New High-Angulation Omni-Directional Sensor Mount

By Mark E. Rosheim* and Gerald F. Sauter, Ph.D*
Ross-Hime Designs, Inc. Minneapolis, Minnesota

© Ross-Hime Designs, Inc.

Abstract

Presented is a new revolutionary free-space optical communication sensor mount. Featuring 70 arcseconds average repeatability, this gimbal-like pointing mechanism provides over 180 degrees azimuth and declination singularity-free pointing capability for a wide range of sensors for the entire electromagnetic spectrum. Applications include air, sea, and space as well as land based vehicles.

Keywords: Sensor, mount, gimbal, tracker, joint, universal, wrist, omni-directional.

1.0 Introduction

A key component of any practical, covert laser communications system will be the pointing and tracking mechanism. Such a system requires a mechanism that is (1) easily manufacturable therefore inexpensive; and which has (2) high slew rates; (3) high accuracy; and (4) wide, singularity-free range of acquisition. Ross-Hime Designs, Inc. (RHD) believes that it has developed a new, innovative mechanism that can meet these demanding requirements. RHD has constructed and tested a free space optical communication sensor mount based on the Omni-Wrist series of robotic wrist mechanisms. The Omni-Wrist III Sensor Mount is a revolutionary, patented, low-cost, lightweight, compact, high dexterity pointing device for space, land, and sea-based communication applications ref. [2,3,4]. It is a major advancement over conventional azimuth/elevation mounts because it offers 180 degrees of unimpeded (singularity-free), hemispherical movement; low manufacturing costs; and a streamlined parallel mechanism architecture. These features combine to produce a reliable, high-performance pointing device capable of greater precision than existing designs.

Fig. 1 Omni-Wrist III Sensor Mount

The Omni-Wrist III (Figure 1) address technology gaps in existing approaches to sensor pointing. Typical solutions have involved the use of either a two-phase system, a gimbal or a universal joint for position. Both of these types of conventional azimuth/elevation mounts suffer from slowed response times and/or limited range of motion, or from singularities (voids in the work envelope). The first type of design has a sensor mounted on a fork for elevation (pitch) that in turn is mounted on a rotating base for azimuth (yaw). Such a design requires the positioning of two independent systems, which increases response time by introducing unnecessary path optimization calculations into every pointing operation. The resultant "singularity cone" results in very poor pointing precision in this area.

Presently two forms of antenna pointing systems dominate the ground and space based markets (Figures 2 & 3). Schaffer Magnetics Inc. "Biaxial Drive" represents the most common design, also known as an azimuth and elevation or yaw-pitch type drive ref [1,5]. Schaffer's system is composed of two harmonic drives powered by DC motors with redundant windings and electronics. Wiring for electronics, motors and antenna is integrated within the structure. The simple mechanical design is very stiff and rugged.

	Fig. 2 Schaffer Biaxial Drive
	Fig. 3 Honeywell Mount

Although compact and precise, within limits, this design has a singularity (it jams) when the sensor is pointed straight-up. Yaw becomes roll and the resultant singularity degrades precision and complicates control. This is particularly evident when tracking high-speed objects or in the case of platforms that are pitching and yawing due to the dynamic media that they inhabit.

The second common antenna pointing system (Figure 3) is represented by the "APS" manufactured by the Honeywell Space Systems Group ref. [1]. Two perpendicularly mounted actuators produce + 110 degrees each. Singularity is mitigated when the antenna is pointing straight-up because of the perpendicular orientation of the two actuators. However, singularity or gimbal lock occurs when the pointing system attempts to move in circumduction (a combination of pitch and yaw motion) at the extremes of its range of motion. The Honeywell design is less stiff than the Schaffer Magnetics design because the orientation of the joints create longer lever arms that could cause deflection under high inertia loading. It is also heavier than the Omni-Wrist III due to the larger amount of metallic structure. Two harmonic drives with redundant motor windings and electronics power the system. Complex flexible cables (a major design problem) are found in both designs.

A myriad of other sensor mounts have been tried over the years on planes, missiles, ground based and shipboard systems. Gimbals, universal joints, and ball-and-socket have been tried in an effort to develop a cost-effective design that is flexible, stiff and rugged. Similarly, their modern day equivalent of rotating mirrors, raster-scan devices and dreams of phased arrays may be to futuristic due to limitations of range-of-motion, cost, and complexity [ref 6 & 7].

