Technical Report

Design of Meso-Scale Robotic Systems with Miniature Actuators Kemal Berk Yesin

UMN-AML-TR-00-02

Advanced Microsystems Laboratory Department of Mechanical Engineering University of Minnesota 111 Church St. SE Minneapolis, MN 55455

April 2000



2000 University of Minnesota

This report is based upon work supported by the Defense Advanced Research Projects Agency, Electronics Technology Office (Distributed Robotics Program), ARPA Order No. G155, Program Code No. 8H20, Issued by DARPA/CMD under Contract #MDA972-98-C-0008.

Abstract In this report, design of meso-scale robotic systems using miniature actuators is investigated. Specifically, design of an active video module for a meso-scale mobile reconnaissance robot is discussed. The small size of the robot presents strict requirements on size and power consumption of the module. Available technologies for video sensors, wireless transmitters and miniature actuators are discussed with an emphasis on their applicability to meso-scale mobile systems having limited volume and working on battery power. Alternative mechanical designs for the pan-tilt mechanism of the module are presented. Computer vision techniques are implemented to perform visual tracking of targets and dynamic characteristics of the system are experimentally evaluated. Finally, a simple motion detection and tracking algorithm that was developed to track people moving inside a room is presented.

iii

iv

Table of Contents 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. Distributed Robotics Using Reconfigurable Robots . . . . . . . . . . 2.1 The Ranger . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 The Scout . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 Video Reconnaissance Module . . . . . . . . . . . . . . . . 3. Video Sensor and Wireless Transmission . . . . . . . . . . . . . . . 3.1 Video Sensor . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.1 CCD sensors . . . . . . . . . . . . . . . . . . . . . 3.1.2 CMOS sensors . . . . . . . . . . . . . . . . . . . . 3.2 Wireless transmitter . . . . . . . . . . . . . . . . . . . . . . 4. Actuators for Miniature Systems . . . . . . . . . . . . . . . . . . . . 4.1 Piezoelectric Actuators . . . . . . . . . . . . . . . . . . . . 4.2 Shape Memory Alloy (SMA) Actuators . . . . . . . . . . . 4.3 Electro-Mechanical Actuators . . . . . . . . . . . . . . . . 5. DESIGN WITH MINIATURE ACTUATORS . . . . . . . . . . . . . 5.1 Specifications for the Video Reconnaissance Module . . . . 5.2 Alternative Designs . . . . . . . . . . . . . . . . . . . . . . 5.3 First Generation Video Reconnaissance Module (VRM-1) . 5.3.1 Mechanical construction . . . . . . . . . . . . . . . 5.3.2 Driver electronics . . . . . . . . . . . . . . . . . . 5.3.3 Test results . . . . . . . . . . . . . . . . . . . . . . 5.4 Second Generation Video Reconnaissance Module (VRM-2) 5.4.1 Mechanical design . . . . . . . . . . . . . . . . . . 6. Active Vision with the Video Module . . . . . . . . . . . . . . . . . 6.1 Experimental set-up . . . . . . . . . . . . . . . . . . . . . . 6.2 Motion Detection and Tracking . . . . . . . . . . . . . . . . 7. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8. References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

v

. . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . .

. 1 . 3 . .3 . .4 . .5 . 7 . .7 . .7 . .9 . 10 . 12 . 12 . 15 . 17 . 22 . 22 . 22 . 24 . 24 . 27 . 31 . 32 . 32 . 36 . 36 . 42 . 47 . 49

vi

List of Figures Figure 1: Micro-aerial vehicle (a) and miniature motor (b) . . . . . . . . . . . . . . . . . . . 1 Figure 2: Distributed robotic system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Figure 3: The Ranger . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Figure 4: The Scout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Figure 5: Scout jumping over obstacle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Figure 6: Wireless Image Transmission . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Figure 7: CCD camera . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 Figure 8: Video transmitter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 Figure 9: Dipole orientation in PZT ceramic . . . . . . . . . . . . . . . . . . . . . . . . . . 13 Figure 10: Stacked piezoactuator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 Figure 11: Ultrasonic Motor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 Figure 12: Shape Memory Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 Figure 13: Electric Motor Types. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 Figure 14: DC motor operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 Figure 15: Permanent magnet motor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 Figure 16: Brushless Motor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 Figure 17: DC motor torque-speed curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 Figure 18: Scout payload volume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 Figure 19: Video camera. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 Figure 20: Static camera position and tilt control by spring arm . . . . . . . . . . . . . . . . 23 vii

