Indoor radio communications for factories of the future...

Indoor Radio Communications for Factories of the Future Theodore S. Rappaport

A HE BOOM IN FACTORY AUTOMATION HAS created a need for reliable real-time communications. In 1984, the Manufacturing Automation Protocol (MAP) networking standard was established by industrial leaders to encourage commercialization and standardization of high data rate communications hardware for use in computer-controlled manufacturing [l]. In late 1985, the Technical and Office Protocol (TOP) was developed for computer communications in office buildings [2]. Both MAP and TOP are capable of supporting 10 Mb/s data rates for short periods of time and rely on coaxial cable or fiber optic cable to interconnect users. Twisted-pair interconnection of computer terminals is also commonly used as a communications channel in modem factories.

Methods for transporting parts-in-process in a futuristic, but realistic, JIT manufacturing environment is one of the thrust research areas at the ERC for Intelligent Manu fact u ring Systems.

To address concerns about the international competitive- ness of the United States, the National Science Foundation (NSF) established large interdisciplinary Engineering Re- search Centers (ERCs) at ten universities in 1985. The charter of these ERCs is to conduct pioneering research aimed at vastly improving some of the manufacturing technologies and meth- odologies currently used in U.S. industry [3] [4]. Work at the ERC for Intelligent Manufacturing Systems at Purdue Univer- sity revealed that Just-In-Time (JIT) manufacturing techniques can result in cost-savings of several orders of magnitude when applied to small and medium batch manufacturing pro- cesses. Industries that produce small-quantities of light-weight ( < 25 kg) metal parts, medium-sized ( < 1,000 units) production runs of electronic parts, and durable goods are particularly well-suited to JIT techniques [5].

0 163-6804/89/0OO5-OO 15 $0 1 .OO ' 1989 IEEE

Methods for transporting parts-in-process in a futuristic, but realistic, JIT manufacturing environment is one of the thrust research areas at the ERC for Intelligent Manufacturing Systems. Analysis has shown that an inexpensive, agile, mobile robot fleet, capable of navigating without any type of track, could easily accommodate a majority of the material flow required for a JIT manufacturing system. Navigational techniques and corresponding navigational error analyses have been conducted for a variety of autonomous guided, mobile robot systems [6-81.

A truly Autonomous Guided Vehicle (AGV) that does not use a tether requires a radio system for control. Optical systems are viable, but become inoperable when obstructed. Further- more, radio systems are useful for providing fast and inexpensive connections for often-moved manufacturing equipment and computer terminals. Radio will also accommodate reconfigurable voice/data communications for other facets of factory operation. For example, RF tags, which can be easily attached to parts-in-process, provide a paperless data storage medium and are becoming an increasingly popular replace- ment for paper work order cards [9]. Eventually, radio communications may be used in homes and offices to provide universal, digital, portable communications [lo]. Leading telecom- munications firms, such as Bell Communications Research, AT&T Bell Laboratories, and Motorola are now exploring the viability of indoor radio communication systems for homes and offices [IO- 141.

Narrow-band (RF bandwidths < 25 kHz) VHF digital radio systems are presently marketed by a number of compa- nies and are beginning to find use in data communications for inventory management (bar code readers, shelving dispatch) and dedicated control (for overhead cranes, wireguided vehicles, etc.). Most of these systems employ binary Frequency Shift Keying (FSK), a form of FM modulation. By retrofitting commercially available FM transceivers, many factory radio system vendors provide digital radios by using Audio Frequen- cy Shift Keying (AFSK). In AFSK transmission, a data bit is mapped into a particular baseband tone having a duration equal to the reciprocal of the bit rate. There are two tones for each of the two possible bit values. The resulting baseband analog audio signal is then applied to the input of a standard FM transmitter. At the receiver, bandpass filters or tone decoders, which follow the FM demodulator, are sampled to determine which audio frequency (i.e., bit) was sent. Although AFSK is not as efficient as other forms of FSK, it is very simple to install

May 1989 - IEEE Communications Magazine 15

on readily available analog FM transceivers since only audio circuitry must be added. Some factory radio system vendors use true FSK or Minimum Shift Keying (MSK), which is the most spectrally efficient form of binary FSK [15, p. 3261.

Voice radio systems, which are presently used for paging and personnel dispatching inside buildings, employ FM at VHF and UHF frequencies. In many applications, analog FM transceivers are used in factory environments that contain both electromagnetic and acoustic noise. In the metal-working and textile industries particularly, operators in noisy environments use headphones that contain a microphone element. The ear-mounted microphone detects aural vibrations that pass through the eardrum and jawbone, while background noise is shielded from the operator and the microphone by the headset. At the receiving base station, signal processing is per- formed on the received speech to filter baseband background noise and to undo distortion induced by the throattear channel.

With modem speech coding algorithms, it is possible to greatly enhance voice communications in noisy, multipath radio channels. Several vendors have recently demonstrated powerful speech coding algorithms, which code speech at rates of between 6.5 kbls and 13 kbls, for next-generation U.S. digital cellular telephones. The Jet Propulsion Laboratory, in con- junction with the University of California and the Georgia In- stitute of Technology, have shown the viability of 4.8 kbls speech codecs for satellite voice channels. Advancements in speech coding are discussed in the February 1988 issue of the IEEE Journal on Selected Areas in Communications. Informa- tion on research in mobile satellite systems is available in the MSAT-X Quarterly, which is published by the Jet Propulsion Laboratory, Pasadena, California.

It is important to note that voice systems can tolerate much greater bit error rates than data systems. Where an uncoded data transmission system may require bit error rates on the order of speech communications can tolerate bit error rates on the order of This is due to inherent redundan- cies in human speech and the brain’s ability to interpret lan- guage.

While existing indoor factory radio systems are well suited for human operation and simple digital communication tasks, it is anticipated that for a moderately sized AGV fleet (> 20 vehicles) employing multiple-access digital radio communications within a Computer Integrated Manufacturing (CIM) environment, data rates of several hundred kilobits per second will be needed to accommodate real-time computer control and navigation of the vehicles.

