For underwater vehicles to successfully detect and navigate turbulent flows, sensing the fluid interactions that occur is required. Fish possess a unique sensory organ called the lateral line. Sensory units called neuromasts are distributed over their body, and provide fish with flow-related information. In this study, a three-dimensional fish-shaped head, instrumented with pressure sensors, was used to investigate the pressure signals for relevant hydrodynamic stimuli to an artificial lateral line system. Unsteady wakes were sensed with the objective to detect the edges of the hydrodynamic trail and then explore and characterize the periodicity of the vorticity. The investigated wakes (Kármán vortex streets) were formed behind a range of cylinder diameter sizes (2.5, 4.5 and 10 cm) and flow velocities (9.9, 19.6 and 26.1 cm s−1). Results highlight that moving in the flow is advantageous to characterize the flow environment when compared with static analysis. The pressure difference from foremost to side sensors in the frontal plane provides us a useful measure of transition from steady to unsteady flow. The vortex shedding frequency (VSF) and its magnitude can be used to differentiate the source size and flow speed. Moreover, the distribution of the sensing array vertically as well as the laterally allows the Kármán vortex paired vortices to be detected in the pressure signal as twice the VSF.
Complex underwater environments such as rivers or the near shore coastal zone present significant navigation challenges to submersible vehicles. Navigation in complex underwater environments is a key goal for ocean technologies. Flow fluctuations can disrupt the trajectory of a craft or an animal. Turbulence at different length scales, such as on a global, oceanic scale in the formation of gyres or at a local, riverine scale as small eddies, will impact an object in different ways. Detecting and reacting to a flow perturbation requires an online sensing strategy that will enable real-time adaptation to the changing environment.
For an underwater robot to successfully characterize the flow, it needs to measure and compute the source of any flow fluctuations, the size of the source, its own distance downstream and the relative size and timing of any flow fluctuations. Flow sensors could provide measured information on the environment, on which decisions for reactions could be based.
Some swimming animals have a flow sensory system and therefore provide a useful model of what is required for successful navigation by an autonomous underwater vehicle. Fish have a lateral line array, where each sensory mechanoreceptor unit, called a neuromast, is distributed over the surface of their bodies, in a lateral line as well as over the body as an array of sensors, to help them navigate their environments [1–3]. The network of neuromasts is both surface (superficial neuromasts) and subsurface (canal neuromasts) and are for the most part innervated by an anterior and a posterior nerve. The surface neuromasts are sensitive to flow velocities, and the canal neuromasts are sensitive to the pressure difference across pore openings that connect the canal network to the fish–fluid interface. Each neuromast has an axis of increased sensitivity providing directionality. On some fish, these directions are largely composed of rostral–caudal and dorsal–ventral alignment . The geometry of the fish modifies and redistributes the flow over the sensors on its body. Previous biological investigations have hypothesized that the more variable orientations of superficial lateral line sensors observed at the head of the fish are present to sense the more varied flows at those regions . Sensors at the nose region directly in front of the eye can sense dorsal–ventral flows, such as those generated by vertical movements of the fish or shed vortices . Bioinspired underwater robotics is one field of research that can address robotic navigational challenges and provide us insights into how animals achieve successful aquatic navigation [5–9].
Using the fish lateral line as bioinspiration, a flow sensory array was implemented for a craft with the aim of characterizing a complex flow environment. If steady flow and unsteady flow can be distinguished using a sensory array, then adjustments can be implemented to provide a locomotive strategy. To further our knowledge of this three-dimensional fluid–structure interaction, data on fluid motion need to be coupled with three-dimensional sensing.
To achieve this, pressure sensors were distributed on a movable fusiform head, in three dimensions around the nose. This was used to further understand and build algorithms from the on-board flow analysis, to achieve the goals of:
(1) source localization,
(2) determining size of the source and
(3) source wake characterization.