Figure 21: Smoovy motor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 Figure 22: Motor-gearbox assembly. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 Figure 23: VRM-1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 Figure 24: VRM-1 in up, tilted and panned configuration . . . . . . . . . . . . . . . . . . . 29 Figure 25: Smoovy motor speed-torque characteristic . . . . . . . . . . . . . . . . . . . . . 30 Figure 26: Square wave 3 phase driver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 Figure 27: Pulse Width Modulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 Figure 28: System diagram for scout-VRM interface . . . . . . . . . . . . . . . . . . . . . . 31 Figure 29: Dual action mechanism operation . . . . . . . . . . . . . . . . . . . . . . . . . . 33 Figure 30: VRM-2 cross-section. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 Figure 31: VRM-2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 Figure 32: VRM-2 up, tilted and panned . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 Figure 33: Experimental setup. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 Figure 34: SSD error surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 Figure 35: Pinhole camera model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 Figure 36: Controller system diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 Figure 37: Step response of VRM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 Figure 38: Visual targets. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 Figure 39: System response to sinusoidal inputs . . . . . . . . . . . . . . . . . . . . . . . . 42 Figure 40: Frequency response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 Figure 41: Motion detection algorithm flow chart. . . . . . . . . . . . . . . . . . . . . . . . 45 Figure 42: Motion tracking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

viii

List of Tables Camera Specifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 Miniature Video Transmitters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 NiTi Alloy Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 Comparative Summary of Actuators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 Miniature Motors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 Smoovy Motor Specifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

ix

x

Figure Credits Figure 1a is courtest of Aeronvironment. Figure 10 is courtesy of Physic Instrumente. Figures 1b, 21, 22, 25 are courtesy of RMB Miniature Bearings.

xi

xii

Design of Meso-Scale Robotic Systems with Miniature Actuators

1

1. Introduction From giant powerplant turbines to silicon micromotors, a broad range of actuators have been realized for an even broader range of applications. A new class of miniature actuators with millimeter to centimeter dimensions has recently emerged. Improvements in these actuators and their increased availability has generated a variety of applications. Many diverse fields will benefit from these miniature actuators such as portable devices, medical systems and robotics. For example, a centimeter size gear-pump has been developed by MEMStek for medical applications. This device uses a miniature electric motor for actuation [3]. Micro-aerial vehicles built by Aeronvironment are remote controlled reconnaissance airplanes about six inches in length that use miniature electric motors for control surface actuation [5]. Figure 1 shows a micro-aerial vehicle and a miniature electric motor.

(a) (b) Figure 1: Micro-aerial vehicle (a) and miniature motor (b)

This report investigates a new class of miniature sized electromagnetic actuators for meso-scale electro-mechanical systems. The actuators form a key component of a miniature robotic system. Several of these miniature robots combine with medium sized carrier robots to create a novel distributed robotic system. The distributed robotic system is itself a joint research project between the University of Minnesota, MTS Systems Corp., and Honeywell Inc. The main objective of the system is reconnaissance with a group of remotely operated mobile robots that are linked to one other and to an operator via a wireless communication network. The system is primarily developed to operate in an urban environment. A number of medium sized carrier robots are deployed to the area of reconnaissance, each carrying up to twelve miniature mobile robots. These miniature robots are deployed by the carriers through a special launching mechanism

2

Tech Report UMN-AML-00-02

and each contain sensory devices such as audio, video and vibration sensors. This way a distributed sensory network is established. This report specifically discusses the design and development of a video reconnaissance module, a sensory component carried by the miniature robots. The module consists of a video camera with wireless image transmission and pan-tilt capabilities. Miniature electromagnetic actuators are used on the pan-tilt mechanism. Chapter 2 introduces the distributed robotics project, briefly summarizes its components and introduces the video reconnaissance module. Chapter 3 describes the video camera and wireless video transmitter components of the module and summarizes available technologies. Chapter 4 discusses three common forms of actuation that are suitable for miniaturization, emphasizing their applicability to mobile robots of this scale. Chapter 5 describes alternative mechanical designs for the video reconnaissance module, discusses design issues with miniature actuators and presents test results. Finally, Chapter 6 integrates the pan-tilt device with real-time computer vision algorithms to form a closed-loop visual servoing system for autonomously tracking observed motion.

Design of Meso-Scale Robotic Systems with Miniature Actuators

3

2. Distributed Robotics Using Reconfigurable Robots A joint project between the University of Minnesota, MTS Systems Corporation and Honeywell Inc., called “Distributed Robotics Using Reconfigurable Robots”, is developing a novel system of mobile robots for distributed reconnaissance [10]. The main objective of this project is to create a fleet of medium and miniature sized mobile robots for distributed reconnaissance. The system is primarily intended for use in an urban environment. The medium sized robots, called “Rangers” carry and deploy the miniature robots called “Scouts”. All robots have wireless communication capability. Due to their small size the scouts have limited communications range, and exchange information with human operators via the rangers. Figure 2 illustrates this overall concept.

