**2. Radiation sensors**

Since UAS has weight and size limitations for onboard components so that the flight efficiency is not reduced due to the payload, the choice of radiation sensors is important. Radiation sensors should be swappable (attached and detached from the robotic platform) in the field conditions. The sensors should have low power requirements so that the integration into the platform does not affect the battery power. The data processing limitations should be also taken into account. To address this necessity, two types of flight ready ambient temperature radiation sensors were developed: a Cadmium Zinc Telluride (CZT) semiconductor sensor for high-resolution gamma spectrometry and the elpasolite scintillation sensor Cs2LiYCl6:Ce3+ (CLYC) for neutron and gamma measurements. Both sensors were designed as plug-and-play interchangeable modules.

#### **2.1 CZT sensor**

A CZT is a wide bandgap semiconductor [7]. Gamma rays that interact with this detection medium deposit their energy into the CZT operating in a directconversion mode at an ambient temperature. Therefore, no cooling is needed which significantly simplifies the sensor's integration into the robots and their practical use. These semiconductors are able to process more than 10 million photons per square millimeter per second.

The CZT module GR1-A (Kromek) [8] was integrated into a multicopter aerial platform. The GR1-A includes a 1 cm3 cubic CZT crystal. Electric signals generated by the CZT proportionally to energy of the incident gamma ray are amplified by a preamplifier and then processed by a shaping amplifier. A 4096 channel analyzer produces the discrete data array of the gamma spectrum that can be processed further. The scheme of the CZT sensor operation is shown in **Figure 1**. The USB interface is used for the data communication as well as power. The power consumption of this module is about 250 milliwatts. The module's volume is 2.5 × 2.5 × 6.1 cm3 and the weight is about 50 grams. The sensor's energy range is from 30 keV to 3 MeV with the Full Width at Half Maximum of the peak (FWHM) energy resolution of less than 2% at photon energy of 662 keV. The module's

**87**

**Figure 2.**

*CZT sensor mounted on the UAS platform.*

*Sp*′′

*Gamma Ray Measurements Using Unmanned Aerial Systems*

platform (the DJI S1000+ octocopter) is shown in **Figure 2**.

spectrum using a 60Co gamma-ray source is shown in **Figure 1**.

temperature range is from 0 to 40°C. The CZT sensor mounted onto the UAS

Robot Operating System (ROS) [9] was used for fusion of the data sent from multiple sensors connected to the robotic platform (radiation sensors and a GPS sensor). ROS is an open source tool, consisting of libraries used for robot applications. It allows including a number of independent nodes which communicate to each other, using a publishing and subscribing messages to the topics. Once a data array was sent from the CZT sensor via USB to an onboard Linux minicomputer (the Odroid model) and processed using ROS, a gamma ray spectrum is generated. The example of a measured

The gamma ray spectrum is then analyzed using a code to extract the peaks of interest along with their intensities. This allows for identification of isotopes (sources) that emit these gamma rays, and also for estimation of source strength. Computational algorithms are widely used for photon spectral analytics for various applications [10–14]. For the rapid spectral analysis on board of robotic platforms with a limited computing power, a simple, but robust algorithm is needed. The algorithm based on Mariscotti's technique [15] was developed for the peak identification in the presence of a background in the measured spectrum. In this technique, peaks

Within a small interval of several peak's widths, which is applicable for CZT detectors and scintillation detectors, the spectrum's background is represented as a linear polynomial. In this interval, number of counts in the spectrum channel *<sup>i</sup>* is reconstructed as *Sp*(*i*) <sup>=</sup> *Gauss*(*i*) <sup>+</sup> *<sup>B</sup>*<sup>1</sup> <sup>+</sup> *<sup>i</sup>* <sup>∙</sup> *<sup>B</sup>*2, where *Gauss*(*i*) is a Gaussian function, and constants *B*1 and *B*2 are associated with the background. This method locates a peak

centered at a channel *p*. The value of *σ* is found as *FWHMpeak*/2.355. For a discrete data *y*(*i*), the 2nd derivative of *Sp*(*i*) was approximated as a centered finite-divided difference:

, where *h* is the channel's width*.* Since *y*(*i*) carries statistical errors, the *Sp*(*i*) data will also exhibit statistical fluctuations about expected values at each channel. Hence, a peak cannot be resolved in case of an expected value of *Sp*(*<sup>p</sup>*) which is equal to the standard deviation. For *<sup>σ</sup>* <sup>=</sup> <sup>4</sup> it corresponds to the minimum resolvable *I* value of 600. To resolve much smaller peaks, function *Sp*(*i*) should be 'smoothed', consequently reducing the standard deviation. The smoothing was done by calculating an average of *Sp*(*i*) values over *<sup>w</sup>*

]) includes the height*I* of the peak that is

*DOI: http://dx.doi.org/10.5772/intechopen.82798*

are approximated using a Gaussian function.

The function *Gauss*(*i*) <sup>=</sup> *<sup>I</sup>* <sup>∙</sup> *exp*(−[*<sup>i</sup>* <sup>−</sup> *<sup>p</sup>*]

(*i*) = [ *y*(*i* + 1) − 2 ∙ *y*(*i*) + *y*(*i* − 1)

where the 2nd derivative of the function *Sp*(*i*) is not zero.

]/*h* 2 2 /[2*σ* 2

**Figure 1.** *CZT sensor operation.*

*Use of Gamma Radiation Techniques in Peaceful Applications*

designed as plug-and-play interchangeable modules.

Since UAS has weight and size limitations for onboard components so that the flight efficiency is not reduced due to the payload, the choice of radiation sensors is important. Radiation sensors should be swappable (attached and detached from the robotic platform) in the field conditions. The sensors should have low power requirements so that the integration into the platform does not affect the battery power. The data processing limitations should be also taken into account. To address this necessity, two types of flight ready ambient temperature radiation sensors were developed: a Cadmium Zinc Telluride (CZT) semiconductor sensor for high-resolution gamma spectrometry and the elpasolite scintillation sensor Cs2LiYCl6:Ce3+ (CLYC) for neutron and gamma measurements. Both sensors were

A CZT is a wide bandgap semiconductor [7]. Gamma rays that interact with this detection medium deposit their energy into the CZT operating in a directconversion mode at an ambient temperature. Therefore, no cooling is needed which significantly simplifies the sensor's integration into the robots and their practical use. These semiconductors are able to process more than 10 million photons per

The CZT module GR1-A (Kromek) [8] was integrated into a multicopter aerial

by the CZT proportionally to energy of the incident gamma ray are amplified by a preamplifier and then processed by a shaping amplifier. A 4096 channel analyzer produces the discrete data array of the gamma spectrum that can be processed

The USB interface is used for the data communication as well as power. The power consumption of this module is about 250 milliwatts. The module's volume is

from 30 keV to 3 MeV with the Full Width at Half Maximum of the peak (FWHM) energy resolution of less than 2% at photon energy of 662 keV. The module's

and the weight is about 50 grams. The sensor's energy range is

further. The scheme of the CZT sensor operation is shown in **Figure 1**.

cubic CZT crystal. Electric signals generated

**2. Radiation sensors**

**2.1 CZT sensor**

2.5 × 2.5 × 6.1 cm3

square millimeter per second.

platform. The GR1-A includes a 1 cm3

**86**

**Figure 1.**

*CZT sensor operation.*

temperature range is from 0 to 40°C. The CZT sensor mounted onto the UAS platform (the DJI S1000+ octocopter) is shown in **Figure 2**.

Robot Operating System (ROS) [9] was used for fusion of the data sent from multiple sensors connected to the robotic platform (radiation sensors and a GPS sensor). ROS is an open source tool, consisting of libraries used for robot applications. It allows including a number of independent nodes which communicate to each other, using a publishing and subscribing messages to the topics. Once a data array was sent from the CZT sensor via USB to an onboard Linux minicomputer (the Odroid model) and processed using ROS, a gamma ray spectrum is generated. The example of a measured spectrum using a 60Co gamma-ray source is shown in **Figure 1**.

