The trend of biomimetic underwater robots has emerged as a search for an alternative to traditional propeller-driven underwater vehicles. The drive of this trend, as in any other areas of bioinspired and biomimetic robotics, is the belief that exploiting solutions that evolution has already optimized leads to more advanced technologies and devices.
In underwater robotics, bioinspired design is expected to offer more energy-efficient, highly maneuverable, agile, robust, and stable underwater robots. The 30,000 fish species have inspired roboticists to mimic tuna, rays, boxfish, eels, and others.
The development of the first commercialized fish robot Ghostswimmer by Boston Engineering and the development of fish robots for field trials with specific applications in mind mark a new degree of maturity of this engineering discipline after decades of laboratory trials.
METHOD OF BIOINSPIRED ROBOT DESIGN
The biological data were obtained from both the biology literature and experiments. The model animal used for the biological experiments was a rainbow trout. Fish behavior was recorded in a controlled hydrodynamic environment with a high-speed video camera and a digital particle image velocimetry (DPIV) system (Figure 1). Various flow conditions were investigated to measure the response of animals to variations in the sizes and strengths of wakes.
Canal neuromasts are situated in the canals under the skin, and each of them measures the pressure difference between adjacent points where the canal emerges at the surface of the skin. This project has created several variations of two types of lateral line systems. The first type is based on microelectrome-chanical system (MEMS) stress-driven nitride-based bilayer design (Figure 3), equipped by a strain gauge.
Figure 4. (a) The pressure sensors mounted on circuit boards with onboard electronics. The pressure sensors of an artificial lateral line use MS5407-AM diver’s watch sensors by Intersema Sensoric SA. The sensing unit is connected as a Wheatstone bridge to give the sensor a high sensitivity of 56 mV/bar in the full scale (0–7 bar). We are using a 22-b differential analog-to-digital converter (ADC) with 124.5-mV reference voltage so that we can measure pressure with a least significant bit (LSB) of about (0.106 Pa). (b) The schematic of a fish robot prototype with a 3-D pressure-sensing lateral line.
Figure 5. The KVS. (a) Schematics m is the wavelength of the KVS. (b) A snapshot of a DPIV image of the flow obtained during fluid dynamics experiments. This data gets analyzed using (http://www.mathworks.com/matlabcentral/fileexchange/37323) and is an input for (c) and (d). (c) Instantaneous vorticity obtained from the DPIV image. The blue and red regions show high vorticity in opposite directions. The plot is obtained from two consequent DPIV snapshots, and a Gaussian filter is applied to smoothen the plot. (d) Velocity readings averaged more than 10 s (500 frames obtained with 50-Hz frequency): 1) suction zone and 2) reduced flow zone.
Figure 6. The pressure cues recorded from a robotic platform immersed in uniform flows (red-filled circle) and vortex streets (green-filled circle). (a) In vortex streets, most of the pressure signals were dominated by the vortex shedding frequency. The number of sensors detecting vortex shedding frequency decreased gradually when the robot was moved away from the vortex street.
In contrast, in uniform flows, each pressure measurement had a different frequency with maximum amplitude, so there was no agreement among sensors. (b) The pressure difference between nose and side sensors is distinct between uniform flows and vortex streets. (c) Through an analysis of the turbulence intensity and amplitude of the dominant frequency, we were able to estimate the position of the robot with respect to the cylinder unambiguously. (d) The amplitude of the pressure difference between the right and left sides of the robot was linearly correlated to the robot’s orientation with respect to the oncoming flow. The slope of the lines was different in uniform flows and vortex streets.
FLOW-AIDED CONTROL AND NAVIGATIONThe experiments of flow-aided control of the FILOSE robot are conducted in uniform flow and in KVS in a flow tank, where the flow and the trajectories of the robot are recorded (Figure 7). The experimental setup is described in Figure 8. Experiments are conducted in a flow tunnel with a 0.5-m wide, 0.5-m high, and 1.5-m long working section. The robot is freely swimming but its motion is limited to two dimensions to permit trajectory tracking and motion analyses using an overview camera. The following experiments of flow-aided control were conducted.
The tail bends against the high-pressure zone created by the vortex, and the robot takes advantage of the increased perpendicular component of the lift force created by pressure differences on both sides of the flexing tail (Figure 11). The results are compared with those of the tail fin propulsion in the steady flow with the same incoming flow speed. In comparison, 100% more thrust is created in KVS with the appropriate tail beat timing than in steady flow.
Though all 30,000 fish species have a lateral line organ, so far, there have been no technological counterparts to lateral line sensing in use for controlling underwater robots. The contribution of this article is to give a new sense, svenning, to aid the control of underwater vehicles. Once flow can be perceived, it can be analyzed and exploited for a variety of purposes. Traditionally, flow is treated as a disturbance in underwater robotics, to be compensated by the vehicle’s control algorithms. With flow sensing, flow becomes a source of information, and, with clever sensing–actuation coupling, flow becomes a source of energy. Flow information can be fused with other sensor modalities and used for vehicle control. Knowing flow direction and strength permits movement with respect to the flow-relative reference frame as opposed to the global reference frame.
Flow information can also be incorporated into higher- level behaviors. Salient flow features, such as wakes of objects or steady currents, could be identified and classified and used as landmarks. Again, this information can be fed into a vehicle’s navigation algorithm and used for map building and localization. Flow sensing permits identification of flow conditions where a vehicle’s control is more stable and energy efficient. Coupling of flow sensing and actuation opens up new opportunities to exploit flow for energy-efficient motion. Flow-relative control could also make traditional rigid hull underwater robots more efficient, but there is more to gain from flow perception and robot–fluid interaction from small devices, with flexible fins or rudders.
This aligns flow-sensing robots with the increasingly popular trend of soft robotics. Hydrodynamic forces can be best exploited if the craft is continuous, flexible, and able to vary its stiffness to adapt to different conditions. For example, our FILOSE robot would need a stiffer tail to produce enough thrust in uniform flow and a floppy passive tail to bend in KVS between the vortices. Designing such a robot can be a challenging problem for mechanical engineering.
Authors: Otar Akanyeti | Jennifer C. Brown | Lily D. Chambers | Hadi el Daou | Maria-Camilla Fiazza | Paolo Fiorini | Jaas Ježov | David S. Jung, Maarja Kruusmaa | Madis Listak | Andrew Liszewski | Jacqueline L. Maud | William M. Megill | Lorenzo Rossi | Antonio Qualtieri | Francesco Rizzi | Taavi Salumäe | Gert Toming | Roberto Venturelli | Francesco Visentin | Massimo De Vittorio