In the U.S., the Federal Communications Commission (FCC) has allocated spectrum for narrow-band industrial radio communications in the VHF (450 MHz) and UHF (900 MHz) bands. More recently, the FCC authorized the use of suitably designed spread-spectrum systems for 900 MHz, 2,400 MHz, and 5,725 MHz [ 161. Provided that transmitters meet FCC ap- proval, unlicensed 1 W transmitter power levels may be used ovtr Sandwidths greater than 25 MHz. In Japan, spectrum has been sei aside for 300 mW, 4,800 bls indoor radiQ systems operating in the 400 MHz and 2,450 MHz bands [ 171. Federal regulatory agencies have recently recognized the potential of wideband radio Local Area Networks (LANs), and radio propagation inside factories and office buildings is becoming increasingly understood. However, the current method for in- stalling factory radio links requires lengthy surveys and often involves redundant equipment. Robust signalling techniques and system designs for wideband indoor radio systems do not currently exist, although this is an active area of research [i.e.,

Accurate characterization of the operating channels is a mandatory prerequisite for the development of reliable indoor radio systems. Radio channel propagation data from factory

18-20].

buildings have recently been made available through research at the ERC for Intelligent Manufacturing Systems, and a statistical impulse response channel model, based upon the details being developed at Virginia Tech. As is shown in this article, it is not environmental noise, but rather multipath propagation that limits the capacity of radio links operating above 1 GHz. The severity of multipath largely depends on factory inventory, building structure, and surrounding topography.

Factory Noise Although much of the radio noise encountered in factories

arises from weakly emitting sources, measurements have revealed that some types of industrial equipment produce harmonic RF energy and can radiate substantial noise (up to several hundred megahertz) [21]. Equipment such as RF- stabilized arc welders, induction heaters, and plastic bonders are acute sources of noise. Although interference is significant at H F and VHF, noise signatures of such equipment fall off rapidly above 1 GHz [21]. This trend has also been found in urban mobile-radio channels [22, p. 2971. Recent measurements have confirmed that typical machine-generated noise levels in operational factories are much less severe at higher frequencies [23]. Figure 1 shows results of peak noise power spectrum measurements (measured 4 m from engine cylinder machining line [23]) made along an engine manufacturing transfer line in full operation. When compared with the VHF spectrum, worst-case noise levels are 40 dB lower at UHF/ microwave frequencies and are only a few dB above the ther- mal noise floor of the spectrum analyzer receiver. Although a careful study of impulsive noise has not been conducted, these results are encouraging and indicate machine-generated noise will not severely hamper most factory radio systems operating at UHF and above.

Multipath Propagation Due to the large metal content of a factory, multipath inter-

ference is created by multiple reflections of the transmitted signal from the building structure and surrounding inventory. The resultant received waveform is a sum of time and frequency shifted versions of the original transmission, and, depend- ing on parameters of the signal and the channel, the received signal may be greatly distorted.

Multipath has historically been identified as the most important factor limiting mobile and portable radio communication systems [22] [24]. For narrow-band systems (where the baseband digital symbol duration is several times greater than the extent of the multipath-induced propagation delays, i.e., flat-fading conditions), multipath causes large fluctuations (fading) in the received signal voltage due to the changing phasor sum of signal components arriving at the receiver antenna via different paths. Temporal variations of the channel as well as changing multipath geometry seen by a mobile user are the mechanisms for the fading. Additional signal loss will occur when an AGV or portable user is shadowed by inventory and equipment. In wideband systems, the scatterers create intersymbol interference and cause the channel to be frequency selective. Consequently, the maximum data rate supported by a multipath channel is limited. Typical flat-fading channels require 30 dB more transmitter power to achieve low bit-error rates ( lop4) than do systems operating over ideal channels

In order to determine radio propagation characteristics inside factory buildings, radio wave propagation experiments at 1,300 MHz were conducted by the author in five operational factories in the spring and summer of 1987 [26]. Over 30,000 narrow-band fading measurements and 950 wideband impulse response measurements were made in a diverse collection of

1251.

16 May 1989 - IEEE Communications Magazine

VHF

-20-oo

n

E * -40.00 - E ti n 2 -60.00-

E g -40.00 v L a, 3 0 -60.00 n ‘0 a, > .- -80.00 8 2 1

-100.0 4 I 27.00 87.00 147.0 207.0 267.0 327.0

Frequency (MHz)

UHF -20-oo -

w a, > 0 a, K

-80.00 -

-1 00.0 4 I 1.2 1.3 1.4 1.5 1.6 1.7

Frequency (GHz) Fig. 1. Peak noise spectrum measurements.

industries and building structures. Factories that participated in the research included engine and automobile parts manufac- turers and dry-foods producers. Detailed descriptions of the experiments and the measurement equipment are given in [26] [ 321 [40]. Briefly, four distinct topographical scenarios were identified in each of the five factories. These ranged from Line- Of-Sight (LOS) transmission paths along lightly cluttered aisles to heavily cluttered obstructed paths between adjacent aisles. In each topography, three measurement locations were selected having graduated transmitter-receiver separations between 10 and 80 m. The transmitter was positioned (in the clear) in the center of the particular topography while the receiver was moved along a 1 m track at each measurement loca- tion. Wideband channel impulse responses were measured in the time domain by repetitively transmitting a 10 ns pulse (7.8 ns rms duration) and receiving on a digital storage oscilloscope the attenuated, distorted, and delayed versions of the pulse.

The factory channel measurement apparatus consisted of a periodic pulse-modulated transmitter with a peak output power of 1 W. The receiver consisted of a low-noise amplifier followed by a square law envelope detector and a 350 MHz digital storage oscilloscope. A directional coupler allowed CW envelope measurements to be made simultaneously by a modi- fied communications receiver. The measurement system block diagram is shown in Figure 2. Wideband discone antennas [28]

were usea at both transmitter and receiver and were mounted at heights of 2 m above the floor. Channel measurement results obtained using this apparatus are given subsequently.

