Sensing Shape

Within days of the collapse of the twin towers in New York on September 11, 2001, dozens of urban search and rescue robots were crawling through the rubble to search for survivors. These robots had any number of means to find the survivors: microphones to detect voices or other sounds of possible human presence within the ruins, thermal cameras to register body heat, other cameras to search for colors distinctive from the gray dust that had blanketed the debris, and so on.

Unfortunately, these robots were of limited use; when they successfully located a survivor, the presence of rubble effectively shielded the robot from the omnipresent global positioning system (GPS) signals that they required to assess their own location. In other words, when the robots found survivors, rescue workers could not find the robots.

It was this event and other defense applications that compelled the development of a new type of tracking system: fiber GPS. Fiber GPS is a high-spatial-resolution, distributed fiberoptic sensing cable designed to provide real-time feedback on its own dynamic shape and position. If we know the shape of the fiber over its entire length and the point of its origin, we can determine the location of the end of the fiber; fiber GPS thus enables shape feedback of structures of interest and a means by which a tethered object may be accurately tracked. The sensing cable is minimally intrusive, lightweight, immune to electromagnetic interference, and extremely robust, making deployment in harsh environments practical.

Figure 1. In fiber GPS, axially co-located FBGs form sensing elements that function as discrete 3-D bend sensors. The density of sensing elements ensures the accurate reconstruction of global shape.

The sensing cable consists of high-density, linear arrays of fiber Bragg gratings (FBGs) written simultaneously into the cores of a multicore fiber (MCF) and aligned in the dimension of the cross-section. The FBGs act as strain sensors, detecting strain introduced to the silica media by local bending. Algorithms use the strain differential to calculate 3-D local bend at every discrete element along the length of the sensing cable. Because of the sensor density, each individual sensing element can be integrated to reconstruct the total shape of the cable (see figure 1).

MCF is more complex than standard telecommunica-tions fiber but is fabricated in much the same way. Though numerous methods can be used to achieve the desired geometry, the preferred methods are the multi-chuck over-cladding procedure and the stack-and-draw process.

Both processes begin the same way. First, we design and model the optical parameters (refractive index profile, core/cladding diameters, etc.) of the preform to obtain the desired waveguiding performance. Next, for each fiber in the finished bundle, we fabricate a preform with the desired dopants and numerical aperture. These preforms are first stretched, then sectioned to the appropriate lengths and inserted into a silica tube with additional glass rods to fill the voids in the tube.

The variation in the two procedures arises from the method by which the preform rods are inserted into the tube. In the multi-chuck method, glass rods and preforms are positioned in the tube on a glass-working lathe, using a double chuck to align them in the tube. Once the constituent parts are positioned, the tube is collapsed on the glass rods to create the preform, which is fiberized in the draw tower by the standard procedure. In the stack-and-draw process, the preforms and bait rods are stacked up together in the silica tube, and the interstitial space is filled with additional glass rods. The resultant glass assembly is then fiberized.

The fundamental elements of our system are the FBGs, which are produced by exposing a photosensitive fiber to a pattern of pulsed UV light to form a periodic change in the refractive index of the core (see oemagazine, January 2001, p. 38). This pattern, or grating, reflects a very narrow frequency band of light that is dependent on the modu-lation period formed in the core. In its most basic operation as a sensor, an FBG is either stretched or compressed by an external stimulus. This results in a change in the modulation period of the grating, which, in turn, causes a shift in the frequency reflected by the grating. By measuring this shift in frequency, we can determine the magnitude of the external stimulus applied (see figure 2).

Figure 2. In a fiber Bragg grating, a portion of the incident light traveling through a fiber reflects from a refractive-index grating written in the fiber (left). If the fiber is stressed (right), the grating period changes and the wavelength of light reflected changes also.

One of the most significant challenges in the development of the fiber GPS concept was the need to read the large FBG arrays necessary to accurately reconstruct global fiber shape. Distributed fiberoptic sensing techniques have gained increasing popularity in recent years, and it was here that a solution would be found.

Optical time-domain reflectometry involves launching a pulse of light into a fiber and tracking the time it takes a signal to undergo Rayleigh or Fresnel reflection and return to the detector. The time required for travel yields the distance to the sensing mechanism. Although this technique works very well over kilometer-scale distances, spatial resolution tends to be coarse (on the order of meters).

An alternative was a wavelength division multiplexing (WDM) technique in which each sensor reflects a different wavelength of light. Although this approach improves the spatial resolution significantly, the number of sensors tends to be limited to the order of tens due to the finite bandwidth of broadband or swept-wavelength laser sources and the need to allocate significant portions of that bandwidth to each sensor to accommodate the dynamic range of the application.

Our sensing application requires dynamic ranges of ±10,000 µm, which is equivalent to nearly 20 nm of spectral bandwidth. Using a WDM-based technique would require either that each FBG be allocated the full spectral bandwidth expect-ed, or that the dynamic range of the measurements be constrained. Furthermore, WDM sensor fabrication is very labor intensive, and thus, very costly.