The gamma ray spectrum is then analyzed using a code to extract the peaks of interest along with their intensities. This allows for identification of isotopes (sources) that emit these gamma rays, and also for estimation of source strength. Computational algorithms are widely used for photon spectral analytics for various applications [10–14]. For the rapid spectral analysis on board of robotic platforms with a limited computing power, a simple, but robust algorithm is needed. The algorithm based on Mariscotti's technique [15] was developed for the peak identification in the presence of a background in the measured spectrum. In this technique, peaks are approximated using a Gaussian function.

Within a small interval of several peak's widths, which is applicable for CZT detectors and scintillation detectors, the spectrum's background is represented as a linear polynomial. In this interval, number of counts in the spectrum channel *<sup>i</sup>* is reconstructed as *Sp*(*i*) <sup>=</sup> *Gauss*(*i*) <sup>+</sup> *<sup>B</sup>*<sup>1</sup> <sup>+</sup> *<sup>i</sup>* <sup>∙</sup> *<sup>B</sup>*2, where *Gauss*(*i*) is a Gaussian function, and constants *B*1 and *B*2 are associated with the background. This method locates a peak where the 2nd derivative of the function *Sp*(*i*) is not zero.

The function *Gauss*(*i*) <sup>=</sup> *<sup>I</sup>* <sup>∙</sup> *exp*(−[*<sup>i</sup>* <sup>−</sup> *<sup>p</sup>*] 2 /[2*σ* 2 ]) includes the height*I* of the peak that is centered at a channel *p*. The value of *σ* is found as *FWHMpeak*/2.355. For a discrete data *y*(*i*), the 2nd derivative of *Sp*(*i*) was approximated as a centered finite-divided difference: *Sp*′′ (*i*) = [ *y*(*i* + 1) − 2 ∙ *y*(*i*) + *y*(*i* − 1) ]/*h* 2 , where *h* is the channel's width*.*

Since *y*(*i*) carries statistical errors, the *Sp*(*i*) data will also exhibit statistical fluctuations about expected values at each channel. Hence, a peak cannot be resolved in case of an expected value of *Sp*(*<sup>p</sup>*) which is equal to the standard deviation. For *<sup>σ</sup>* <sup>=</sup> <sup>4</sup> it corresponds to the minimum resolvable *I* value of 600. To resolve much smaller peaks, function *Sp*(*i*) should be 'smoothed', consequently reducing the standard deviation. The smoothing was done by calculating an average of *Sp*(*i*) values over *<sup>w</sup>*

**Figure 2.** *CZT sensor mounted on the UAS platform.*

adjacent channels near the channel *i*. This smoothing can be repeated *z* times yielding *Si*(*z*,*w*) <sup>=</sup> <sup>∑</sup> *<sup>j</sup> Cij*(*z*,*w*)*yj*, where *Cij* are weighting factors. Its standard deviation is *Di*(*z*,*w*) <sup>=</sup> [<sup>∑</sup> *<sup>j</sup> Cij* 2 (*z*,*w*)*yj*] 1/2 . Values *z* and *w* were determined for the CZT sensor based on the desired *I* value.

This algorithm makes it possible to distinguish photopeaks from Compton shoulders in the spectral data. The peak centroid was found at the center of the Gaussian function. The intensity of the peak was calculated using the area under the Gaussian. The C code implementing this algorithm was written as a function within ROS. The spectrum data array can be processed using this function returning the energies of found peaks and their intensities, as a result minimizing the data transfer from the UAS platform to a ground control station.