The term “path loss” is used to describe the relative power attenuation (in dB) seen by a receiver with respect to a conve- nient (close) reference distance between transmitter and receiver. In H F communication systems, a reference distance of 1 km is often used [29, ch. 21. In recent wideband propagation measurements inside ofice buildings, a reference distance of 1 m was used [30]. With a reliable path loss model, one merely needs to know the transmitter power to accurately estimate the received power level for a particular distance from the transmitter, provided that the transmitting antenna used in the field is comparable to the type used to arrive at the path loss model.

CW measurements inside factory buildings revealed that, on the average, path loss increases according to a power law (in other words, received power falls off as an inverse power of distance) and is statistically described by a log-normal distribu- tion about a mean value given by:

Path loss (d) a d“ (1)

where d is the distance between transmitter and receiver in meters and n is the mean path loss exponent. For free space, n = 2. This model, which is sometimes called a large scale attenuation model, has been found to describe UHF channels inside and around houses and office buildings [IO] [ I l l [31]. The term large-scale indicates that the model holds for a large range of distances between transmitter and receiver and accounts for the gross attenuation characteristics of the channel with distance. Large-scale statistics are derived by measuring average or median received signal levels at many local areas throughout a desired coverage area. This is different from small-scale fading models, which statistically describe the signal level seen by a mobile receiver while it is moved about a small area (typically on the order of a few square meters) at a typical operating velocity [32] [33]. For small-scale fading models, the large- scale effects are ignored since the distance between transmitter and receiver is virtually constant over the local area, and the model describes fading due to temporal variation of the receiver position. This phenomenon is often referred to as fast- fading, as the signal statistics over relatively short time inter- vals are characterized.

It is interesting to note that inside buildings, where surrounding objects in the channel move very slowly, fading is primarily due to motion of the receiver. Historically, ionospheric and tropospheric communications systems were used (and are still in use) between fixed stations, and it was the temporal variation of the propagating medium (i.e., the ionosphere) that induced signal fading. consequently, fading statistics based on long-term and short-term temporal variations of the channels are used to determine link performance of many systems [29, ch. 6,7]. A definite distinction exists between models that describe signal strength as a function of distance as opposed to a function of time. The former is used in determining coverage areas and co-channel interference levels whereas the latter is valuable in determining bit-error rates and outage probabili- ties. By assuming that path loss (a distance-related phenomenon) and fast fading (a time-related phenomenon) are independent, predictions of instantaneously received signal levels for particular receiver locations and times can be computed through use of a probability density function that is the product of the two individual densities [31]. Preliminary work by the author indicates that such a modeling approach, when con- ditioned on the topography surrounding the receiver, accurately describes fading conditions inside factories.


Figure 3 shows the received signal strengths relative to a reference measurement made at 10 wavelength distance (2.3 m) in five factories. The figure is useful for computing large scale attenuation models for different factory topographies. Tables I and I1 indicate that in all cases, received power and Transmitter-Receiver (T-R) separation are highly correlated. In the tables, the mean path loss exponent A has been estimated from the slope of a linear least square error fit to the empirical data, where both distance and path loss are expressed logarithmically [3 1 1 [35]. In the tables, odenotes standard de- viation about the mean power law. Although path loss increases with distance more rapidly in obstructed topographies, large-scale attenuation in factory buildings does not appear to be as severe as in partitioned homes and office buildings, where path loss exponents typically range from 4 to 6 [ 131 [30] [3 11. This is most likely due to the large ceiling expanses, wide aisles, and metal ceiling truss work and inventory that support, rather than impede, radio wave propagation.

LOS Light x LOS Heavy 1 LOS Wall * OBS Light

Signal .= .. *I

I I I I I I I I I

5 10 20 50 100 Distance (m)

Fig. 3. Large-scale attenuation.

Because accurate descriptions of path obstacles were kept during the measurements, it is possible to extract from the data the RF signal loss caused by typical factory surroundings. By comparing the received signal levels for shadowed locations with the ensemble average of the factory measurements in a particular topography, average shadowing losses have been computed. Table I11 indicates typical shadowing losses that occur when a receiver is placed directly behind an obstruction (deep shadowing). Diffraction theory is used to predict the amount of RF energy that is received by an object that does not have a direct LOS path to the source. In many instances, an obstruction can be modeled as a knife-edge having a very thin width and a height equal to the physical height of the obstruction [27] [29]. Narrow-band measurements in factories reveal that path loss predicted by knife-edge diffraction theory is pes- simistic; deeply shadowed locations experience received signal levels consistently 5 to 20 dB larger than predicted by diffraction theory. For geometries where obstructions are located to- wards the middle of the direct path, knife-edge diffraction is in closer agreement with the empirical data. This indicates that, just as in urban radio channels, multipath from surrounding structures can illuminate receivers when a direct LOS does not exist.

Measurements made with a moving CW receiver over many small areas (1 m long tracks in the middle and sides of the aisles) reveal that fading is usually Rayleigh in heavily clut-

TABLE 1. Path Loss Exponent as a Function of Factory Building (1,300 MHz)

I SiteF I 1.92 I 4.79 I 17 1 .98 I

TABLE II. Path Loss Exponent as a Function of Factory Topography (1,300 MHz)

TABLE 111. Shadowing Effects of Common Factorv EauiDment (1.300 MHz)

tered LOS and lightly cluttered obstructed topographies; Rician for paths along perimeter walls and on lightly cluttered LOS paths; and log-normal for paths that traverse heavily cluttered obstructed topographies. Figure 4 illustrates some of these typical fading distributions and their fit to some of the observed fading data. Rician distributed channels are highly desirable as they require a much smaller fading margin in system design. There is also evidence that suggests Rician fading channels support significantly larger bandwidths than do Rayleigh fading channels [34]. LOS paths along factory aisleways were consistently found to be Rician (K = 6 dB), thus suggesting that for radio communications wideband radio LAN terminals should be mounted along and at the ends of aisles.