The periodic shedding of vortices behind a cylinder placed in uniform flow, called a Kármán vortex street (KVS), provides a repeatable, unsteady flow in a laboratory setting. The sensing of the turbulent wake behind a cylinder using a virtual sensor has been described previously by Akanyeti et al. , and object identification using pressure sensors has been completed for static arrays [11–13]. Further wake navigation algorithms have been previously developed based on the predicted pressure signal . Research by Venturelli et al.  sampled a KVS using two symmetric linear lateral lines running on either side of a three-dimensional ‘boat-shaped’ craft, as an analogue of the posterior lateral line. Characterization of the KVS was performed using calculations of turbulence intensity and vortex shedding frequency (VSF) and by measuring the magnitude of the fluctuations present in the sensor readings. Vortex travelling speed using cross-correlation among the sensors was also calculated . In this study, a fish-like craft is investigated, with sensors distributed at the nose, along the sides and up the front of the craft. The sensors are not limited to a two-dimensional horizontal array and extend up into the vertical plane. Sensing in three dimensions using an artificial lateral line has been previously demonstrated on a static cylinder localizing a dipole source ; in this research, dynamic sensing is investigated.
To recreate paired vortices in the laboratory, the flow investigated is a steady-free stream flow around a cylinder. The interaction of the cylinder wake impinging on the moving craft is investigated; in particular, the dependence of the hydrodynamic pressure signal on the direction to and relative distance from the object. By tracking across and through the wake of a cylinder, a new approach to unsteady flow sensing at a scale relevant to our underwater craft is detailed. In this study, we provide a data-driven methodology, where the pressure signal in three dimensions can be simultaneously recorded while moving. By logging the sensed pressure signal on our craft, we can determine our position in the wake. We use the measured pressure signal to distinguish between nine wakes using a range of cylinder sizes and flow speeds. A range of flow speeds and cylinder sizes were used to explore the sensed pressure signal to detect which cylinder size and what flow velocity the craft was in. The three-dimensional distribution of the pressure sensors provides us a unique view of the flow field from within the fish-like head.
2.1. Craft and sensor array
An array of 33 pressure taps was distributed over the surface of the head. Sensors were connected to the taps using stiff rubber tubing. The pressure sensors (MS5401-AM, Measurement Specialties) had a 240 mV = 1 bar sensitivity and 0 to 1 bar full-scale range. The sensors were calibrated in still water, at a fixed depth for these experiments with the craft, and had 2 Pa accuracy. The data acquisition of the pressure signal is described in further detail in Akanyeti et al. . For these experiments, a three-dimensional configuration with 16 sensors was used at a sampling rate of 200 Hz.
The distribution of pressure taps was used to provide an array mimicking a natural system. The biological distribution of the goldfish lateral line was thoroughly measured by Schmitz et al.  (figure 1a,b). The sub- and supraorbital canal pores were used to generate a simplified representation of their distribution for the model (figure 1c). To recreate this in a sensing array, sensors were distributed in three dimensions as a linear (rostral–caudal) array in the x–y plane and out into the vertical, dorsal–ventral (z-plane). There is increased evidence that due to the geometry of the fish body the pressure gradient is greatest at the front of the fish [17–19]. This is also supported by the increased variance in the axis of sensitivity for the surface neuromasts around the head of the fish . We positioned our sensors around the tip of our craft to mimic natural systems and measure pressure signals from local water flow.
2.2. Craft movement
The craft consisted of a fish-like head made of polyamide plastic using a three-dimensional printer (figure 2). The head was attached to a watertight metal housing which contained the sensors and electronics. The housing was attached using a 6 mm rod to a linear motor (MT480P Aerotech Inc.). This was mounted above the flow tank and the length of the rod was selected to ensure the craft was placed centrally within the water column with the front tip of the craft positioned at 13 cm depth.
The linear motor moved the craft three times with a periodic motion at 0.02 Hz, both laterally across the tank (y-direction) and longitudinally up and down stream in the tank (x-direction). The amplitude of lateral and longitudinal motion was 18 and 12 cm, respectively (figure 3). This corresponds to an average of 0.72 cm s−1 velocity for the lateral path and 0.48 cm s−1 for the longitudinal path. The front tip of the craft was positioned three cylinder diameter lengths downstream. The sensor signal while moving in uniform 19.6 cm s−1 flow was also investigated.