#### **2.2 CLYC sensor**

A 2.54-cm diameter, 2.54-cm long cylindrical CLYC [16, 17] crystal was utilized in the design of a gamma-neutron sensor. The CLYC's density is 3.31 g/cm3 . Its scintillation light wavelength range is from 275 to 450 nm (the peak is at 370 nm). The crystal's refractive index is 1.81 at 405 nm. This crystal was coupled with a super bialkali photomultiplier tube matching the CLYC's emission wavelength range, a miniature digitizer, and a high voltage generator; all packaged in a custom housing (**Figure 3a**). The 6 *Li* isotope enrichment of the CLYC was 95%. Neutron detection was achieved via <sup>6</sup> *Li*(*n*,α)*t* reaction. The scintillation light yield of CLYC is 20,000 photons per 1 MeV for gamma rays and 70,000 photons per thermal neutron. CLYC's scintillation properties allow for gamma spectrometry. This sensor operates without cooling. The measured gamma-ray FWHM energy resolution was 5% at 662 keV and 3.3% at 1332 keV.

Moreover, CLYC exhibits neutron-photon pulse shape discrimination (PSD) properties. The photon-induced excitation in the CLYC medium leads to a fast core to valence (CVL) decay and prompt cerium decay with 1 and 50 ns decay constants, respectively. A neutron event in CLYC causes a slow cerium self-trapped excitation (Ce-STE, 1000 ns decay constant) [18].

The digitized neutron and photon signals of the CLYC sensor are shown in **Figure 3b**. Each signal was analyzed using the eMorpho digitizer, generating three values saved in a list mode: the time stamp; the integral under the front of the signal's curve assessed using the partial integration time; and the integral under the entire curve that is proportional to the energy of measured radiation. In order to segregate neutron signals from gamma-ray signals, a PSD value (calculated as ratio of the area under the tail part of the signal to the front part of the signal) was used. Because neutron signals have longer tails, it leads to larger PSD values than in the case of gamma-ray signals. The experimental plot of neutron-gamma PSD for the CLYC sensor using a PuBe source is shown in **Figure 3c**. Neutron and gamma ray signals are well separated in this plot. The neutron signals appear at the value of 3200 keV, electron equivalents (keVee). The figure of merit of neutron-gamma PSD for CLYC was evaluated as 2.3 using an approach described in [19].

#### **2.3 Sensor integration with UAS**

The 'plug-and-play' concept was used to integrate the radiation sensors into the UAS using ROS. This approach supports 'hot swapping' of the sensors into the UAS platform, meaning that the user does not need to set up sensor's parameters each time the UAS is powered on. When the CZT or CLYC sensor is plugged, the operating system recognizes it, and then installs the associated driver to read and process the sensor's data. The data is published and the message appears. Similarly, when the

**89**

**Figure 3.**

*(a) CLYC sensor, (b) CLYC signals, and (c) neutron-gamma PSD plot.*

*Gamma Ray Measurements Using Unmanned Aerial Systems*

*DOI: http://dx.doi.org/10.5772/intechopen.82798*

*Gamma Ray Measurements Using Unmanned Aerial Systems DOI: http://dx.doi.org/10.5772/intechopen.82798*

*Use of Gamma Radiation Techniques in Peaceful Applications*

transfer from the UAS platform to a ground control station.

*Di*(*z*,*w*) <sup>=</sup> [<sup>∑</sup> *<sup>j</sup> Cij*

the desired *I* value.

**2.2 CLYC sensor**

(**Figure 3a**). The 6

was achieved via <sup>6</sup>

662 keV and 3.3% at 1332 keV.

(Ce-STE, 1000 ns decay constant) [18].

**2.3 Sensor integration with UAS**

2 (*z*,*w*)*yj*] 1/2

adjacent channels near the channel *i*. This smoothing can be repeated *z* times yielding *Si*(*z*,*w*) <sup>=</sup> <sup>∑</sup> *<sup>j</sup> Cij*(*z*,*w*)*yj*, where *Cij* are weighting factors. Its standard deviation is

This algorithm makes it possible to distinguish photopeaks from Compton shoulders in the spectral data. The peak centroid was found at the center of the Gaussian function. The intensity of the peak was calculated using the area under the Gaussian. The C code implementing this algorithm was written as a function within ROS. The spectrum data array can be processed using this function returning the energies of found peaks and their intensities, as a result minimizing the data