99 I OBS Light Clutter, Site C

LOS Heavy Clutter, Site E * LOS Light Clutter, Site D

20 1 . 7. ,f .y** . I :

O . I 0.2 Log-normal U = 7.5 dE Rayleigh Rician K = 6 dB

0 . 1 I / I ' I I I I I 1

-30 -20 -1 0 0 11 Signal Level (dB about median)

Fig. 4 . Cumulativedistributions of typical measurements and theirfit to various distributions.

Antenna diversity is a well-known technique for combating fading on narrow-band channels [36-381. The basic principal is to use several antennas separated in distance so that when one antenna receives a deeply faded signal, the other antenna(s) receive only a slightly faded version of the signal. The receiver then requires some level sensing circuitry that selects the best signal or various combining techniques to be used [38]. Diversity reception can provide several dB of gain against multipath when the antennas are positioned to achieve highly uncorrelated signal strengths. In manufacturing environments, measurements made over identical paths with different receiver antenna heights show that received signal strengths are often highly correlated (not independent) for vertical separations of two wavelengths (0.5 m). As seen in Figure 7, however, close- spaced antenna diversity may be useful when antennas are located with at least a quarter-wavelength separation along hori- zontal planes parallel to the ground. This can be seen by observing that the received signal levels on a particular antenna, at displacement differentials of approximately 0.06 m, consistently differ by more than 10 dB. Energy density antennas that couple both electric and magnetic fields are also useful in combating multipath fading in flat channels [ 391.

l5 1 . Crank Shaft Wash Line in Site F T-R separation is 25 rn

l o 1 Signal Level

(dB from median) -5

-10 -f

-15 -1 I I I I I 0 25 50 75 100

Time (s) Fig. 5 . Temporal fading measurement in engine assembly.

Factory channel impulse response measurements (also called power delay profiles) reveal that for LOS paths there typically exist only a few specular multipath components, with the

direct signal having a significantly larger signal level than the latter components. Over obstructed paths, however, when ei- ther the transmitter or receiver is shadowed by large equipment or by stacks of inventory, the predominant energy arrives 50 to 150 ns after the first observable signal [40]. To determine how individual signal components change with receiver motion, 19 equally-spaced power impulse response measurements were made along 1 m tracks throughout various topographies in five factories. Figure 8 illustrates how specular reflections from perimeter walls, etc. are easily distinguishable. In Figure 9, typical spatially-averaged multipath power delay profiles from various factories and topographies are shown.

% Probability Signal Level c Abscissa

99

f 0 Aisle in Site E * Aisle in Site F

Core Production, Site E rn Wash Line, Site F

Raylelgh O ? Riclan K = lOdB

0 1 1 -10

Signal Level (dB about median)

Fig. 6. Cumulative signal level distributions for temporal fading.

One measure of multipath conditions in a mobile radio channel is the root mean square (rms) delay-spread (q) which is inversely proportional to the maximum usable data rate of a channel [22] [41]. In [41], Bello characterized frequency selective channels in terms of a Taylor-series expansion of the average multipath power spectrum about the carrier frequency. He showed that for Wide Sense Stationary Uncorrelated Scatter- ing (WSSUS) channels, the rms delay-spread describes the ratio of power in the first (linear) frequency term of the series expansion to the power in the (flat fading) constant term. Un- derlying assumptions are that the frequency selectivity (the slope of the channel transfer function with frequency) changes slowly over the operating bandwidth.

Root-mean-square delay-spread is computed as the square root of the second central moment of an averaged power delay profile. This average is usually computed over time [41], but for a slowly varying channel such as is found inside factories, this average may be computed over space by using many profile measurements in a local area [30] [40]. Analysis of the propagation database from inside factories has revealed that multipath characteristics can be correlated with building structure and type of inventory. Measurements in a food processing factory that manufactures dry-goods and has considerably less metal inventory than most factories had oT values consistently half of those observed in factories that produce metal products. Newer factories which incorporate steel beams and steel- reinforced concrete in the building structure have larger delay spreads than older factories which use wood and brick for perimeter walls. Summarizing the results of the wideband measurements, the worst-case oT value was 300 ns in a modem engine plant. Typical values ranged between 100 ns and 200 ns.


Antenna Height 2 m 1.5 m ......

A 2 0 4 -

10 - Relative

Received

Strength Signal 0 -

(d B)

-10 -

-20 -

-0.5 0 0.5 D i s p I a cem e n t (m)

Fig. 7. Received signal levels using antenna height diversity.

Unlike in office buildings [ 301, delay-spread values in factories do not appear to depend on whether or not there exists a LOS path [40]. This is due to the open expanses and vast amount of reflecting material that readily supports multipath propagation.

In [22], DPSK bit-error-rate analyses for mobile radio channels undergoing both Rayleigh envelope (flat) fading and frequency selective fading indicate that as signaling bandwidth increases, multipath-induced intersymbol interference becomes the main cause of performance degradation. Our measurements show a worst-case rms delay-spread of 300 ns inside modern factory buildings. From Figure 4.2.9 in [22], it is easy to show that in order to achieve an irreducible bit error rate of better than 1 O P 3 , baseband data rates must be limited to below

Fig. 8. Close1.v-spaced power impulse response measuremen ts.

1 .o

0.5 B n

0 0 2 5 0 5 0 0

Exwss Delay (ns)

lal

LOS Light Clutter LOS Heavy Clutter

1.0, l - O l

0 0 250 500 Excess Delay (ns) LOS Along Wall

Excess Delay (ns) OBS Light Clutter

B 5 0.54

0 h 0 250 so0

~xwss Delay (ns) OBS Heavy Clutter

Fig. 9. Power impulse responses.