By way of contrast, RHD's Omni-Wrist III uses a new, innovative, unique, double universal joint that reduces response time by enabling a single calculation to drive the motion of the joint. RHD has designed and built a number of gimbal systems for the robotic industry since 1987. Typically these systems are used in applications where dexterity is required to follow a complex contour path, such as in robotics and RF antenna pointing. Thousands of RHD RF pointing systems are in use world-wide for telecommunication purposes based on early first generation technology.

1.2 Advantages and Features

The Air Force Research Labs/Ballistic Missile Defense Organization (BMDO) provided a contract to develop and build and test the Omni-Wrist III Sensor Mount. The titanium unit was subsequently tested for repeatability and accuracy at Ross-Hime Designs and Oceaneering Space Systems.

	Fig. 4 Singularity-Free Omni-Wrist III Sensor Mount in Circumduction

The technology created by Ross-Hime Designs resulted in a break-through solution to these issues. The Omni-Wrist III features a unique mechanical design that eliminates singularity or gimbal-lock (Figure 4). In witness to the unique mechanical design, during the design process, using a major PC based software analysis package, we encountered a problem with generating animation. The software did not recognize the kinematic nature of the design, and interpreted it as a static linkage incapable of motion. A visit to Minneapolis by a representative of the software company soon resulted in a new revision.

It's small parts count of only three unique pieces and use of commercial bearings and acuators addresses the issue of manufacturing cost and maintenance. Two different individuals easily calculated the kinematic solution in a short period of time. The through-the-center hole design allows all wiring and hoses to pass through the center of the Omni-Wrist III, eliminating the need for special shrouding and protection. Also, through the center routing provides the minimum bend radius for the wiring, increasing its life. More compact with lower inertia compared to other designs, it is a simple matter to shroud the device in a commercial flexible covering or "boot."

Three redundant fail-safe features were created to make the wrist rugged and reliable. First, mechanical stroke limits were added to the Exlar linear actuators. Second, magnetic Hall effect sensors triggered by magnets in the lower arms provide further over-travel protection. Finally, software limits prevent commands that could send the wrist beyond it's range of motion. In operation, a pair of high precision electric linear actuators drive the double jointed Omni-Wrist III. Encoders on the actuators provide relative position information. Optical range finder sensors mounted on the support column provide absolute position information of the lower arms.

1.3 Specifications

Payload Capability: 5 lbs.
Material: Titanium
Range of motion: 180 degrees hemisphere
Output Speed: 60 degrees/sec
Accuracy: .06 degrees = 3.6 arcminutes = 216 arcseconds
Physical Envelope: Cylindrical, 10" diameter X 18" long
Weight of Sensor Mount: 20 lbs.

These specifications were stated as goals and were based on the characteristics of the Phase I design. Our efforts resulted in a prototype that meets or exceeds each of these design goals. The actual test results are included in two parts. First, we include a coordinate measuring machine (CMM) testing data developed by our Controller partner, Oceaneering Space Systems, Houston, Texas. Second, we present the test results generated at the home office of Ross-Hime Design, Inc. In general terms, our results are as follows:

Payload capacity: All testing was done with a simulated payload with a weight of 4.6 lb. We also experimented with payloads as heavy as 10 lbs with similar results. It is noted that as the weight is increased, there is some deflection of the mechanism.

Range of Motion: The device achieves the stated goal of 180 degrees of capacity. These are magnetic limit switches located to limit the range of motion to less than 181 degrees but more than 180 degrees. The software included the provision to not accept any command that would cause the declination to exceed 90 degrees, and it will not accept a manual declination command larger that 90.00 degrees.

Output speed: The speed is currently set at approximately 60 degrees per second. There is a provision in the software code to raise or lower the output speed. The velocity can also be lowered via the GUI.

Accuracy: As detailed below, the goal of 216 arc-seconds was met and exceeded. In general terms the accuracy is around 70 arc-second, and is very repeatable.

Physical Envelope: The wrist portion of the mechanism is basically as sphere of less than 10" diameter. The height, not including the base is less that 18" in height. The major factor affecting the height is the length of the commercial linear actuators. These actuators are the smallest commercially available that meet the linear accuracy and repeatability required.

2.0 Test Results at Oceaneering Space Systems, Inc. (OSS)

	Fig. 5 Omni-Wrist III Sensor Mount Attached to CMM at OSS

A Coordinated Measuring Machine (CMM) was used to determine the accuracy of pointing for the Omni-Wrist III Sensor Mount. The CMM is capable of high precision measurements of any point on the Sensor Mount. Figure 5 shows the Sensor Mount attached to the CMM. The tip of the Sensor Mount was used for these measurements. As the Sensor Mount was exercised the position of the tip was precisely determined. These positions were then compared with expected values and the various errors were calculated. The Sensor Mount was placed within the CMM and exercised at two azimuth settings, (0 and 45 degrees). At each of these positions the declination angle was changed from 0 to 81 degrees in several steps. All told there were 58 separate positions.