OPERATOR Ranger scout

scout

scout scout

scout scout

scout

Ranger Figure 2: Distributed robotic system

2.1

The Ranger

Rangers are medium sized (65x62x36 cm) all-terrain mobile robots with wireless ethernet and video links. Their main purpose is to carry and deploy the scouts and to provide a long range communication link between the scouts and the operator. Each ranger can carry twelve scouts in a special

4

Tech Report UMN-AML-00-02

magazine and deploy them through a launcher. The launcher uses a spring coil mechanism with a motorized tilt. Figure 3 illustrates a ranger with the launcher on top.

Figure 3: The Ranger

2.2

The Scout

Scouts are cylindrical in shape with an outer diameter of 40 mm and length of 110 mm. The wheels at both ends are turned independently by small dc motors for mobility. A small spring arm protrudes from the side and can be wrapped around the cylindrical shell by a separate winch mechanism controlled by a third motor. Upon quick release of the spring arm a hopping action is achieved. This mode of locomotion is intended mainly to overcome large obstacles and climb stairs. Figure 4 shows a scout. Figure 5 shows the scout hopping over an obstacle. The scout is able to jump over obstacles 25 cm high. The scout has two Scenix microprocessors (8-bit, 50MHz), one for data processing and the other for wireless communications protocol, driver electronics for its motors, a magnetometer and a tiltmeter for direction and roll sensing. Nine 3V lithium batteries are used to supply power. The quiescent power consumption is 1.7 W, however it increases to 3.4 W while winching in the spring arm.

Design of Meso-Scale Robotic Systems with Miniature Actuators

5

Figure 4: The Scout

Figure 5: Scout jumping over obstacle

2.3

Video Reconnaissance Module

The basic scout robot described above can be reconfigured with various types of modular sensors depending on its mission such as vibration, chemical, audio and video sensors. The most important

6

Tech Report UMN-AML-00-02

reconnaissance mode envisioned for the scouts is visual. Therefore, the key sensing component developed for the scout is a video reconnaissance module. The basic purpose of the video reconnaissance module is to provide visual feedback from the scout to the operator. This definition identifies two main components of the system, a vision sensor for acquiring images and means for their wireless transmission. Another requirement is dictated by the dynamic nature of the system and its operation. Unlike the common surveillance cameras on buildings which monitor a fixed location, the scout’s vision sensor must be able to orient itself to explore its surroundings. A mechanism using miniature actuators are used for this purpose. Remaining chapters of this report will investigate the components and design of the video reconnaissance module.

Design of Meso-Scale Robotic Systems with Miniature Actuators

7

3. Video Sensor and Wireless Transmission Two basic components of the video reconnaissance module are the video sensor and wireless transmitter. A lens system focuses light rays from the viewed scene onto the sensor. A video signal representing the image is produced by the sensor and is sent to the wireless transmitter. The transmitter emits an RF signal that represents the video signal. A receiver some distance away receives the RF signal and extracts the video signal. This video signal is used to generate the image on a display device. Figure 6 illustrates the process.

Video

RF Signal

Signal Lens System

Video Sensor

Wireless Transmitter

Video Signal Receiver Display Device

Figure 6: Wireless Image Transmission

3.1

Video Sensor

Early image sensors were vacuum tube type devices with significant weight and size. By the introduction of CCD technology about twenty-five years ago there has been a sharp decrease in their cost, size and power consumption. Recently, CMOS technology has been used to fabricate image sensors resulting in further reduction in all three of these parameters. These two dominating technologies are introduced and compared below. 3.1.1 CCD sensors CCD stands for Charge-Coupled-Device. The front face of the sensor is a rectangular grid array of small photodiode elements that are used to convert incident light (photons) to electrical charge (electrons). Integrated for a small duration, these charges represent the light intensity on that photodiode or pixel (picture-element). The individual charges on each photodiode are then transferred in a clocked fashion towards an amplifier stage. This way the light intensity pattern (the image) on the sensor plate is “scanned” row by row. When all rows are scanned a frame has been generated and the scanning of the next frame starts. Under the NTSC standard for video, 30 frames are generated each second. The resulting video signal also contains synchronization signals to label the beginnings of rows and frames.

8

Tech Report UMN-AML-00-02

Image sensors are manufactured in standard sizes and named by the approximate diagonal size of the sensor plate. A typical 1/3 inch CCD sensor will have a sensor plate of about 4.82mm x 3.64 mm in size and contain 492 rows and 512 columns of pixels [7]. The total number of pixels and the sensor size determine the resolution of the sensor, therefore the larger the number of pixels per unit area the higher the resolution. As described above, the CCD sensor requires other functions to operate, such as clocks, timers and amplifier stages. It is feasible but not economical to integrate these other functions into a single circuit using the CCD manufacturing technology [4]. Therefore, CCD based video cameras require external circuitry with the image sensor. Nevertheless, small single board CCD cameras with simple lens systems are widely available on the market. Figure 7 shows such a camera. The board size of the camera shown is 32x32 mm.