150 kb/s for rectangular pulse shapes (this is found from equa- tion 4.2.71 in [22]). Raised-cosine pulse shapes can increase the worst-case throughput to 250 kb/s for the same channel. It must be stressed that measurements in [40] are the only known wideband measurements from within factories, and it may be that some buildings impose more severe frequency-selective fading conditions. In [30], Devasirvatham found an instance where rms delay-spread outside of a building was approximately 600 ns. It will be possible to improve the capacity of indoor digital radio communication systems through robust modulation techniques [42] [43], distributed antennas [44], and antenna diversity techniques [36] [37].

Important issues in the area of wideband indoor radio propagation measurements include measurement dynamic range and the effects of operating frequency on multipath. To date only three wideband indoor radio propagation works have been reported [I I] [26] [30]. In [ 1 I], an apparatus that uses both a frequency swept technique and a pulsed transmitter was used to measure average multipath power delay profiles. In [26] and [40], a similar but simpler apparatus was used, and the average power profile was computed from a spatial average of individual profile measurements over several wavelengths. In [30], a spread-spectrum system was used to achieve greater dynamic range at the expense of temporal resolution. or values reported in [ 1 I] suggest that indoor radio channels could support bandwidths several times as large as those indicated by


measurements in [26] and [30]. Consequently, Bit Error Rate (BER) calculations based on measurements in [ 1 11 are proba- bly optimistic. It is interesting to note that measurements in [ 1 11 were made at 1.5 GHz, measurements in [26] were made at 1.3 GHz, and measurements in [30] were made at 850 MHz. Because of the limited amount of wideband indoor propagation data, it is unknown at this point what effect carrier frequency has on indoor radio propagation. One would think that in office buildings the delay spread would decrease with frequency due to increased attenuation by the structure. In factory buildings, however, this may not be the case.

Recently, rapid changes in channel group delay have been found to cause burst error in digital communication systems due to shifts in eye pattern timing [ 141 [45] [62]. As described in [ 141, it is necessary to use a phase-lock loop to accurately track the mean delay of the channel in order to detect digital symbols at the point where the signal-to-noise ratio is largest. In high-data-rate indoor radio systems, where channels vary slowly, the phase-lock loop will track the instantaneous centroid (first moment) of the impulse response. A phenomenon known as jitter occurs when the centroid of the impulse response changes rapidly, causing loss of bit synchronization and resulting in burst errors as the receiver attempts to reacquire bit synchronization. Differential-delay jitter, which is the rate at which the impulse-response centroid changes over time (or distance for a mobile moving at a constant velocity), is a useful design parameter since it indicates the required slew rate for a bit synchronization circuit. When the duration of a data bit is much larger (at least a factor of four or five greater) than the temporal extent of the indoor channel impulse response, it has been shown in [45] [62] that burst errors are caused primarily by envelope fades and not by timing jitter. In [ 181, empirical measurements indicate that inside factories, the impulse response centroid can change by as much as 180 ns with just a few centimeters of movement at the portable.

Multiple-Access Networking Considerations

Future factory radio communication systems will rely on multiple-access techniques to accommodate many fixed and mobile terminals. Multiple-access techniques such as Frequen- cy Division Multiple-Access (FDMA), Time-Division Multi- ple-Access (TDMA), Code Division Multiple-Access (CDMA, also called spread-spectrum), and Carrier-Sense Multiple- Access (CSMA, also called packet radio) partition channel users into non-overlapping signal spaces [46-5 11. Because of non-idealities, however, there is inevitable overlapping of sig- nals and the resulting co-channel interference can appear as lengthy message delays or as degradation of the desired received signal [47-5 11.

Random access (CSMA, CDMA) radio LANs are attractive because they have relatively few synchronization requirements and support a distributed architecture. Unlike fixed assignment networks (FDMA, TDMA), which assume all users require their own channel, random access techniques rely on “bursty users” and assume that the likelihood of many users using the network at one time is small. For AGVs using reliable dead-reckoning systems, infrequent position updates suggest the use of random access networking for AGV control. In fact, for mobile robot systems which navigate by buried wire or paint strip, CSMA is a popular choice because communications traffic is generally very light. On the other hand, direct control of a fleet of vehicles by a central dispatching station (such as would be warranted for cleaning a contaminated nuclear reactor) might require a fixed assignment scheme. There also exist many types of demand- assignment access techniques that appear particularly suited to indoor radio networking [49] [53].

Considerable progress has been made in determining realistic limitations on the delay characteristics of packet networks [54], and a powerful product-form solution model for packet radio systems has been developed [55]. CSMA and CDMA strategies can be merged to enhance multiple-access communication performance in a multipath environment while providing some ranging capability [56] [57]. In [56], fundamental ex- pressions that permit the calculation of BERs in packet spread-spectrum systems have been provided; these expres- sions permit system designers to analyze throughput and delay as a function of number of simultaneous users.

Work on indoor radio communication systems has revealed that by using a slotted packet radio technique (slotted ALOHA), overall system throughput and delay characteristics can be improved through capture [58]. A packet reservation technique, which accommodates both voice and data, is an example of a slotted, packetized multiple-access technique that exploits the burstiness of speech and gives priority to data transmission [59].

The July 1987 issue of the IEEE Journal on Selected Areas in Communications [49] contains several works on recent developments in multiple-access networks and their performanc- es. In [60], a modeling approach has been used to include exact propagation timing for events on broadcast channels. Results from [60] show that the topology of a linear (bus-like) network should be considered when analyzing the performance of packet radio protocols, since previous protocol models underesti- mate performance. This indicates that for futuristic wideband indoor packet radio systems, the locations of repeaters may be an important design consideration not only from the standpoint of link budget and minimization of multipath, but also from the standpoint of channel capacity due to the inherent physical structure of the network. In [61], a CSMA protocol that uses a tone-modulated preamble has been developed to re- solve contentions on packet radio channels.