A 2.54-cm diameter, 2.54-cm long cylindrical CLYC [16, 17] crystal was utilized

*Li* isotope enrichment of the CLYC was 95%. Neutron detection

*Li*(*n*,α)*t* reaction. The scintillation light yield of CLYC is 20,000

scintillation light wavelength range is from 275 to 450 nm (the peak is at 370 nm). The crystal's refractive index is 1.81 at 405 nm. This crystal was coupled with a super bialkali photomultiplier tube matching the CLYC's emission wavelength range, a miniature digitizer, and a high voltage generator; all packaged in a custom housing

photons per 1 MeV for gamma rays and 70,000 photons per thermal neutron. CLYC's scintillation properties allow for gamma spectrometry. This sensor operates without cooling. The measured gamma-ray FWHM energy resolution was 5% at

Moreover, CLYC exhibits neutron-photon pulse shape discrimination (PSD) properties. The photon-induced excitation in the CLYC medium leads to a fast core to valence (CVL) decay and prompt cerium decay with 1 and 50 ns decay constants, respectively. A neutron event in CLYC causes a slow cerium self-trapped excitation

The digitized neutron and photon signals of the CLYC sensor are shown in **Figure 3b**. Each signal was analyzed using the eMorpho digitizer, generating three values saved in a list mode: the time stamp; the integral under the front of the signal's curve assessed using the partial integration time; and the integral under the entire curve that is proportional to the energy of measured radiation. In order to segregate neutron signals from gamma-ray signals, a PSD value (calculated as ratio of the area under the tail part of the signal to the front part of the signal) was used. Because neutron signals have longer tails, it leads to larger PSD values than in the case of gamma-ray signals. The experimental plot of neutron-gamma PSD for the CLYC sensor using a PuBe source is shown in **Figure 3c**. Neutron and gamma ray signals are well separated in this plot. The neutron signals appear at the value of 3200 keV, electron equivalents (keVee). The figure of merit of neutron-gamma PSD

The 'plug-and-play' concept was used to integrate the radiation sensors into the UAS using ROS. This approach supports 'hot swapping' of the sensors into the UAS platform, meaning that the user does not need to set up sensor's parameters each time the UAS is powered on. When the CZT or CLYC sensor is plugged, the operating system recognizes it, and then installs the associated driver to read and process the sensor's data. The data is published and the message appears. Similarly, when the

for CLYC was evaluated as 2.3 using an approach described in [19].

in the design of a gamma-neutron sensor. The CLYC's density is 3.31 g/cm3

. Values *z* and *w* were determined for the CZT sensor based on

. Its

**88**

**Figure 3.** *(a) CLYC sensor, (b) CLYC signals, and (c) neutron-gamma PSD plot.*

#### **Figure 4.**

*Plug-and-play operation scheme for radiation sensors.*

#### **Figure 5.**

*Scheme of the RTK GPS technique.*

sensor is unplugged, the operating system terminates the process and deletes the sensor's driver. The scheme of the plug-and-play operations with CZT and CLYC sensors is shown in **Figure 4**.

The ROS programming environment was also utilized for fusion of the radiation sensor's data with time and position data to enable the spatiotemporal analytics of measured radiation fields. To determine the UAS coordinates at a time of the gamma-ray measurement, the real-time kinematic (RTK) positioning technique was used. This navigation method enhances precision of the position data obtained using satellite positioning systems. Based on the measurements of the phase of the signal's carrier wave, it employs a single reference 'base' station for real-time corrections of the UAS position. The scheme of the RTK GPS technique is shown in **Figure 5**.

A base station contains a fixed Swift Duro RTK GPS receiver, having known coordinates. Multiple UAS platforms carry the RTK GPS and L1/L2 GPS antenna. Correction data are transmitted from the base station to the UAS's RTK Piksi Multi GPS receiver thus allowing calculating the ionospheric error. This method permits raw data measurements with a frequency of 20 Hz. Accuracy of the UAS position estimation using the RTK technique is 10 mm horizontally and 15 mm vertically.

