Robotic manipulators are expected to perform various functions in space operations, such as assembly, inspection, repair, retrieval, and exploration. As such, they would be a vital resource on the future U.S. space station. But these robots must be light, have high-load capacity, and be able to manipulate objects smoothly and precisely. Unfortunately, these conflicting requirements rule out conventional robotic systems because either they are too heavy or they don’t operate smoothly enough.
To overcome this problem, a robotic manipulator must have:
• A mechanical drive with low friction and backlash.
• A sophisticated control system that compensates for lessthan- ideal mechanical performance and provides precise control.
Down to earth
A mechanical drive and control system that meets these requirements would also be suitable for servomechanism applications back on Earth. It could perform industrial tasks such as mechanical assembly, grinding and deburring, polishing, scribing, and riveting. Many of these tasks are now done with hand-held power tools because robots using power tools don’t provide satisfactory control of contact forces In an effort to develop a suitable robotic manipulator for space, a Cleveland engineering firm, called NASTEC, designed and built a robotic joint for the Structural Dynamics Branch of the NASA Lewis Research Center, also in Cleveland. Called a pitch-yaw joint, this flexible device is part of a 2-degree-of-freedom (DOF) robotic assembly that manipulates objects by providing rotation (pitch and yaw) around two mutually perpendicular axes.
Using a combination roller-andgear assembly to transmit torque, the pitch-yaw joint, Figure 1, combines the smooth, backlash-free characteristic of a roller traction drive with the high-torque capacity and lower bearing loads of a gear drive. At the bottom of this figure, a pair of bidirectional motors and gearboxes with a 90:1 speed-reduction ratio drives the two input stages of the joint. At the top, a 33- in.-long link is mounted to the output stage of the pitch-yaw joint, creating a 2-DOF manipulator arm capable of moving a 50-lb payload within a spherical workspace.
How it works
The pitch-yaw joint consists of a series of bevel-shaped rollers and bevel gears as shown in Figure 2. Each rollergear pair is mounted as a parallel set.
Each input stage asssembly, consisting of a bevel roller and a bevel gear (bottom), drives additional roller-gear assemblies including an intermediate and transverse stage. Finally, both transversing roller-gears mesh with the output roller-gear assembly (top). When the motors turn in the same direction at equal speed, they produce a pure yaw output motion. Turning in opposite directions at equal speeds produces a pure pitch motion. Any other combination of input motions produces both pitch and yaw motions. The ratio of input speed to output speed in either pure pitch or pure yaw is 3.43:1.
The pitch-yaw joint, which is approximately 6 in. square and 9 in. long, delivers a maximum torque of 1,650 lb-in. to the manipulator arm.
Roller functions
The bevel-shaped rollers, which act as traction-drive components, do two things that are essential to smooth operation and proper control:
First, they remove gear backlash (clearance between mating teeth) from the system. Contacting roller pairs are compressively loaded against each other by springs to obtain smooth, backlashfree motion. In each input stage, a spline locks the bevel roller with its corresponding input bevel gear, letting the rollers move axially in response to spring force.
At startup, when the gear teeth are not fully engaged (because of clearances between teeth), the spring-loaded rollers transmit the torque. Though each roller moves at the same theoretical speed as its corresponding gear, rollers experience a small loss of motion, known as creep, when transmitting torque. This allows the gears to “catch up” and begin transmitting torque soon after the initial motion — usually within a fraction of a revolution.
Continue on Page 2
Second, the rollers attenuate gear cogging or ripple, an uneven output motion that occurs as mating gear teeth go into and out of mesh.
Other features of the roller-gear design include adjustable spacers on each bevel-gear pair to minimize backlash, and high-capacity needle bearings (thrust and radial) to counteract input roller forces.
All gears are 20-deg pressure-angle Zerol bevel gears made from case-carburized AISI 9310 steel. Engineers chose Zerol gears because they provide smoother, quieter operation than straight bevel gears. And, they exhibit less sliding and higher efficiency than spiral bevel gears.
The rollers are also made from casecarburized AISI 9310 steel with the contact surfaces slightly crowned to prevent concentrated edge loading due to misalignment.
Design tradeoffs
The roller-gear arrangement represents a compromise between an all-roller traction-drive design, which transmits torque through frictional contact, and a conventional all-gear design. An all-roller design produces smoother motion that is free of gearinduced ripple. But it requires at least 2.5 to 3 times higher forces between the contacting rollers to transmit the same torque as a rollergear design. This, in turn, requires a camactuated adjustableload system to get practical efficiency, plus larger bearings and heavier housings to minimize system deflections.
Because the roller-gear design requires less force between rollers, it reduces the bearing loads. Also, the constant-load spring mechanism used in the roller-gear assembly is simpler than the adjustable-preload mechanism used in all-roller drives.