There has been much interest in the use of TDMA for mobile and portable radio communications. In Europe, a Pan- European TDMA standard has been agreed upon for cellular radiotelephone, and several vendors are proposing TDMA for U.S. digital cellular systems. TDMA is also a viable technique for multiple-user indoor radio communications, and TDMA communication system prototypes are being developed at major research laboratories [ 121 [ 131 [ 191 [62]. TDMA offers many advantages, including communications flexibility and cheaper hardware. Some presently available factory radio systems employ low data-rate TDMA and use some form of priority scheduling to reallocate time slots that would normally go un- used. A subtle but potentially important advantage of TDMA is that a mobile user can listen to the base as it transmits to other users within the frame. During this listening period, the mobile can use small-scale antenna diversity to choose the antenna for best reception (and transmission) for its upcoming time slot. A further advantage is that TDMA implies that only one user transmits at a time. Thus, class-C amplifiers, which are more efficient but have rich harmonic and intermodulation components, can be used. In FDMA, class-C amplifiers can create adjacent channel interference if users are in close proximity to one another (although feedback techniques exist which can reduce intermodulation emissions of class-C amplifiers).

FDMA is advantageous because the multiple-access system can be built up around existing and field-proven narrow-band technology. Although FDMA requires more hardware for a given number of users (because of a greater number of communication channels), the data rates in each of these channels is small. Thus, users in an FDMA system are not subjected to multipath-induced intersymbol interference, which can affect high bit-rate TDMA channels. Equalization, which becomes an issue for TDMA systems, is usually not needed for FDMA because each channel undergoes flat fading.


As the number of users increases, the real-time communications capability of random access techniques diminish, and a fixed assignment approach is required. Furthermore, if central computers using parallel processing architectures are required to simultaneously communicate, navigate, and control many simultaneous users on a virtually continuous basis, TDMA or FDMA approaches may be desired. Portable/mobile users transmitting large blocks of data (i.e., MIS, video transmissions, high resolution graphics, maps, etc.) are accommodated best by a fixed assignment network. Selection of networking suategies for radio links inside buildings will depend heavily upon the number of users, the duration of transmissions, the limit of sophistication at each terminal, the importance of real- time control, and whether or not it is desired to use radio for AGV navigation.

Conclusion As part of the research mission of the NSF Engineering Re-

search Center for Intelligent Manufacturing Systems, measurement, characterization, and modeling of indoor factory radio channels have been carried out. The work reveals that man- made noise is not a serious problem for indoor factory radio systems at frequencies greater than 1 GHz, and that fading characteristics are highly dependent upon local topography in the workplace. Shadowing data and large scale path loss models have been developed and form the basis for designing reliable narrow-band indoor radio LANs for portable communications and AGV control. Wideband measurements reveal that commercially available technology currently limits data rates to on the order of 150 kb/s in typical factory channels. While this accommodates current needs, it is anticipated that greater capacity will be required for the highly automated and flexible factories of the future. Ongoing work at Virginia Tech is aimed at developing robust wideband multi-access communication system designs and signaling techniques for indoor radio communications.

Acknowledgments The author wishes to thank Clare McGillem and Heidi

Peterson of Purdue University, and Charles Bostian, Tim Pratt, and Warren Stutzman of the Virginia Tech Satellite Communications Group for their helpful suggestions during the preparation of this article. The thoughtful comments of the reviewers are appreciated.

References M. Kaminski, Jr., "Protocols for Communicating in the Factory,- IEEE Spectrum, vol. 23, no. 4, pp. 56-62, Apr. 1986. S . Farowich, 'Communicating in the Technical Office,- IEEESpectrum, vol. 23, no. 4, pp. 63-67, Apr. 1986. National Academy of Engineers, Engineering Research Centers: Leaders in Change, Washington, D.C.: National Academy Press, 1987. L. Mayfield, 'NSF's ERC Program-How it Developed,' Engineering Education, pp. 130-132, Nov. 1987. J. Solberg and C. McGillem, "Automated Guided Vehicle Systems,- ln- ternational Trends in Manufacturing Technology Series, R. Holler, Edi., United Kingdom: IFS Publications Ltd., 1986. T. Rappaport, "The Design and Development of a Mobile Robot Loca- tion System,- Masters Thesis in Electrical Engineering, Purdue Univer- sity, West Lafayette, IN, Dec. 1984. C. McGillem and T. Rappaport. "An Infra-red Navigation Technique for Mobile Robots," Proc. 1988 IEEE lntnl. Conf on Robotics andAutorna- tion, Philadelphia, PA, Apr. 26, 1988. C. McGillem, J. Callison, and T. Rappaport, -Autonomous Roving Ve- hicle Navigation, Control, and Communication,' 9th lntnl. Conf. on Au- tomation in WarehousinglGth lntnl. Conf on Automated Guided Vehicle Systems, Brussels, Belgium, Oct. 25, 1988.