2.3. The fluid environment
The testing environment was an open-topped recirculating tank. The working section (100 × 40 × 36 cm3) was bounded by collimators to smooth the flow generated by a propeller. The air–water interface of this open-channel design did generate surface waves with microturbulence evident at the higher freestream velocities. Although for engineering purposes, this introduces a more complex environment by increasing the noise level of the background signal, from a biological perspective, it is more indicative of real-world conditions.
The wake behind a cylinder or other bluff body object in uniform flow is termed a KVS when it is composed of paired point vortices which are shed sequentially in alternating rotation into the wake at a frequency relative to the flow speed and the size and shape of the object. A KVS is one fluid phenomenon that generates a flow which has repeatable, measurable characteristics in a laboratory setting; this is why it was chosen as the fluid regime to create an unsteady environment in which to test our craft. The relationship between the VSF, flow velocity (U), and cylinder diameter (d) is given in equation (2.1), where for the majority of cases, the Strouhal number is equal to 0.2. 2.1
A series of cylinders with diameters of 2.5, 4.5 and 10 cm and a series of flow velocities of 9.2, 18.6 and 26.1 cm s−1 were used to generate nine KVS in total. Figure 3 shows an example of a wake field measured using digital particle image velocimetry (DPIV).
A snapshot of the velocity magnitude and vorticity plots of the craft in the KVS is shown in figure 4. The two-dimensional rotational elements in the flow are discrete and are reflected in the velocity vector plot as a region of fast flow.
2.4. Measured parameters
Turbulence intensity, dominant frequency and magnitude were calculated from the on-board pressure data. The raw pressure data were filtered to remove noise originating from sources in the environment or from the craft and sensors themselves. A band-pass filter in the range 0.1–5 Hz was applied. The speed of motion was selected to be below the distortions generated by self-motion.
The turbulence intensity was calculated as the ratio of the standard deviation of the sensor reading to the mean value over a 0.2 s window. This short time window of 0.2 s was chosen for two reasons: to have a fast response time reducing delays and to reduce the average effect over the motion and thus focus analysis on localized turbulence.
In , it was shown that nose sensor Pnose minus the mean of the side sensors Pside provides a unique signature to detect flow velocity in steady flows. Owing to the geometry of the fish head, the pressure is greater at the nose than at the sides, creating a pressure difference that varies with flow velocity. Here, the measure of nose sensor pressure minus the side sensor pressures is investigated to detect variance in a KVS while moving. The self-motion signal of the craft was analysed by a series of previous tests of the same craft in still and steady flow , where the craft oscillated along the longitudinal path. When the craft moved at a slow constant speed (less than 5.4 cm s−1), it was found to be the best to detect external sources of flow . In this experiment, the craft moved at less than 0.5 cm s−1 for the corresponding longitudinal path (figure 3, motion path 2).
The VSF was computed by analysing the frequency spectrum of each pressure recording. The frequency with the highest amplitude was extracted to determine the VSF and its amplitude. The estimated VSF was then compared with its theoretical value as predicted by equation (2.1). In Venturelli et al. , it has been shown that VSF can be detected more robustly by analysing the pressure difference across the craft (pressure on the right-hand side minus the pressure on the left-hand side). These pressure difference data were used in our analysis of three discrete points in the KVS, the left-hand side, centre and the right-hand side.
A vehicle or animal navigating a flow does not have access to the world view of the flow, as it is limited to the sensors it can carry with it. Here, we investigate how craft motion affects a sensor's ability to identify local flow conditions. To identify vortex streets from steady flows, to detect flow features within the vortex street and to estimate wake source parameters (cylinder size and flow speed). The results presented in §§3.1 and 3.2 were obtained in a vortex street behind a 4.5 cm cylinder in a 19.6 cm s−1 flow. The electronic supplementary material provides the pressure profile and graphs of the pressure data for each of the 16 sensors over the range of craft motions. Owing to the dimensions of the experimental set-up, this flow and cylinder size provided a wake width which the craft could navigate into and out of in our flow tank. Characterization with DPIV was used to gain a global sensing perspective. Here, it is demonstrated how these measurements can be recovered using on-board sensors.
