requirement of these devices is the capability of carrying out brief segments of unsupervised tasks, 5

6 adapting a general model to specific tasks. This requirement is essential both in space and on Earth. For example, in space it will refer to the repetition of simple actions, such as bolting and unbolting, replacement of ORU, and cleaning of optical surfaces. On Earth it may refer to the automatic execution of a suture during a robot-aided surgical operation, or the refill operation of different car brands, or to the picking of different types of fruits. In space as well as on Earth, the robot needs to adapt a memorized, general sequence to the specific case, thus the need of introducing the new concept of robotic awareness, which has been briefly detailed above. This concept is being pursued by various research groups, in particular in the context of robotic surgery, and it may be a precursor to the introduction of limited automation in the operating room. 6. REFERENCES [1] H Das, H. Zak, J. Johnson, J. Crouch, D. Frambach, "Evaluation of a telerobotic system to assist surgeons in microsurgery", Computer Aided Surgery, 4(1), [2] R. Taylor, et al., An Image-Directed Robotic System for Precise Orthopaedic Surgery, Computer Integrated Surgery, R. Taylor, S. Lavallee, G. Burdea and R. Mosges Editors, MIT Press 1996, pp [3] H. Kazerooni, The Human power amplifier technology at the University of California Berkeley, Robotics and Autonomous Systems, 19 (1996) pp [4] C. Casadei, P. Fiorini and S. Martelli, A Workcell for the Development of Robot-Assisted Surgical Procedures, Journal of Computer and Control Engineering (submitted). [5] L. Leifer, Tele-service-robots: Integrating the social-technical framework of human service through the Internet World-Wide-Web, International Workshop on Human-Robot Symbiosis, May 18-19, 1995, Tsukuba, Japan. [6] R. Peters et al., Visual Servoing with the Bridgestone soft arm, International Workshop on Human-Robot Symbiosis, May 18-19, 1995, Tsukuba, Japan. [7] T. Sato, The Robotic Room: a Room taking care of sick people with human-centered Robotic components, Workshop on Healthcare Robots, ICRA 97, April 21, 1997, Albuquerque, NM. [8] P. Fiorini, A. Giancaspro, S. Losito, and G. Pasquariello, Neural networks for the segmentation of teleoperation tasks, PRESENCE, Teleoperators and Virtual Environments, MIT Journal, 2(1), pp. 1-13, [9] J. E. Colgate, W. Wannasuphoprasit, M. A. Peshkin, Cobots: Robots for Collaboration with Human Operators, Proceedings of the International Mechanical Engineering Congress and Exhibition, DSC-Vol. 58, Atlanta, GA, 1996, pp [10] W. Wannasuphoprasit, P. Akella, M. Peshkin, J. E.Colgate, Cobots: A Novel Material Handling Technology, Proceedings of International Mechanical Engineering Congress and Exposition, Anaheim, CA, [11] Y. Yamamoto, H. Eda, X. Yun,"Coordinated Task Execution of a Human and a Mobile Manipulator", Proceedings of the 1996 International Conference on Robotics and Automation, pp , April 1996, Minneapolis, Minnesota [12] K. Kosuge, H. Yoshida, D. Taguchi, and T. Fukuda, "Robot-human collaboration for New Robotic Applications", Proceedings of the 1994 International Conference on Robotics and Automation, pp , 1994 [13] R. Ikeura, H. Inooka, "Variable impedance control of a robot for cooperation with a human", Proceedings of the 1995 International Conference on Robotics and Automation, pp , 1995 [14] R.G. Bonitz and T.C. Hsia, "Robust Internal Force-tracking Impedance Control for Coordinated Multi-arm Manipulation - Theory and Experiments", 6th International Symposium on Robotics and Manufacturing, 2nd World Automation Congress, [15] H. Seraji, D. Lim, and R. Steele, "Experiments in Contact Control" Journal of Robotic Systems, 1996, 13(2), pp