G. Schettler, "Developments in Radio Frequency Transponders,- 1988 Materials Handling FOCUS Conference, Materials Handling Re- search Center, Georgia Institute of Technology, Atlanta, GA, Sept. 15, 1988. D. Cox, H. Arnold, and P. Porter, "Universal Digital Portable Communications-A System Perspective,- IEEE Trans. J. Se/. Areas Comm., vol. SAC-5, no. 5, pp. 764-773, June 1987. A. A. M. Saleh and R. Valenzuela, "A Statistical Model for Indoor Multipath Propagation," I€€€ Trans. J. Se/. Areas Cornrn., vol. SAC-5, no. 2, pp. 138-146, Feb. 1987. A. Saleh, A. Rustako, Jr., and L. Cimini, Jr., "A TDMA Indoor Radio Communications System using Cyclical Slow Frequency Hopping and Coding-Experimental Results and Implementation Issues. IEEE Globecom '88 Proc., pp. 41.3.1-6. R. Bernhardt, "Two-way Transmission Quality in Portable Radio Sys- tems," IEEE Globecorn '88 Proc., pp. 41.2.1-6. J. C-l Chuang, "The Effects of Multipath Delay Spread on Timing Re- covery," I€€€ Trans. Veh. Tech., vol. VT-36, no. 3, pp. 135- 140, Aug. 1987. L. Couch 11, Digital and Analog Communication Systems, MacMillan, Second Ed., 1987. M. Marcus, "Regulatory Policy Considerations for Radio Local Area Networks," IEEEComrnunicationsMagazine, vol. 25, no. 7. pp. 95-99, July 1987. H. Nogami and M. Wakao, "Premises Radio Systems in Japan," IEEE Communications Magazine, vol. 25, no. 10, pp. 49-55, Oct. 1987. T. Rappaport, -Delay Spread and Time Delay Jitter in Manufacturing Environments," 1988 IEEE Veh. Tech. Conf Record, Philadelphia, PA, pp. 186-189, June 15, 1988. J. C-l Chuang and N. Sollenberger, 'Burst Coherent Detection with Ro- bust Frequency and Timing Estimation for Portable Radio Communica- tions,- IEEE Globecom '88 Proc. pp. 26.1.1-6. T. Takeuchi, F. Ikegami, S . Yoshida, and N. Kikuma, 'Comparison of Multipath Delay Characteristics with BER Performance of High Speed Digital Mobile Transmission,- 1988 IEEE Veh. Tech. Conf Proc., pp. 119-126, June 15, 1988. E. Skomal, Man-Made Radio Noise, NY: Van Nostrand Reinhold Co., Ch. 3 and 4, 1978. W. Jakes, Microwave Mobile Communications, Wiley, 1974. L. Duran, "Characteristics of Electromagnetic Interference in Industrial Environments,- Masters Thesis in Electrical Engineering, Purdue Uni- versity, Dec. 1987. R. S . Kennedy, Fading Dispersive Communication Channels, Wiley, Ch.

R. Bultitude, "Measurement, Characterization and Modelling of Indoor 800/900 MHz Radio Channels for Digital Communications,' IEEECom- munications Magazine, vol. 25, no. 6, pp. 5-12, June 1987. T. Rappaport, "Characterizing the UHF Factory Multipath Channel,- Ph.D. Dissertation, Purdue University, Dec. 1987. Also available as Tech. Report TR-ERC88-12. NSF Engineering Research Center for In- telligent Manufacturing Systems, Purdue University, June 1988. R. Colin, Antennas and Radiowave Propagation, McGraw-Hill, pp. 369- 374, 1985. T. Rappaport, "Simple-to-Build Wide-Band Antennas," RFDesign, pp. 35-41, Apr. 1988. J. Griffiths, Radio Wave Propagation and Antennas, Englewood Cliffs, NJ: Prentice-Hall International, 1987. D. Devasirvatham, 'Time Delay Spread and Signal Level Measure- ments of 850 MHz Radio Waves in Building Environments, IEEE Trans. Ant. Prop., vol. AP-34, no. 1 1 , pp. 1,300-1.308. Nov. 1986. D. Cox, R. Murray, and A. Norris, '800-MHz Attenuation Measured In and Around Suburban Houses,- AT&TBellLaboratories TechnicalJour- nal, vol. 63, no. 6, pp. 921-954, July-Aug. 1984. T. Rappaport and C. McGillem, -UHF Fading in Factories.- IEEE J. Se/. Areas Comm.. Jan. 1989. D. Cox, R. Murray, and A. Norris. -Measurements of 8W-MHz Radio Transmission into Buildings with Metallic Walls,' Bellsystem Tech. J., vol. 62, no. 9, pp. 2,695, 2,716, Nov. 1983. R. Bultitude, "Measurement, Characterization and Modelling of 8001 900 MHz Mobile Radio Channels for Digital Communications, Ph.D. Thesis, Carleton University, Ottawa, Feb. 1987. W. Dixon and F. Massey, Jr., Introduction to Statistical Analysis. McGraw-Hill, Fourth Edition, Ch. 11, 1983. R. G. Vaughan and J. B. Andersen, -Antenna Diversity in Mobile Com- munications," IEEE Trans. Veh. Tech., vol. VT-36, no. 4, pp. 149-1 72, Nov. 1987. J. Winters, "Optimum Combining for Indoor Radio Systems with Multi- ple Users," IEEE Trans. Comm., vol. COM-35, no. 11, pp. 1,222- 1,230, Nov. 1987. S. Stein, "Fading Channel Issues in System Engineering (Invited Tutori- al).- IEEE J. Sel. Areas Comm., vol. SAC-5, no. 2, pp. 68-89, Feb. 1987. K. It0 and S . Sasaki, "A Small Printed Antenna Composed of Slot and Wide Strip for Indoor Communication Systems,- 1988 IEEE Antennas and Propagation Symposium. pp. 7 16-7 19, June 12, 1988.

1-3, 1969.