3.1. Crossing a wake: boundaries of steady and unsteady flow
Subtracting the mean of all the side pressure signals (Pside) from the nose pressure signal (Pnose) provides a clear instantaneous signal to identify instances when the head moves through the cylinder wake corridor. Figure 5a shows the lateral motion profile of the craft with respect to the centre of the KVS. The wake characteristics were verified using DPIV, with a calculated wake width of 6.7 ± 0.5 (s.d.) cm for a 4.5 cm cylinder in 19.6 cm s−1 flow (as outlined by dotted lines in figure 5a). The pressure difference between the nose and side sensors decreased as the craft was moved laterally through the KVS (figure 5b), reaching its minimum at the centre of the street. This occurred six times in agreement with the three lateral oscillations (figure 5a). The duration of the pressure drop at the nose can be related to the lateral speed and width of the KVS. In freestream, flow pressure was higher on the nose than the sides, but there was minimal pressure difference between the nose and side sensors when in a KVS (figure 5b). By comparing figure 5a and b, it can be seen that the pressure gradient is negative in line with the craft's position at the centre of its movement, which is within the wake as calculated using DPIV.
The turbulence intensity is approximately one order of magnitude larger in the KVS wake then outside of it (figure 6). When the head was drawn laterally through the wake behind a cylinder, there were substantial changes in the turbulence intensity which corresponded to the crafts position behind the cylinder (figure 6). The motion of the craft was repeated across the tank in a steady-free stream flow to analyse the effects, if any, of the motion. As expected, there was no change in the turbulence intensity at any of the sensors under steady flow conditions (figure 7).
Figure 8 shows the pressure measurements at the extremes of this motion. The change in pressure on the nose sensor (sensor ID no. 5, at centre position in figure 8) was clear as the craft moved from the high pressure region of the freestream (left-hand side of the wake) to the low pressure region directly behind the cylinder. The subsequent increase in the turbulence intensity provides the inverse picture as the head exits the wake on the opposite right-hand side.
Turbulence detection is important to track a wake but to be able to exploit features in the flow the periodic nature of a KVS needs to be sensed. As evident in figure 6, the craft detected the highest turbulence intensity at the centre of the KVS. If the craft remains positioned in the centre of the street for a finite time, then the VSF can be accurately identified. The craft may need to maintain position in the street to detect the VSF, because the frequency information of the flow features behind a cylinder are in the order of a few hertz or less; to allow a minimum number of events to elapse in order to achieve accurate data analysis. Active sensing by crossing the wake allows the craft to detect the wake width. This highlights the additional information available using dynamic sensing at this craft speed. The goal for vehicle autonomy is to move towards real-time measurements. The measure of pressure difference between the sensor at the nose and the side sensors requires a minimal time interval for calculation; this measure can then be used, in near real-time, to detect a KVS when the craft is crossing a turbulent wake.
3.2. Inside the wake: distance downstream
The successful identification of the KVS and the suction zone was possible using turbulence intensity, VSF and pressure difference. To demonstrate this, we moved the craft longitudinally, towards and away from the cylinder. Figure 9a highlights the distance travelled and the timing of the motion. In this scenario, as the craft is kept within the KVS throughout its movement, the pressure at the nose minus the mean of the other sensors (figure 9b) does not create a difference as it did in the previous translational motion (figure 5b), and the values remain consistently negative.
Using the same turbulence intensity calculation for the longitudinal motion confirms it as a strong indicator of position within the cylinder wake. Turbulence intensity decreases each time the craft approaches the cylinder and enters the suction zone (figure 10). When in the suction zone, only the sensors at the front of the craft decrease their turbulence while the others on the sides (sensors 1-3 and 7-9) maintain a significantly high value when compared with that within a free stream flow (figure 7).
As the craft moves towards the cylinder, a characteristic decrease in the pressure difference between the left and right sides (figure 11b) and, by consequence a decrease in the normalized magnitude of the VSF signal, occurs (figure 11d). The pressure difference decreases between the left and right sides as a result of the craft moving out of the region with fluctuating vortices and into the region directly behind the cylinder. The VSF (0.88 ± 0.17 Hz) for a 4.5 cm cylinder in 19.6 cm s−1 flow was detected throughout the whole longitudinal motion when the craft neared the cylinder but as it moved closer, reducing the distance between the nose and the cylinder edge a reduction in the normalized VSF magnitude was evident (figure 11d).