Optical frequency-domain reflectometry (OFDR) provides an attractive alternative to WDM. Originally developed by researchers at NASA Langley Research Center for testing on the X-33 and the space shuttle, OFDR has recently been commercialized for numerous monitoring applications. The technique allows thousands of sensors with overlapping spectra to be read with very-high spatial resolution, giving the most complete picture of any of the viable distributed fiberoptic sensing techniques. By applying OFDR to the detection of the FBG arrays in the MCF, we can feasibly achieve the sensor densities necessary for high-resolution shape sensing or object tracking.

In an OFDR system, we interrogate the gratings with a high-coherence, widely tunable swept-wavelength source (see figure 3). Each grating is spaced a unique distance from a broadband reflector such that each grating/reflector combination forms an interferometer with a unique optical path difference. This interference modulates the reflected spectrum of each grating with a unique frequency that is directly proportional to the optical path difference. When a Fourier transform is performed on the raw data, the signal from each grating then occupies a unique band in the frequency domain. We can extract the signal from each grating using a bandpass filter. This data is then inverse-transformed and the center wavelength of each grating's reflected spectrum is determined. We determine the magnitude of the applied stimulus by interpreting the shift in the center wavelength from the nominal center wavelength.

Figure 3. In optical frequency-domain reflectometry, each grating G₁, G₂, ..., G_n forms an interferometer with a broadband reflector R at some distance L₁, L₂, ..., L_n (left). The interference causes the signal from each sensor to be modulated by a unique frequency ƒ₁, ƒ₂, ..., ƒ_n, which can be mapped to a unique frequency location and windowed by a bandpass filter function, allowing the retrieval of each individual sensor signal (right).

This demodulation technique does not require gratings to have high reflectivity, but it can accommodate extremely weakly reflecting gratings (-34 dB is typical). In addition, each of the FBGs measured can have identical nominal reflected spectra. Spectral overlap and sensitivity allow us to write the gratings in the fiber while it is in the draw tower, reducing average sensor cost significantly by automating the process (see figure 4). Writing the gratings at this point also allows us to avoid the strip and recoat process, which can weaken the fiber structurally. OFDR allows us to multiplex many more sensors - thousands instead of tens - on a single fiber and achieve higher spatial resolutions than those of competing methods (1 cm center-to-center is typical). The technique also requires less spectral bandwidth for complete detection while widening the dynamic range of the parameter being sensed.

Figure 4. We can write FBGs in multicore fiber during the fiber draw process, significantly reducing sensor manufacturing cost through automation.

One of the most significant challenges in the development of this type of sensing system is the need to develop methodologies and procedures for determining and verifying accuracy. Given that no comparable technology exists, there is little historical context for metrics that are universally accepted and understood. This characterization has been carried out to date primarily by forcing the shape-monitoring cable into known shapes with known spatial coordinates and then measuring deviations from fixed spatial points. Circles, sinusoids, sharp and gradual curves, as well as more complex shapes have been produced with results to date yielding an average error of 1.2% of total shape-cable length (see figure 5). (Here error is the vector deviation from the actual, normalized to the actual.)

Figure 5. The shape sensor "knows" its own shape along its entire length, from which we can determine position. The shape measurement is direct; it does not need to be inferred from other parameters, does not require recalibration, and is completely stand-alone.

Additional experimentation has been carried out by forcing the shape cable to conform to structures of interest and then subjecting these structures to a variety of shape-altering events including cantilever bending, three-point bending, and dynamic excitation. The shape cables used in these tests were approximately 1 m in length with FBGs written into each core of a tri-core fiber at 1-cm center-to-center spatial resolution; these shape cables consisted of 300 FBGs (100 per core). The results showed that the fiber GPS was able to accurately reconstruct the shape of the structures undergoing cantilever and three-point bending (and, thus, end-point location) with an error of 1.2% of the length of the tether. In addition, when the dynamic displacement data was Fourier transformed, the frequency of excitation could be reproduced with sub-Hertz accuracy.

Though the technology was originally developed to assist in the location of survivors trapped in rubble piles, it has numerous other applications. It is no great stretch of the imagination to picture how a system that provides highly accurate positional information on tethered robots might be used in more proactive defense applications than search and rescue; for example, tethered robots have been developed for exploration of cave systems where GPS signals are unavailable. With the additional functionality enabled by fiber GPS, a team located a safe distance away from a hazard can track the robots' positions with ex-tremely high accuracy.

Unmanned vehicles are becoming increasingly relevant as they move from reconnaissance plat-forms to tools that can take the fight to the enemy, particularly in urban war-fare situations in which each advance brings new dangers from booby traps and hidden adversaries. In such cases, increased posi-tional information means increased situational aware-ness and a reduction in risk for the combatant. Additional applications include real-time shape feedback for aircraft with adaptable attributes (morphing wings), towed sonar arrays, and deployable space structures, as well as medical applications such as minimally invasive surgical techniques. oe

This work was funded by the National Aeronautics and Space Administration's Langley Research Center and Goddard Space Flight Center.

1. R. Duncan et al., Materials Evaluation 61, p. 838 (2003).

2. R. Duncan et al., Space Technology and Applications International Forum 2005, AIF Conference Proceedings 746, p. 880 (2005).