Hardware tests
Engineers from NASTEC and NASA tested the pitch-yaw joint for backlash and efficiency, using a PC to control motion and obtain feedback data. The PC was linked, via interface cards, to various measuring devices including tachometers, torque meters, resolvers, and a load cell.
Backlash tests were conducted by locking the input pinions and measuring the manipulator arm pitch at various loads. The amount of backlash was found to be zero.
Joint efficiency (output vs. input power), was determined by measuring input torque and speed while the manipulator arm lifted weights through a 30-deg angle. As expected, efficiency increased with increasing load, Figure 3. At 50% of rated full load, efficiency ranged from 88 to 97% (average 92.5%). At 73% of full load, efficiency was 97.5%.
Advanced control closes the loop
The control system for the pitch-yaw joint must ensure smooth, precise operation of the mechanical system despite external disturbances — gravity effects, forces on the end-effector (gripper at the end of the manipulator arm), and collisions with objects — as well as internal dynamic factors — friction, stiffness, sensor noise, and time delay. In particular, friction, stiffness, and backlash in the mechanical components often degrade controller performance.
Typically, either of two methods is used to control a manipulator arm: position control or force control (contact force). However, position control is not suitable where the path is uncertain and force control is often plagued by instabilities. An alternate approach controls the relationship between position and force using a programmable compliance (characteristic that provides a soft or gentle contact between the end effector and external objects). In effect, the controller smooths the manipulator arm’s motion to reduce high contact forces (avoid collisions) by simulating a spring and damper system.
In this approach, however, friction reduces the sensitivity of the system to small end effector forces and backlash causes instability. Robots can be designed with lower friction and directdrive (low-backlash) actuators, but this requires larger motors, which reduces payload capacity.
Continue on Page 3To circumvent these shortcomings, two things were done. First, gear backlash was removed from the robotic joint by adding roller drive components as described earlier.
Second, a sophisticated force-control system, called natural admittance control (NAC), was used to control the mechanical system. This technology is too complex to describe in detail here. However, such a controller uses torque-feedback information from the mechanical system to help the manipulator arm operate as though it were driven by a frictionless transmission, thus producing smooth, precise, and stable motion of the arm. Using NAC, researchers at the Center for Automation and Intelligent Systems Research, Case Western Reserve University, Cleveland, developed control system algorithms for the pitch-yaw joint.
Optimum feedback control
The control system uses torque-feedback information to help it compensate for friction or backlash in the mechanical components. Torque transducer location is important in determining the effectiveness of this control function.
Experimenters have shown that closing torque or force loops around transmissions, not around whole manipulators, produces more effective force control. Therefore, the engineers placed the torque transducers between the motor- gearbox assembly and the pitch-yaw joint, Figure 4. In this configuration, the pitch-yaw joint and manipulator arm mass act as a filter for environmental disturbances, thereby reducing the stiffness of contact with the object being manipulated and preventing high contact forces or collisions.
Because the torque transducers are between the gearbox and the pitch-yaw joint, the controller can compensate for friction or backlash in the motor or gearbox, but not in the pitch-yaw joint or the arm. This arrangement provides the most effective compensation because most of the friction (and backlash) resides in the gearbox. Further, because the pitch-yaw joint is outside the control loop, its lowfriction, low-backlash characteristics are crucial to good system performance.
Performance with feedback control
Engineers tested the pitch-yaw joint to determine how well the feedback control system compensates for friction and how well it maintains stability under harsh conditions — stiff contact with an external object.
Force measurements were taken as the joint performed a slow, constant-velocity pitch motion. With the controller off, a force of 15 lb at the end effector was required to overcome friction and begin motion. Turning the controller on reduced the required force to 0.6 lb. This reduced effort at the end effector indicates that friction (in the gearbox) was attenuated by more than an order of magnitude. And, it illustrates the advantage of placing most of the mechanical friction in a manipulator between the actuator (motor), and the torque sensor.
To test the controller stability when interacting with a stiff environment, the controller instructed the arm to collide with a rigidly mounted angle bracket at moderate speed. The resultant contact force, measured by a load cell on the end of the arm, was damped to a stable value within a few cycles.
In summary, when the feedback-control program was linked with the mechanical drive, the system exhibited good force control and maintained stable interaction with a stiff environment. Further, the controlled joint compensated for a large part of the gearbox friction.
Future in space?
Currently, NASA engineers plan to refine the performance of the pitch-yaw joint somewhat more. But, then what? Its future in space depends mainly on the outcome of NASA’s proposal for Space Station Freedom, which President Clinton has ordered scaled down. NASA recently responded by submitting three lower-cost proposals which are being considered by the government.
Information for this article was provided by William J. Anderson, NASTEC Inc.; Douglas A. Rohn, NASA Lewis Research Center; and Professor Wyatt S. Newman, Case Western Reserve University.