3.3. Identifying a wake and the size of its source
By analysing the frequency content of the flow we were able to identify a unique hydrodynamic signature peculiar to vortex streets. As the craft enters the corridor formed between the vortices, it was able to detect the alternating vortices. The side sensors (sensor ID nos 1–4, 6–9) could detect only the vortices on one side of the street as vortices from the other side were obscured by the presence of the craft. Therefore, the side sensors detect the VSF. By contrast, the foremost sensor (sensor ID no. 5), at the tip of the craft detected twice the VSF (2VSF) as it was detecting both vortices alternating on each side of the street within the same period (figure 12a). This distinction was also evident when moving from the horizontal plane into the vertical plane; sensor 13 detected 2VSF, whereas sensors 10–12 and 14–16 detected VSF. We demonstrate that by analysing the frequency content of the pressure readings we were able to create a clearer map of the stimuli than by just using the average pressure readings (figure 12b,c).
To identify the size of the source, the analysis was extended to different flow speeds and cylinder sizes. It was observed that VSF, both measured and calculated using equation (2.1), were not sufficient to map each cylinder wake individually (figure 13). The same VSF value could correspond to a number of different combinations of flow speed and cylinder size; the VSF remained constant, for instance, between 2.5 cm diameter cylinder at 9.9 cm s−1, 4.5 cm at 19.6 cm s−1 and 10 cm diameter cylinder at 26.1 cm s−1 (figure 13).
To resolve this ambiguity, two methods are proposed. The first method is based on estimating the width of the wake as it is determined by the diameter of the cylinder. When the craft was in the vortex street two distinct patterns emerged in the pressure recordings; low pressure difference between the nose and side sensors and a high turbulence intensity. Knowing the average craft's speed to cross the tank (0.72 cm s−1) and the duration of the reduced pressure difference and high turbulence intensity signals (average 10 s), then the wake width was estimated to be 7.2 cm. This quick approximation is close to the DPIV measured wake width of 6.7 cm. Once the wake width was estimated, the cylinder diameter was determined by using the diameter–wake width relation and the flow velocity estimated using equation (2.1).
The second approach was to incorporate the amplitude (the magnitude of the Fourier transform) of the dominant frequency into our analysis. We observed that the strength of the vortices increased as the flow speed and the cylinder size increased. This was reflected in the strength of the VSF and 2VSF signals plotted as amplitude (figure 14). By using the amplitude of the signal coupled with the VSF, the object size can be detected more accurately. However, it is important to note that the strength of the vortices also varies within the street. The strength of the vortices is usually highest just after the formation zone and then decrease gradually as the vortices travel downstream. To accurately determine the cylinder diameter using this approach, the craft sensors should record at a fixed position downstream from the cylinder, such as the vortex formation zone.
Figure 15 highlights the amplitude of VSF and 2VSF signals at three different locations in the vortex street; one, two and three diameter downstream from the cylinder. It is clear that for all three distances, sensor ID nos. 4 and 6 were maximally sensitive to the VSF amplitude. The sensitivity decreased gradually towards the side sensor. As the array neared the cylinder from three diameters to two diameters downstream, there was an initial increase in the magnitude of the VSF, suggestive of stronger vortices just beyond the vortex formation zone. Closer to the cylinder, at one diameter downstream, the magnitude across all sensors dissipates as the array enters the suction zone.
The picture is different with regard to the 2VSF signal. The nose sensor (sensor ID no. 5) detects the highest signal level (figure 15b) which attenuates along the sensors distributed further along the craft head. In contrast to the VSF magnitude signal, the highest 2VSF magnitudes are detected at the nose sensor and at the sensor in the vertical dimension directly above (sensor ID no. 13) in figure 12a. The ability of the vertical sensors to also detect the 2VSF signal highlights their use as a unique flow feature indentifier for the KVS. The laterally distributed sensors follow a similar pattern of magnitude detection as for the VSF detection (figure 15a,b). However, on the nose, the highest magnitude is at three dimensions, reduced at two dimensions and finally is undetectable at one dimension. The minimal detection of vortices in one dimension is because there are no fully formed vortices 4.5 cm downstream from the cylinder.
