Figure 1. Schematic diagram of the cylindrical gel actuation system. The apparatus was composed of a single mechanical joint, single stage cable driven transmission, passive antagonist spring, optical joint position sensor, tendon force gauges, and servo controlled valves.

Figure 2. Schematic diagram of control system.
Figure 3. Schematic diagram of the cylindrical polymer gel actuator using PAN gel fibers. Perforated Teflon irrigation tubes were connected to a common fluid manifold allowing only two input lines from the fluid control valves. The distribution of tubes among the fibers attempted to minimize fluid diffusion time to maximize overall actuator performance. The second view shows the arrangement of gel fiber bundles, and acid and base conduits in cross section.

Figure 4. A planar array of PAN gel fibers was positioned between two reservoirs containing acid and base respectively. The reservoirs could alternatively flood the fiber chamber simultaneously through perforated sheets. A common mechanical coupling transmitted and balanced the loads from the individual fiber bundles.

  • Novel stimulation schemes It became clear, from our initial research, that a suitable gel material was only part of the problem. Polymer gel is essentially {\em passive; that is, it only responses to {\em changes in the surrounding environment. Thus changing the external environment rapidly is central to effective gel based actuator design. Our approach has been to use an integral irrigation system to affect simultaneous pH changes across arrays of gel fibers. This produces a dramatic response, as well as allows the use of an inherent chemical energy storage system. We have explored novel irrigation schemes including arrays of interstial porous tubes, parallel porous planar sheets, flexible fluid reservoirs, miniature modular gel/tube units, and gel-based fluid conduits. In collaboration with Dr. James Weaver in the Biophysics department and Prof. Alex Klibanov in Chemistry, we have begun the study of electrically controlled biological membranes, to regulate the release of enzymes which catalyzes a reaction whose produces produce pH changes.

  • Mechanical modeling One of our strengths is mechanical modeling. In collaboration with Sandia National Laboratory, we have developed physical models of the actuator, including finite element analysis (FEA) and lumped parameter models of gel, semi-continuous models of the fluid system, and parametric models of the surrounding mechanics and transmission system. The construction of complete, integrated physical representations is unique in gel based actuator design. Our models and simulations have allowed us to test and optimize actuator design, including irrigation system dimensions, pore size, fiber placement, flow rates, and pH levels. For example, acid and base infusion as a function of pore size and placement is illustrated in figure 5. The use of dynamic models and computer simulations will continue to allow us to optimize actuator parameters and maximize system performance.
    Figure 5. Optimizing pore diameter produces a more uniform hydrogen ion concentration along the fiber. Different line types are used to represent flow and ion concentration at six locations along the length of the fiber.

  • Advanced control algorithms The nonlinear nature of polymer gel, suggests nonlinear techniques for actuator control. An area of expertise, in which the Robotics Group in the AI Lab is particularly adept, is the application of advanced control theory to robotic systems. We have, in simulation, successfully applied a nonlinear sliding mode control system to a gel based actuator. A well controlled actuator is critically important for any practical application of polymer gel systems. Therefore, we propose to fully exploit our ability in advanced control theory to optimize the performance actuators which we construct.

    D. Deliverables

    The proposed project will yield scientific and technological results. These results will be presented in journals articles, major conference proceedings, and ARPA reports. At the end of each year of the project, a demonstration will be conducted and a professional quality videotape will be produced.

    E. Technical Plan

    E.1 Statement of Work

    F. Cost Estimate

    G. Appendix

    G.1 MIT Artificial Intelligence Laboratory

    The AI Laboratory has a long history of making fundamental contributions in the fields of Artificial Intelligence and Robotics. For more than 20 years the lab has demonstrated the ability produce innovative mechanical systems, algorithms and computer architectures. Currently employing over 20 faculty and full time research staff, as well as over 100 graduate students, it provides a vast resource of technical expertise in computational systems, sensory interpretation, mechanical devices and realtime control systems directly relevant to this project's needs. In addition, the facilities of the Department of Mechanical Engineering and the Artificial Intelligence Laboratory shop facilities will be available on an as-need basis. These include extensive mechanical and electronic fabrication facilities. At the AI Lab a large network of Sun Workstations is available and provide powerful computational resources

    G.2 Robotics Group at the MIT AI Lab

    Over that past 10 years at MIT and elsewhere, Dr. Salisbury and his associates have focused on the development of electromechanical systems utilizing novel transmission, actuation, and control schemes. These devices have ranged from joysticks to robotic hands and arms. The underlying capability emphasized in all these devices has been the ability to control and sense contact forces. Mechanical and servo design techniques have evolved through these experiences that enable our team to design relatively complex force-exerting devices with high performance and low cost. Central to our view of performance are stringent demands on bandwidth, dynamic range (ratio of maximum to minimum exertable and sensible forces), and backdrivablitity. These have resulted in techniques utilizing stiff, backlash-free, and extremely low friction transmissions to efficiently couple responsive electric motors to stiff, low mass mechanisms. To date, we have employed traditional engineering materials such as aluminum structures and stranded steel cable; it is clear that use of higher performance composite structural materials (carbon fiber) and high strength synthetic cables (Spectra, Vectran) will further improve our systems. A complimentary area of expertise that has evolved within Dr. Salisbury's group is the ability to design multi-axis force sensors. Used to monitor and control contact force interactions, we have found these devices to be a robust and performance enhancing adjunct to our mechanical systems.

    G.3 Investigators

    David Brock is a Research Associate at the MIT AI Lab and a Postdoctoral Associate at the MIT Research Laboratory for Electronics. He completed his Ph.D. at MIT in 1993 on methods for automated robot grasping and is currently responsible for a significant portion of the Virtual Environment Training Technology (VETT) haptics project. In 1991, with funding from the Sandia National Laboratory and EXOS Corporation, he formed an Aritificial Muscle Laboratory within the MIT AI Lab for the research and development of polymer gel based actuators.

    Ken Salisbury is Principal Research Scientist at the MIT Artificial Intelligence Laboratory. He has been at MIT since 1982 when he received his Ph.D from Stanford University. He is well known internationally for the design of robotic devices and algorithms for force interaction. He designed the JPL six-axis force-reflecting hand controller and the Stanford-JPL three-fingered hand ("Salisbury Hand"), and is co-inventor of the WAM robot arm, and the MIT Thimble-Gimbal force feedback device ("PHANToM"). He holds significant design patents and patents pending for his work on robotic arms, hands, and force sensing technology and has served extensively as a consultant to industry and government laboratories.