Figure 7: CCD camera

The working principle for a CCD camera is similar for both grayscale and color sensors. Advanced color systems have optical filters to separate red, green and blue components of the image and then use three separate CCD sensors to generate a color signal. Simple single sensor systems microfabricate the filters onto individual pixels of the sensor. Single board CCD cameras typically require a single 5 V DC power supply to operate and consume 100 to 150 mA current. An important feature found on many of the CCD cameras is automatic exposure control circuit. This circuit adjusts the integration time of the pixels (the duration while the photons hit the pixels and charges are collected before they are sampled and flushed) and eliminates the need for external mechanical shutter components. In other words, the camera electronically adjusts to ambient lighting conditions and no mechanical aperture in the lens system is needed.

Design of Meso-Scale Robotic Systems with Miniature Actuators

3.1.2

9

CMOS sensors

CMOS (Complementary Metal-Oxide Semiconductor) technology is widely used for semiconductor device manufacturing. It has recently been applied to image sensors resulting in considerable improvements in size, cost and power consumption. Both CMOS and CCD sensors use the same photoconversion process to convert incident photons to electrical charge. For the CMOS sensor, however, these charges are not transferred from the pixels but amplified at the point of collection by dedicated CMOS transistors [4]. Thus, every pixel has its own amplification circuitry. The two main advantages of CMOS sensors over CCDs result from the wide application of CMOS technology in industry. In addition to reducing the manufacturing cost, the use of well developed CMOS technology facilitates the integration of all camera functions into a single VLSI package. Therefore, several single-chip video cameras are available on the market at very low costs. Additionally, they typically consume three to five times less power than similar CCD based cameras. CCD sensors are known to have better image quality compared to CMOS sensors. One major problem with the CMOS sensor was the fixed pattern noise resulting from the unmatched transistor characteristics at each pixel [4]. However, this problem has been greatly reduced. Today, CMOS sensor manufacturers claim to have reached the image quality of common CCD sensors. An analog video signal is not the only output option on single-chip cameras. Several manufacturers make sensors with digital output. These are especially useful with PC based systems since the necessary digitization is already done in the camera. However, transmission of digital images requires a high bandwidth communication link. The current wireless link on the scout is limited to 2.4 kbps which is too low for the real-time video requirements dictated by the distributed robotics project. As previously explained, CMOS sensor based cameras have advantages over the CCDs in term of size, power consumption and cost. Therefore, a CMOS sensor based camera system was selected for the video reconnaissance module. The monochrome sensor selected is a BV5016 by OmniVision. A pinhole lens with a 5.7 mm focal length is used to focus the image onto the video sensor. The resulting sensor-lens package is approximately 15x15x16 mm in size and weighs 3.5 grams. Table 1 summarizes the specification

10

Tech Report UMN-AML-00-02 Table 1: Camera Specifications

3.2

Sensor type

Single-chip monochrome CMOS sensor with 320x240 pixels

Size

15 x 15 x 16 mm

Power consumption

20 mA at 6-9 VDC

Output

Composite video signal, 2 V p-p at 30 frame/second

Lens

Pinhole lens 5.7 mm focal length

Wireless transmitter

A number of wireless video transmitters are available on the market with various sizes and power consumption values. Table 2 summarizes the specifications of some commercially available miniature transmitters. Table 2: Miniature Video Transmitters Model/ Manufacturer

Size [mm]

Frequency

Range Specification

Power

Vid-Link 100 Virtual Spy

28 x 20 x 10

434 MHz

150 m

35 mA, 9 V

MV915 Micro Video Products

24 x 17 x 8

915 MHz

300 m

30 mA, 9 V

MP-1 Microplate Supercircuits

50 x 32 x 4

434 MHz

200 m

100 mA, 12 V

T900V Applied-wireless

40 x 20 x 7

900 MHz

150 m

20 mA, 3 V

An important design criterion for the transmitter is the operating frequency. Commercial transmitters of small size operate in three different bands reserved by the FCC for unlicensed operation, namely the 440 MHz ATV (Amateur TV) band, the 900 MHz ISM (Industrial , Scientific, Medical) band, and the 2.4 GHz band. Generally, the higher the frequency the longer the transmission range for a given transmitting power. On the other hand, higher frequency waves concentrate more of their power in one direction. Therefore high frequency transmitters perform better in line-of-sight applications (i.e. when the transmitter antenna and the receiver antenna are in line-of-sight) but are more sensitive to alignment of the transmitter and receiver antennas. Most transmitters use Frequency Modulation (FM) while some use Amplitude Modulation (AM). In general, FM systems have less noise and signal degradation compared to AM systems.