4.1. Geometry of craft and sensor array
In engineering, the majority of fluid–fusiform structure interactions have been considered for fins and aerofoils with respect to their force interactions [20–22]. Little is known about the sensory aspect of the interaction. In this research, we have used a three-dimensional fish-like head in a complex fluid environment to sense the flow. This geometry approaches the more realistic configuration of a craft and provides a closer mimic of the conditions experienced by a fish. It also allows us to investigate the flow signal on a three-dimensional object and any unique signal aspects of this configuration. Using an artificial lateral line composed of a pressure array, we have determined that the edge of a KVS wake can be identified while in motion. Signal processing of the pressure signal over the head, distributed in the vertical as well as horizontal plane, provides an analysis of what is available for the craft to sense in a fluid volume and what a fish mechanoreceptor array may be capable of detecting.
Static sensing is limited, because a reference point is required in order to characterize unsteady wakes. Allowing the craft to move and sense, as shown here, highlights that dynamic sensing increases the sensory range. The motion and geometry of the craft itself alters the available signals and so can add complexity. To account for this the craft motion was very slow in this study, to increase the motion speed a comprehensive analysis of self-generated signals is required, as in , where the inertial effects of a craft were found to dominate below 0.2 l s−1 and the velocity effects dominate above this. Using an on-board pressure-guided exploration method, we identified hydrodynamic cues for discriminating steady and unsteady flow boundaries while in motion. The use of a single lateral pressure sensor array to aid navigation has been previously trialled on an untethered robot . However, the coupling of pressure sensing in the vertical as well as horizontal plane while in motion, to characterize an unsteady flow has, to the best of the authors’ knowledge, not yet been realized. The sensor configuration and motion, investigated here, may be what is required to more successfully mimic the biological behaviour of fish and to exploit the vortices in the flow.
For navigation of a wake to be successful, using the strategy highlighted in this research, of the foremost sensor minus the mean of the sides, the head must be orientated directly into the flow at all times. In this investigation, linear motion was investigated. The difficulties of realizing a rotating and translating craft such as the actuated fish–robot or fish will need to be overcome in the development of a future free-swimming craft.
The importance of the third dimension to our understanding of fish propulsion has recently begun to be appreciated ; our work now brings a similar three-dimensionality to on-board flow sensing. Here, we used vertically distributed sensors on a three-dimensional craft to examine what is usually studied as a lateral two-dimensional wake, namely the KVS. The three-dimensional craft in this study was rigidly mounted in the centre of the water column, restricting the motion to a two-dimensional plane and the flow was visualized using DPIV. The shape of the craft or animal and its movements will affect the hydrodynamics in the wake of the cylinder. It is important to remember that the craft itself redistributes the cylinder wake and its energy, and it is for this reason that the in situ measures of pressure fluctuations are critical. In this work, experimental results were used to determine the hydrodynamic features in a KVS and their corresponding frequency amplitude. The ability of the sensors placed vertically on the craft to detect hydrodynamic features either side of the craft such as the individual vortices in a shed pair as a 2VSF signal is of direct use for future head shape design and placement of the sensors in the vertical dimension.
4.2. Biological relevance
The localization of an unsteady flow source in nature is key for predator avoidance and prey capture. As well as flow itself, the content of the flow, such as food or chemical cues, may have its tracking further complicated by turbulence. Continuing advancements in the field of robotics are extending our reach into more complex environments where the difficulty in tracking wakes and localizing sources (such as velocity, pressure, temperature or chemical signatures) increases.
Fish are known to react to fluctuations in the flow around them, and their interaction with the flow is mediated to a large extent by the lateral line. This mechanosensory array has been shown to respond to fluctuations in the flow compared with bulk water flow [2,3]. One reason to investigate the wake of a cylinder is inspired from gait analysis of fish behind cylinders in flow [3,25,26]. The ability of a fish to maintain its downstream position relative to a cylinder, or on the edge of a cylinder wake, with little or no muscle activity or body movement highlights an interesting phenomenon. The fish is able to use the periodic vortices to establish a stable trajectory [25–27]. It is suggested here that to achieve or mimic this wake interaction with a craft we must first provide a fish perspective of the flow. This would enable research into the effects of fish shape and movements on pressure sensory input and begin to extend experiments on biological animals .