May 1989 - lEEE Communications Magazine 23

T. Rappaport, -Characterization of UHF Multipath Radio Channels in Manufacturing Environments," IEEE Trans. Ant. Prop.. Aug. 1989. Pre- sented in part at IEEE GLOBECOM '88, Hollywood, FL, pp. 26-5.1,

P. Bello, 'Characterization of Randomly Time-variant Linear Channels, IRE Trans. Comm. Sys. vol. CS-11, pp. 361-393, Dec. 1963. R. Valenzuela, 'Performance of Quadrature Amplitude Modulation for Indoor Radio Communications,' IEEE Trans. Comm., vol. COM-35, no. 11, Dec. 1987. M. Kavehrad and G. Bodeep, "Design and Experimental Results for Direct-Sequence Spread-Spectrum Radio using Differential Phase- Shift Keying Modulation for indoor Wireless Communications," IEEEJ. Se/. Areas Comm., vol. SAC-5, no. 5, pp. 815-823, June 1987. A. A. M. Saleh, A. J. Rustako, Jr., and R. Roman, 'Distributed Anten- nas for Indoor Radio Communications," IEEE Trans. Comm.. vol. COM- 35, no. 12, pp. 1,245-1,251, Dec. 1987. S . Ariyavisitakul et al., -Fractional-bit Differential Detection of MSK: A Scheme to Avoid Outages Due to FRequency-selective Fading,' IEEE Trans. Veh. Tech., vol. VT-36, no. 1, pp. 36-42, Feb. 1987. W. Stallings, Data and Computer Communications, MacMillan. Second Ed., 1985. I. Rubin, -Message Delays in FDMA and TDMA Communication Chan- nels,' IEEE Trans. Comm., vol. COM-27, no. 5, pp. 769-777, May 1979. L. Kleinrock and F. Tobagi, 'Packet Switching in Radio Channels: Part I: Carrier Sense Multiple Access Modes and Their Throughput-delay Characteristics,- IEEE Trans. Comm., vol. COM-23, no. 12, pp. 1.400- i,416, Dec. 1975. Special Issue on Performance Evaluation of Multiple-Access Net- works, IEEE J . Se/. Areas Comm., V. 0. K. Li, Edi., vol. SAC-5, no. 6, July 1987. M. Pursley, 'Performance Evaluation for Phase-coded Speread- spectrum Multiple-access Communication-Part I: System Analysis,' IEEE Trans. Comm., vol. COM-25, no. 8, pp. 795-799, Aug. 1977. L. Merakos, G. Exley, and C. Bisdikian, 'Interconnection of CSMA Local Area Networks: The Frequency Division Approach,' IEEE Trans. Comm.. vol. COM-35, pp. 730-738. July 1987. L. Lee and C. Un, "A Code-division Multiple-access Local Area Net- work,' IEEE Trans. Comm.. vol. COM-35. no. 6, pp. 667-671. June 1987. S . Sachs, -Alternative Local Area Network Access Protocols,' IEEE Communications Magazine, vol. 26, no. 3, pp. 25-45, Mar. 1988. S . Beuerman and E. Coyle, 'The Delay Characteristics of CSMA/CD Networks,' IEEE Trans. Comrn., vol. 36, no. 5, pp. 553-563, May 1988. R. L. Hamilton, Jr. and E. Coyle, -A Two-Hop Packet Radio Network with Product Form Solution,' Proc. 20th Annual Conf. on Info. Scienc- es and Systems, Princeton Univ., pp. 871-876, Mar. 1986.

26-5.7, NOV. 29, 1988.

R. Morrow, Jr. and J. Lehnert, 'Bit-to-Bit Error Dependence in Slotted DS/SSMA Packet Systems with Random Signature Sequences,' ac- cepted to IEEE Trans. Comm., to appear early 1989. J. Fischer, J. Cafarella, C. Bouman, G. Flynn, V. Dolat, and R. Boisvert, 'Wide-Band Packet Radio for Multipath Environments,' IEEE Trans. Comm., vol. 36, no. 5, pp. 564-573, May 1988. D. Goodman and A. Saleh, 'The Near/Far Effect in Local ALOHA Radio Communications.' IEEE Trans. Veh. Tech., vol. VT-36, no. 1, pp. 19- 27, Feb. 1987. D. Goodman, R. Valenzuela, K. Gayliard, and B. Ramamurthi, 'Packet Reservation Multiple Access for Local Wireless Communications,' 1988 IEEE Veh. Tech. Conf Record, Philadelphia, PA, pp. 701-707, June 15, 1988. M. Molle, K. Sohraby, and A. Venetsanopoulos,'Space-Time Models of Asynchronous CSMA Protocols for Local Area Networks,' IEEE J. Se/. Areas Comm., vol. SAC-5, no. 6, pp. 956-968, July 1987. W. Lo and H. Mouftah, 'Collision Detection and Multitone Tree Search for Multiple-Access Protocols on Radio Channels,' IEEE J. Se/. Area Comm., vol. SAC-5, no. 6, pp. 1,035-1,040, July 1987. J. C-l Chuang, 'The Effects of Time Delay Spread on Portable Radio Communications Channels with Digital Modulation,' IEEE J. Se/. Areas Comm., vol. SAC-5, no. 5, pp. 879-889, June 1987.

Biography Theodore S. Rappaport was born on November 26, 1960. He received

his B.S.E.E., M.S.E.E., and Ph.D. degrees from Purdue University, West Lafayette, IN, in 1982, 1984, and 1987, respectively. During his undergradu- ate career, he received three national scholarships from the Foundation for Amateur Radio, including the Radio Club of America Scholarship during his senior year. As a graduate student, he received the GTE Graduate Fellowship in Electrical Engineering and the EXXON Electrical Engineering Fellowship. He was with the Government Aerospace Systems Division of Harris Corporation, Melbourne, FL. during the summers of 1983 and 1986. From 1987 to 1988 he was manager of the Autonomous Guided Vehicle Project at the NSF Engineer- ing Research Center for Intelligent Manufacturing Systems, Purdue University, West Lafayette, IN. There he managed the development of a prototype AGV fleet and conducted research in indoor radio communications for manufacturing environments. He is presently an Assistant Professor with the Bradley De- partment of Electrical Engineering, Virginia Polytechnic Institute and State Uni- versity, Blacksburg, VA where he is teaching and conducting research in the areas of digital and satellite communications, radio wave propagation, and antennas. Dr. Rappaport is a consultant to the AGV and cellular radiotelephone industries, and is a reviewer for the IEEE.


Date post:	14-Apr-2020
Category:	Documents
Upload:	others
View:	1 times
Download:	0 times

Indoor radio communications for factories of the future...

Documents