Design of Meso-Scale Robotic Systems with Miniature Actuators

11

A common reason for signal degradation of wireless transmission occurs when the transmitter signal is reflected from surrounding surfaces and reaches the receiver along different paths, resulting in multiple signals that are slightly out of phase with one another. Some advanced transmitters use phase-locked-loop (PLL) circuitry to reject the out of phase signals, however, this additional circuitry increases the size and power consumption of the transmitter. A 900 Mhz transmitter from Micro Video Products was chosen for the video reconnaissance module. It is 24x17x8 mm in size and consumes 30 mA at 9V. Its range was tested to be 150-200 ft. line-of-sight, indoors. However, the structure of the building can effect this range. Figure 8 shows the transmitter.

Figure 8: Video transmitter

12

Tech Report UMN-AML-00-02

4. Actuators for Miniature Systems With the advent of technologies like VLSI and SMT (Surface-Mount-Technology), there has been a great reduction in size of electronics which has caused a further drive to miniaturize many products. The leading edge in this trend is MEMS (Micro Electro-Mechanical Systems) technology, which is itself an offshoot of silicon manufacturing. MEMS fabrication uses photo-chemical processes similar to the VLSI industry for creating mechanical structures with millimeter to micron dimensions. Almost all successful commercial applications of MEMS are sensors, though research is performed worldwide to develop micro-actuation mechanisms. However difficulties exist in the miniaturization of actuators due to both manufacturing issues and scaling effects. As the parts get smaller, inertial (volumetric) forces tend to loose their dominance over surface effect forces such as electrostatics [25]. Despite these difficulties many attempts to develop microscale actuators have been pursued. Microactuators can be classified into two major groups [11] : those using electrostatic and electromagnetic forces (for example electric motors) and those that use a functional element. Two well known examples of the latter type are actuators using piezo-ceramic materials and those using shape memory alloys. These actuators can also be used in meso-scale systems however, their effectiveness is naturally quite different at this scale than at the microscale. Additionally, the power requirements and control of these actuators will be different and these differences become even more important in miniature robotic systems with limited on-board resources. In this chapter, the three most common types of actuators will be examined from the perspective of their utility for miniature robotic systems.

4.1

Piezoelectric Actuators

The word “piezo” is derived from the Greek word for pressure. Discovered in 1880 by Jacques and Pierre Curie, piezoelectric materials create an electric charge when under mechanical stress. A very common natural piezoelectric material is quartz. Initial experiments with quartz showed that an electric field applied across the crystal would generate mechanical strains as well. This is called the inverse piezoelectric effect and is the working principle of all piezoelectric actuators. Special ceramics that exhibit the piezoelectric effect to a greater extend compared to natural materials have been developed. Perhaps the most common of these special ceramics is Lead Zirconate Titanate (PZT). Raw PZT material is treated in a process called poling to induce piezoelectric properties [19]. In this process, PZT ceramic is heated and a strong electric field is applied (>2000 V/mm). This creates groups of dipoles with parallel orientation called Weiss domains. This alignment is roughly maintained after cooling and the ceramic is poled. After poling, when an electric field is applied across the ceramic, the Weiss domains increase their alignment proportional to the applied voltage. With the increased alignment, the ceramic expands in the direction of the applied

Design of Meso-Scale Robotic Systems with Miniature Actuators

13

field and contracts in the perpendicular axes. This creates a linear actuator. Figure 9 illustrates the poling process.

-

- --

-

+ -

-

+

+

-

+

-

+

+ + -

-

- -

+ ++ - - -

-

+

+ +- + + + + +

+

+

+

-

+ + + Before Poling

--- -

During Poling

-

+ + After Poling

+

Figure 9: Dipole orientation in PZT ceramic

The maximum electric field strength that PZT ceramics can withstand are 1 to 2 kV/mm [19]. Typical values for maximum strains achieved along the direction of applied field are around 0.1 %. One common way to reduce the applied voltage to more practical levels is to use multilayer ceramics. Thin layers of PZT material are stacked together and electrically connected in parallel. The maximum applied voltage is limited by the thickness of a single layer, however, displacements of individual layers are accumulated. This way low voltage PZT actuators are made with layers of 20 to 100 µm thickness operating around 100 V [19]. Figure 10 illustrates a stacked type linear PZT actuator.