The stimulus identified here, for complex flow of the VSF and VSF magnitude, are of interest from the biological perspective. Vortices are shed in the wake of swimming animals as well as static bluff bodies. In the case of vortices shed in a KVS, these parameters provide information on the flow which can be used to understand why a swimming fish prefers a certain downstream position from the cylinder. The increase in VSF and 2VSF signal at a distance twice the cylinder diameter suggests an optimum position to detect the periodicity in a KVS while a distance of one cylinder diameter is too close. The reduced frequency component on the head as it nears the suction zone would leave the posterior of the body interacting with a wake that was undetected on the head.
The distribution of the lateral line on fish is in three dimensions. The development of the lateral line on a fish is such that there is no single line of sensors at the front of the animal, traversing the vertical direction as in the craft. However, the ability of the pressure sensors at the front of the craft to easily detect the 2VSF suggests that a fish pivoting its head in the flow could allow the sensors closest to the nose to orientate directly into the flow and detect a similar paired vortex signal. In this study, the vertical, dorsally orientated sensors above the nose were sampled, but similarly those orientated ventrally are of interest too and may be of use to the animal. Research into three-dimensional spatial cognition by fish has indicated that information in the vertical axis can override landmark cues in the horizontal plane . A previous theory as to how fish determine their depth is by using the rate of change of pressure, i.e. hydrostatic pressure and the fish swim bladder [29,30]. Here, the use of pressure sensors, although not a true mimic of a lateral line network which detects a pressure difference, provides an additional possibility for the animal to navigate volumes by the change in the flow frequencies and how this varies as it is distributed over their head. This work highlights the importance of the vertically placed sensors for characterizing unsteady flow.
Biomimetic robotics offers a controlled platform to use in biological investigations [5,31–33]. The use of robotic tools such as this craft to investigate the sensing map available to an animal may help to better understand the optimum sensors configuration for turbulence detection. To better understand the ecological relevance of lateral line position on biological fish, the local hydrodynamics should be mapped. This study suggests that further research is required into the role that increased head angles have on fish navigation in turbulent environments, such as within a KVS  and other behavioural features from a sensing perspective.
The successful identification of a KVS while in motion could potentially be used to control an underwater vehicle position relative to structures in the flow. Here, we used the instantaneous pressure drop from the nose to surrounding sensors as a cue for when the craft crosses into a turbulent region and demonstrate that this could be achieved for slow craft velocities. This was verified using a short time window (0.2 s) to measure turbulence intensity. The peak turbulence intensity at the centre of the street is clearly visible from the pressure data. The periodicity of the flow was analysed by looking at the VSF and its magnitude on the pressure sensors. The drop in the magnitude of the VSF is a powerful indicator of the sensor array position in the suction zone and vortex formation zone directly behind the cylinder. This can be used to obtain the position of the craft downstream of the cylinder. The VSF magnitude was identified in this study as a unique predictor for KVS identification. We highlight the doubling of the VSF signal, which is a characteristic of a paired vortex wake seen within a KVS or a reverse KVS such as that generated in the wake of a swimming fish. The VSF has increased importance for navigation when exploring a fluid volume. Here, it has been highlighted that the most sensitive sensors to the VSF are those located at the front of the craft, particularly at either side of the nose. The ability of the sensors in the vertical dimension to detect individual vortices of a shed pair has impact on the placement of sensors and provides further evidence that sensory information in the vertical as well as the horizontal plane is important for unsteady flow characterization.
The turbulence in active flow regimes such as the KVS creates unique control and propulsive challenges. This investigation has pressure mapped the local hydrodynamics from a fish perspective highlighting the importance of the orientation of an artificial lateral line array to detect flow features in unsteady flows.
Funding for this research was made available by the FILOSE project, supported by the European Union, Seventh Framework Programme (FP7-ICT-2007-3).
We thank Steve Dolan and Nathan Sells for help with the design and build of the flow tunnel, Ryan Ladd for the linear motor apparatus and Andres Ernits for the pressure sensor array build.
- Received May 3, 2014.
- Accepted July 7, 2014.
- © 2014 The Author(s) Published by the Royal Society. All rights reserved.