2.1 Initial Test Results

Table I is a list of the angles used in these measurements. The tests involved a Commanded Az. and Dec. angle followed by movement to these angles. The actual Az. and Dec. angles were measured by the CMM. Errors were determined and plotted. The next several graphs show these results.

Figure 6 displays the difference between the commanded position and the CMM measured position for azimuth and declination for the initial data. The units for both the X and Y axis are degrees.

Fig. 6 Errors: Commanded Position - CMM Measured Position

Note that the pointing accuracy goal set for the Omni-Wrist III Sensor Mount was 216 arcseconds or less than 0.06 degrees. Several positions exceeded this goal. Following these tests the controller was reconfigured with different Proportional Integral Differential (PID) coefficients to bring the stopping point into tighter control. At the same time the Exlar actuators were remounted on the baseplate to make the movements more symmetrical.

Note: Subsequent to these measurements it was also determined that high friction/stiction caused by an undersized actuator bushing was causing the motors to heat excessively. This friction/stiction in turn caused the actuators to stop before reaching the commanded positions. The bushings were enlarged after the unit was returned to Ross-Hime Design's office.

After the Exlar actuators were repositioned the CMM tests were repeated. Figure 7 illustrates the angular error for the two axis. Now the errors on both axis were well within the 0.06 degree goal set for the Sensor Mount.

Fig. 7 Conservative Angular Error from Exlar Mounting Offset for 58 Data Points Taken at OSS

2.2 Test Procedure at University Technology Center

Fig. 8 Sensor Mount Attached to Support

After the tests were completed at Oceaneering Space Systems, Inc. (OSS) the Omni-Wrist III Sensor Mount was returned to the Ross-Hime Designs office located in the University Technology Center (UTC) for further evaluation. A support for the Sensor Mount had been previously designed and built and now was attached to an outside wall at UTC. This support provided a solid base for the remainder of the testing. Figure 8 shows the support with the Sensor Mount attached. A laser pointer was attached to the Sensor Mount at the center hub. The central hub also incorporates the means to attach various weights. We used regular bar bell weights for the testing. Weights of 4.6 and 6.1 pounds were the primary values used.

The detection stage was an adjustable assembly that is shown in Fig. 9. The stage held a paper "target" like the one shown in Fig. 10. The rings had 0.1 inch increasing radii which represents about 74.06 arcseconds/0.1" resolution when the Sensor Mount and the detector stage were separated by 278.5 inches.

Fig. 9 Detection Stage	Fig. 10 Bullseye Pattern at Detector (0.1 inch feature size)

The laser pointer had a diverging beam whose width was approximately 300 arcseconds at the 23 foot separation. During the tests the beam was projected through the paper target and either a mark was made at the center of the beam from the back side of the target or a circle was drawn around the central beam. In this way the centroid of the beam could be accurately found to within 30 - 50 arcseconds or about +0.5 squares. We found it more convenient to measure the position in "number of squares in the X and Y directions. Conversion to arcseconds was performed as a last step. The laser pointer could be attached to the holder in two separate positions. The first position allowed the "home" position to be at Declination 90 degrees and Azimuth 180 degrees. Here the tip of the Sensor Mount was horizontal at the home position. This position also produced full extension of the Exlar actuators. The second position made the "home" position at Declination 0 degrees and Azimuth 0 degrees. This configuration has the Sensor Mount straight up at the home position. The controller hardware is shown in Fig. 11. The controller system includes power supplies, processors, motor controllers and a laptop computer that displays the graphical user interface (GUI).

Fig. 11 Sensor Mount Controller

2.3 Test Procedure

The typical test procedure consists of 1) centering the beam on the bullseye pattern, 2) translating the Omni-Wrist III Sensor Mount to some "other" position, 3) returning it to the original position and 4) marking the beam center on the bullseye pattern. The "other" positions are small angles, large angles and mixed axis translations. Various weights could also be attached to the Omni-Wrist III Sensor Mount so bending moments and the effects of moving different weights through different angles could be determined. We feel confident that these tests, while simple in nature, give useful information on the "repeatability" of pointing.

2.4 Results

The results of one test is shown in Table II and a detailed description follows.