Figure 10: Stacked piezoactuator

14

Tech Report UMN-AML-00-02

Piezoelectric actuators theoretically have infinite resolution. Practically, the resolution is degraded by control electronics, noise, mechanical inaccuracies, hysteresis effects and creep. Nevertheless, sub-nanometer resolution is achievable making them the actuator of choice in many precision applications with small displacement needs. Actuator designs for higher displacement usually employ a mechanical leverage to increase travel. Typically, a mechanism working close to a kinematic singularity is used for high magnification [16]. Simple PZT actuators (i.e. without mechanical magnification) have high stiffness and have a controlled bandwidth of several kHz. Mechanical amplification may reduce both stiffness and bandwidth. One distinct type of actuator using piezoelectric elements is the ultrasonic motor [16]. These types of motors have a rotor that rests on a stator made of piezoelectric material. The stator is excited by a voltage signal to create travelling waves and cause a rubbing movement between the stator and the rotor. Figure 11 illustrates the working principle. Typical characteristics of these motors are high torque at low speed and high holding torque due to friction between stator and rotor. They are also suitable for hazardous environments since no sparks are produced. The inherent high torque at low speed eliminates the need for a complex gear box in many cases. Rotor

Stator Travelling Wave Figure 11: Ultrasonic Motor

Typically, the ultrasonic motor is excited by two sine waves 90 ° out of phase with an amplitude more than 100 V. Although the motor can be constructed in a small size, the necessary electronics to generate the drive signals is quite complex. As examined above, despite their high resolution and high bandwidth advantages, piezoelectric actuators are not the best match for meso-scale mobile robots that work on battery power and have limited volume for electronics.

Design of Meso-Scale Robotic Systems with Miniature Actuators

4.2

15

Shape Memory Alloy (SMA) Actuators

Shape memory alloys are metallic materials with a unique ability to change their shape at their transformation temperature. When the metal is deformed and then heated above this temperature it recovers its original shape and is able to exert a large force during recovery. This property has been utilized in many different ways including actuation. Shape Memory effect was first observed in a AuCd component in 1932 [9] and in a CuZn component in 1938. However, it was not until 1962 that the effect was discovered in nickel-titanium (NiTi) alloys which are the most common type of shape memory alloy on the market today. The basis for the shape memory effect is the phase transformation that the crystal structure of the alloy exhibits when its temperature goes above or below its transformation temperature. Below the transformation temperature the alloy is in a soft martensite phase and can be deformed up to approximately 8 percent. Above the transformation temperature it changes in to a stronger austinite phase. The soft martensite phase can be deformed easily and upon transformation to the austinite phase the material recovers the undeformed shape. In fact, the transformation does not occur at a certain temperature but within a narrow range. Additionally, it shows hysteresis behavior such that the transformation from martensite to austinite does not occur at the same temperature range as the austinite to martensite transformation. Figure 12 illustrates the shape memory effect with hysteresis. The beginning and the end of martensite to austinite transformation is shown as Ms and Mf respectively. Similarly As and Af are for austinite to martensite transformation. Table 3 lists some important properties of NiTi shape memory alloys. Table 3: NiTi Alloy Properties Density [g/cm3]

6.45

Resistivity, Austinite [microohms.cm]

100

Resistivity, Martensite [microohms.cm]

70

Thermal Conductivity, Austinite [W/cm. ° C]

18

Thermal Conductivity, Martensite [W/cm. ° C]

8.5

Young’s Modulus, Austinite [GPa]

83

Young’s Modulus, Martensite [GPa]

28 to 41

Ultimate Tensile Strength [MPa]

895

Transformation Temperatures [ ° C]

-200 to110

Maximum shape memory strain [%]

8.5

16

Tech Report UMN-AML-00-02

Figure 12: Shape Memory Effect

The alloy composition effects these values. NiTi alloys have 51% nickel and 49% Titanium. Excess nickel strongly depresses the transformation temperature and increases the yield strength of the austinite. Other frequently used elements are iron and chromium (to lower the transformation temperature), and copper (to decrease the hysteresis and lower the deformation stress of the martensite) [9]. The transformation temperature of NiTi alloys can be adjusted from over 100 ° C to cyrogenic temperatures. Actuators using the shape memory effect are mostly linear. Usually, SMA material in wire form is strained by a bias force exerted by a spring or deadweight. Upon heating, which is almost always done by passing current through the wire, shape recovery occurs. Stresses of 170 MPa or more may be exerted during recovery. Initial strain is usually limited to 4% to avoid reduction in shape recovery after many cycles. Tens of millions of cycles are possible at low strain [1]. The thermal nature of shape memory effect limits the bandwidth of actuators utilizing this phenomena. Heating the alloy with current is relatively fast however the cooling phase, which is usually unforced, is slow. Actuators capable of 4 cycles per second have been reported that use SMA wire both for actuation and bias force [6]. However, faster cooling times are expected as the actuator size decreases since heat capacitance decreases faster than surface radiation (volume versus surface area) with miniaturization.