**91**

**Figure 6.**

*based on 1/R<sup>2</sup>*

 *model.*

*Gamma Ray Measurements Using Unmanned Aerial Systems*

**3. Multirobot contour mapping of radiation fields**

with the position data as discussed in the previous section.

It was experimentally verified using a single base station and four UAS platforms. Such precision of position estimation for multiple UAS platforms enables to use the RTK GPS in cooperative multirobot radiation monitoring tasks including contour

One of the approaches of cooperative multirobot sensing is the method of contour mapping of radiation areas. It is based on the use of several UAS platforms (the 'swarm') equipped with radiation sensors. This approach allows for the automatic determination of the contour in space that corresponds to a preset radiation dose. Thus, the multirobot system could locate and follow a boundary of

In this section, the contour mapping algorithm is presented along with a gradient direction estimation and heading angle calculation scheme for the swarm consisting of three UAS that are positioned in a circular formation in the two dimensional space. It is assumed that a gamma-ray sensor, CZT or CLYC, is mounted on each UAS platform. The gamma-ray data are time stamped and merged

The contour mapping is based on two components: the gradient direction estimation and the average radiation level calculated using the radiation measurement data from sensors mounted on the UAS platforms of the swarm. The average of a scalar field is estimated over a circular area of radius *r* centered at a point *c* as shown in **Figure 6**. *Tavg* is the average radiation level calculated using the data from sensors

gamma peak of interest at a point(*x*, *<sup>y</sup>*) by *i*th UAS. The formation center moves toward the increasing (a source-seeking method) or the constant (a contour mapping method) value of the average of sensor readings. To find the required direction

*Formation of three UAS platforms around a circle of radius r. Radiation measurements Tn by three UAS are* 

*<sup>N</sup> Ti*(*x*, *<sup>y</sup>*) \_\_\_\_\_\_\_\_\_\_\_ *<sup>N</sup>* (1)

is the intensity of the measured

*DOI: http://dx.doi.org/10.5772/intechopen.82798*

mapping and source search.

the hazardous zone.

**3.1 Gradient direction estimation**

of UASs flying in a circular formation:

*Tavg*(*x*, *<sup>y</sup>*) <sup>=</sup> <sup>∑</sup>*i*=1

Here, *N* is the number of UAS platforms, *Ti*

*Use of Gamma Radiation Techniques in Peaceful Applications*

*Plug-and-play operation scheme for radiation sensors.*

**90**

**Figure 5.**

**Figure 4.**

*Scheme of the RTK GPS technique.*

is shown in **Figure 5**.

CLYC sensors is shown in **Figure 4**.

sensor is unplugged, the operating system terminates the process and deletes the sensor's driver. The scheme of the plug-and-play operations with CZT and

The ROS programming environment was also utilized for fusion of the radiation sensor's data with time and position data to enable the spatiotemporal analytics of measured radiation fields. To determine the UAS coordinates at a time of the gamma-ray measurement, the real-time kinematic (RTK) positioning technique was used. This navigation method enhances precision of the position data obtained using satellite positioning systems. Based on the measurements of the phase of the signal's carrier wave, it employs a single reference 'base' station for real-time corrections of the UAS position. The scheme of the RTK GPS technique

A base station contains a fixed Swift Duro RTK GPS receiver, having known coordinates. Multiple UAS platforms carry the RTK GPS and L1/L2 GPS antenna. Correction data are transmitted from the base station to the UAS's RTK Piksi Multi GPS receiver thus allowing calculating the ionospheric error. This method permits raw data measurements with a frequency of 20 Hz. Accuracy of the UAS position estimation using the RTK technique is 10 mm horizontally and 15 mm vertically.

It was experimentally verified using a single base station and four UAS platforms. Such precision of position estimation for multiple UAS platforms enables to use the RTK GPS in cooperative multirobot radiation monitoring tasks including contour mapping and source search.