Here the home position was Dec. 0; Az. 0 (0, 0). The initial laser beam position at the detector stage was determined, (Xi = -2 squares; Yi = 0.5 squares). The Sensor Mount was then moved to (70, 0) and returned to home. The beam position was again measured (X = -3; Y = 2). The entire procedure was then repeated five times. The Sensor Mount was then moved to (70, 180) and returned to home where the laser beam position was determined. This operation was also repeated five times. In like manner the Sensor Mount was moved to (70, 90) and (70, 270). Movement to the four "cardinal" compass points were accomplished without interruption and without additional jogging or a homing sequence. Data for these movements are shown in the Table. The average change of beam position during the five repeat movements is shown in columns labeled Avg X and Avg Y. For these averages the last beam position of the preceding row becomes the initial position for the row under consideration, e.g. for the movement to (70, 90) the initial position (Xi, Yi) is (-3, -2) - the last position of the preceding movement to (70, 180).

The average change for all movements during this sequence is shown in the last columns. Finally, a conversion to arcseconds was made. Many similar tests were made. The goal was to test the Sensor Mount's repeatable pointing accuracy over the entire hemisphere of operation. The condensed results are shown in Table III. Only one test yielded a pointing error that exceeded the design goal (bold entry). When the Sensor Mount was exercised at 315 degrees some unexplained glitches occurred that forced a rehoming sequence. The fast response (164 degrees per second) was the likely culprit for these glitches. The speed was reduced to the design goal of 60 degrees per second and movement was much smoother.

Repeatability tests after multiple moves was also made. The home position was (0, 0). The Sensor Mount was first moved to Dec = 70, Az = 0 then to Dec 70, Az 180 and then back to home. This sequence was repeated ten times and the X and Y variations were measured. The average variation for the ten repeats was X = 130 and Y = -94.2 arcseconds. A second test was made where the compound move was Dec = 89, Az = 45 then to Dec = 89, Az = 225. The average variation for these ten repeats was X = -32.1 and Y = -101.2 arcseconds.

3.0 Conclusion

The Omni-Wrist III Sensor Mount was exercised through many different sequences. Two home positions, (0, 0) and (89, 180) were used. A laser pointer was used to shine a light spot on a bullseye target mounted a little over 23 feet from the Sensor Mount. At that distance the position of the laser beam could be determined within about + 35 seconds of arc. The test sequences consisted of large angle deflections, small and medium angle deflections, and multiple angle movements. A weight of 4.6 lb. was attached to the Sensor Mount for all except the first test when a weight of 6.1 lb. was used. For all the tests only one of the tests produced an average repeatability error greater than the goal of 216 seconds of arc, approximately 2.9 squares. In general the reproducibility error was less at the home position (0, 0). The actuators are retracted here and the Sensor Mount movements are more certain than at the actuator position extremes. Speed tests indicated that an angular deflection greater than 160 degrees/second could be accomplished - much faster than the goal of 60 degrees/second.

ACKNOWLEDGMENTS

The authors wish to acknowledge funding provided by Ballistic Missile Defense Organization under contract no. F30602-98-C-0014

4.0 REFERENCES

1. Honeywell, 1996. M597B/CLR "Rotary Actuator Characteristics." Specifications sheet on Honeywell's space qualified rotary actuator.

2. Rosheim, M. 200. U.S. Patent 6,105,455. "Robotics Manipulator." Filed 6 Sept. 14, 1999; issued Aug. 22, 2000.

3. Rosheim, M. 1999. U.S. Patent 5,979,264. "Robotic Manipulator." Filed Oct. 16, 1997; issued Nov. 9, 1999.

4. Rosheim, M. 1999a. U.S. Patent 5, 893, 296. "Multiple Rotatable Links Robotic Manipulator." Filed Mar. 13, 1997 and issued. Apr. 13, 1999.

5. Schaffer, E. 1991. "Application of Harmonic Drive in Space. " Presented at the International Harmonic Drive Symposium. Matsumoto, Japan. 23 May, 1991.

6. Butler, et al. 1978. Stabilized Platform System. U.S. Patent 4,068,538. Filed 12 June 1975 and issued 17 Jan. 1978.

7. Wainwright, L. 1933. Universal Joint. U.S. Patent 1,899,170. Filed 7 Dec. 1931 issue 28 Feb. 1933.

*contact anthrobot.com; phone 651 699-5610; Fax: 651 695-0320; Ross-Hime Designs, Inc. 1313 5th Street South East, Minneapolis, Minnesota 55414.