Design of Meso-Scale Robotic Systems with Miniature Actuators

17

The biggest advantage of SMA based actuators is their simplicity and thus reliability. Also the power to weight ratio is high. However, they are usually on-off type actuators without position control.

4.3

Electro-Mechanical Actuators

Electro-mechanical actuators, or electric motors, dominate the world of actuators in most fields. Electrical energy is easy to transmit and the well established network of power lines providing energy everywhere at relatively low cost is one reason for their popularity. Over the years, various types of electric motors have been developed that address the specific needs of different application areas. Figure 13 shows basic types of electric motors.

Figure 13: Electric Motor Types

All electric motors convert electrical energy to magnetic and then to mechanical. Although linear motors which produce force along a line are common, the most abundant type is the torque motor with a rotating shaft. The two major categories of electric motors are AC and DC motors. DC motors do not actually operate with direct current but they commutate the dc current supply at their terminals either electronically (as in brushless dc motors) or mechanically (brushed motors) to generate a rotating magnetic field.

18

Tech Report UMN-AML-00-02

DC motors are widely used in robotic and other servoing applications due to the ease of their control. For mobile applications running on battery power, they are practically the only choice. The most common type is the brush type dc motor. The operating principle of a brush type dc motor is illustrated in Figure 14. A constant magnetic field B is supplied either by an electromagnet (wound-field) or a permanent magnet and current I passing perpendicular to this field runs through the coil with length L . A net force F is generated on each side of the coil perpendicular to B and I defined by the Lorentz Law: F = I⋅L×B The resulting torque will rotate the coils 90 degrees around the motor axis O′ – O to the equilibrium position. In a real motor there are more than one set of coils (usually each having more than one turns) that have an angular displacement around the motor axis. Using mechanical or electronic commutation, these coils are energized in order to ensure constant rotational motion. Figure 15 illustrates a typical dc motor with a pair of permanent magnets used to supply the constant magnetic field. One set of coil windings is shown terminating at the commutator. A pair of brushes (not shown) will conduct the current through the coils in the correct order as the commutator rotates with the shaft.

I F

B

F

Figure 14: DC motor operation

From this basic definition of a dc motor one can see the produced torque (force multiplied by radius) is proportional to both the diameter and length of the motor. In fact, as the diameter increases, the circumference increases as well enabling the designer to place more coils. Therefore, a general relationship between motor size and torque capabilities can be stated as [8] : T = k⋅D ⋅L 2

Design of Meso-Scale Robotic Systems with Miniature Actuators

19

COMMUTATOR

Figure 15: Permanent magnet motor

In this equation k is a constant dependent on design parameters other than size such as magnetic field source (i.e. electro-magnet or permanent magnet), brushes and bearings. D and L are armature diameter and length respectively. It is apparent that miniaturization of the electric motor is not advantageous for its performance since the torque output decreases by the cube of the linear dimension. Still, with efficient design and new permanent magnets with high magnetic field density, dc electric motors at very small sizes are feasible. The introduction of Alnico and Ferrite magnets in early 1940’s, and the discovery of rare-earth magnets in 1960’s enabled motor manufacturers to decrease the size of pm motors while increasing power capacity and decreasing costs [15]. Although the relative scarcity of some of the rare-earth magnetic materials (particularly cobalt and samarium) puts some practical limits on their use in commercial applications, development of NdFeB (Neodymium-Iron-Boron) magnets promises a source of high energy magnetic materials with relatively plentiful sources [15]. Brush type permanent magnet electric motors can be found down to 8 mm in diameter on the market (i.e. MicroMo series 0816 motors). However, as the size (and thus torque output) decreases, the effect of frictional losses at the brushed commutator increases. Brushless motors, therefore, have the advantage of increased efficiency over brushed ones at all sizes. It is for this reason that the smallest motors (5 to 1.9 mm diameter) on the market are all brushless. Brushless dc motors can be described as an “inside out” version of typical brushed motors. The permanent magnets are on the rotor and the coils are stationary on the stator. A typical three phase brushless dc motor is illustrated in Figure 16. Three pairs of windings are positioned at 120 degree

20

Tech Report UMN-AML-00-02

intervals around the rotor. When they are excited in the correct order, a rotating magnetic field is generated and the permanent magnet rotor rotates to align itself with this field.

M rotor

Stator windings Figure 16: Brushless Motor

For correct excitation of the coils, the rotor position is fed back to the driver electronics. Usually, an optical or hall-effect based encoder that is attached to the back of the motor shaft is used. Another method is based on back emf generated by the rotating magnet. The coils are energized such that only two coils are active at a time, leaving the third open for sensing of the back emf generated. With all methods, current or voltage control is possible in the same way as a brush type motor. An added advantage of brushless operation is the avoidance of spark generation at the commutator. Other than decreasing the operating life of the motor, these sparks generate high electro-magnetic interference (EMI) and can also be dangerous in certain hazardous environments. The efficiency of an electric motor is the ratio of mechanical power output to electrical power input. At steady state this relationship can be stated as T⋅ω ε m = ----------- [%] V⋅I Where T [N-m] is output torque, ω [rad/sec] is shaft speed, V [Volts] is the dc voltage across the terminals and I [Amps] is current through the motor. Small sized commercial motors typically have 60-80 % maximum efficiency. Most small dc motors produce power at high speeds and low torque. A gear box is often used to produce more torque at lower speeds at the expense of decreased overall efficiency due to additional frictional losses. However, high quality gearboxes can operate around 90% efficiency [15].

Design of Meso-Scale Robotic Systems with Miniature Actuators

21

As mentioned before, dc motors are relatively easy to control compared to other types. There is a linear relationship between motor torque (load) and speed at a fixed voltage. As the load increases the speed decreases linearly. Similarly, voltage is related linearly to speed at a constant load. Usually, motor characteristics are shown in torque-speed plots at the nominal operating voltage for which the motor is designed. Figure 17 illustrates such a plot with power output (torque x speed) and efficiency also plotted. As seen from the graph, maximum efficiency of energy conversion occurs at about 10% of the stall (zero speed) torque, which is typical for most motors. However, the maximum power output occurs midway through the speed-torque graph.

Figure 17: DC motor torque-speed curve

22

Tech Report UMN-AML-00-02

5. DESIGN WITH MINIATURE ACTUATORS As an application of miniature actuators, design of the video reconnaissance module is presented in this chapter. The required properties of the system are defined and alternative designs are developed. Issues in mechanical design with miniature actuators and their control electronics are examined.

5.1

Specifications for the Video Reconnaissance Module

As explained in Section 2.3 the video reconnaissance module will transmit live video through a wireless link back to the operator. It is required that the video camera provide a complete view of its surroundings. Additionally, the video reconnaissance module can not protrude from the tubular shell since this would interfere with the launch and mobility of the robot. There is very limited volume available on the robot for sensors and other payload. The shape and dimensions of the payload volume is illustrated in Figure 18. The specifications of the video camera are summarized in Table 1 in Section 3.1. Figure 19 illustrates its shape and size.

Figure 18: Scout payload volume

5.2

Alternative Designs

A trivial design for the video reconnaissance module with minimum complexity is a static camera fixed inside the robot. The camera sees through the transparent shell of the robot and depends on the robots mobility for visually tracking the surrounding environment. A static camera has been placed on some of the first series of prototype scouts in order to test this concept. The robot can adjust the pan and tilt angle of the camera by rotating itself on the ground (through independent

Design of Meso-Scale Robotic Systems with Miniature Actuators

0.393 ”

23

0.59 ”

0.236 ”

0.59 ”

Figure 19: Video camera

tilt Winch wire

camera transparent shell

Spring arm

Figure 20: Static camera position and tilt control by spring arm

control of its wheels) and rolling along its longitudinal axis by winching in and out the spring arm. Figure 20 illustrates the position of the camera on the robot and control of the tilt axis of the camera by the winch mechanism. Despite the simplicity this offers, the static camera design displayed some problems during tests. The pan axis, which is controlled by rotating the robot on the ground, is not smooth and is effected by ground conditions and obstacles. The spring arm winch mechanism that uses an electric motor coupled to a gearbox operates very slowly and is the most power consuming subsystem (~1.7 W) on the robot. An alternative design was sought that would provide pan-tilt capability independent of the robot. However, the limited payload volume prohibits the camera from being panned and tilted complete-

24

Tech Report UMN-AML-00-02

ly inside the robot. Therefore, in addition to pan and tilt action, a third component of motion to raise the camera outside the shell and retract it is needed. The first generation video reconnaissance module (VRM-1) was designed and built to meet these specifications.

5.3

First Generation Video Reconnaissance Module (VRM-1)

5.3.1 Mechanical construction With the specifications on the video camera and payload volume fixed, a mechanical system to add pan, tilt and pop-out/retract capabilities to the camera was designed. A significant design challenge is to choose an actuation technology that will enable the system to fit inside the small payload volume. Three common types of actuators were previously analyzed in more detail. Table 4 summarizes the advantages and disadvantages of these actuation technologies. Table 4: Comparative Summary of Actuators Actuator

Advantages

Disadvantages

Shape Memory Alloy

• • •

Simple and robust High actuation force Simple two-state (on/off) action

• • • •

Low strain (% 4) Low bandwidth (