| Most laryngeally echolocating bats orient and hunt in complex environments such as dense vegetation.This requires considerable motor skills,e.g.maneuverability,as well as sensing capabilities that encode the relevant sensory information in a reliable fashion even in the presence of numerous nuisance echoes.The sensing capabilities involving prey in dense clutter in some bats are based on Doppler signatures.Specifically,species in the families Rhinolophidae,Hipposideridae,Rhinonycteridae and a few species of Mormoopidae have an integrated system that includes adaptations in pulse design,inner ear,auditory system and behavior.But how do these species deal with the other sensing challenges,associated with orientation in clutter?Peripheral dynamics of bats that exploit clutter could underlie the biosonar systems of these bats.Certain species of echolocating bats,Old World leaf-nosed bats and horseshoe bats(families Hipposideridae and Rhinolophidae),display a conspicuous dynamics at the periphery of their biosonar systems.The animals move their noseleaf("megaphone-like"emission baffles)and pinnae(outer ears)while the ultrasonic biosonar pulses are transmitted and the respective echoes are received.These motions are driven by sets of muscles that can generate multiple degrees of freedom,including various deformations and rotations,resulting in multiple different motion patterns of the noseleaf and the pinnae.The structural shape change of the emission side(noseleaf)significantly changes the characteristics of the emitted beam,and also adds a time dimension to the acoustic characteristics of the bat’s biological sonar system.Likewise,the dynamic changes of the reception(outer ear)side enhance the accuracy of sensory information and direction findings.This morphological and behavioral complexity demonstrated that the peripheral dynamics of the noseleaf and pinna of hipposideridae and rhinolophidae bats play a functional role in modulating the encoding of sensory information in biological sonar.However,whether there is a relationship between the noseleaf and the pinnae of bats and the underlying functional mechanisms remain unknown.So far,only Doppler shifts generated by fast pinna motion have been used in bionic systems for direction finding based on individual signal frequencies and individual receivers,but similar dynamic properties and functional relationships between them haven’t been applied in relevant engineering(e.g.,UAVs,engineering sonar or radar).Most of the existing designed bionic sonar robots are static or have limited peripheral deformation degrees of freedom.Most of the designed sonar heads are based on rhinolophidae bats,and there is no comparative model.Hence,to address the relationship between bat peripheral dynamics and potential functional advantages,the thesis investigated the dynamics between the noseleaf and pinnae of the hipposiderid bats during echolocation.Based on this,a dynamic bionic soft-robot based on the hipposiderid bats was developed from the perspective of biomimicry,and a deep learning approach was used to systematically investigate the potential functional properties of different coordinated patterns simulated by this model.In the thesis,we firstly took the hipposiderid bats as the research object,made marker points on the edges of the noseleaf and two pinnae,and recorded the bat’s behavior simultaneously with high-speed camera array and a microphone.The multi-marker tracking algorithm and coordinate 3D reconstruction were introduced to obtain the marker point trajectory and 3D coordinates,and to obtain the information of the markers with the greatest motion characteristics on the noseleaf and pinna by numerical analysis.The k-means clustering algorithm and principal component analysis were introduced to classify the motion patterns of the noseleaf,and the motion patterns of the pinna were classified according to the motion patterns and states.The dynamic relationship between the noseleaf and pinnae was analyzed by numerical analysis and canonical correlation analysis.Secondly,a dynamic biomimetic sonar soft-robot was built by mimicking the non-rigid coordinated patterns of the noseleaf and pinna of hipposiderid bats.A biomimetic dynamic receiver was designed to introduce a pneumatic network elastomer to develop a bionic pinna model that achieved precise control of the pinna shape.Then,the dynamic transmitter and receiver were integrated to develop a bat-inspired dynamic bionic sonar soft-robot.A complete perception-action loop system inspired by bats was developed to provide a comparative model for existing dynamic bionimetic sonar systems and to enrich the variety of biomimetic sonar robots.The methods of energy sum of information,euclidean distance between signals,and normalized mutual information of signals were introduced to analyze the differences in the temporal dimension of signals generated by biomimetic sonar robots.Finally,the deep learning paradigm was introduced to classify the echo specrograms collected in different coordinated motion patterns.At the same time,the robustness of the classification performance of the convolutional neural network classifier on the echo frequency-domain signals was verified,and the visualization method of convolutional neural networks was introduced to verify functional differences of different coordinated motion patterns.The non-rigid deformation of bat noseleaf and pinna has not been directly equated in current sonar technology,and bat echolocation mechanisms simulated using a biomimetic design paradigm could provide more insight into improved acoustic sensing mechanisms.Delving into how bats coordinate their motions and studying the potential sensing tasks they performed could help integrate this novel dynamic dimension into man-made technologies.The main innovative achievements of the thesis are as follows:Firstly,taking hipposiderid bats as the research object,the k-means clustering algorithm was used to classify the noseleaf motion patterns into four categories,included:Close,Open,Random and No motion(No motion).The pinna motion mode was divided into rigid and non-rigid motion,and the non-rigid motion was redefined as close and open motion.The results of the dynamic relationship between the noseleaf and pinnae by numerical and canonical correlation analysis showed that different pairs of motion patterns between noseleaf and pinnae of bats occur at different frequencies,there wss a strong coordinated relationship between the noseleaf and pinnae motion.It is revealed that the biological sonar system of hipposiderid bats includes coordinated transmit and receive dynamics.The positive results of this work lay the foundation for a better understanding of bats and other active dynamic sensing paradigms.Secondly,a dynamic biomimetic sonar soft-robot was developed based on the complex structure of the noseleaf and pinna of hipposiderid bats.By introducing the pneumatic network elastomer,the pinna had a smoother and curled deformation,which achieved the local precise control of the pinna shape and was closer to the non-rigid deformation of the real bat pinna.The coordinated motion patterns of bats and the motion patterns not found in bats were successfully reproduced by the dynamic biomimetic soft-robot.The above findings not only expand the variety of biomimetic sonar systems,but also provide a platform for in-depth study of potential sensing tasks performed by bat coordinated motions.Thirdly,taking the biomimetic dynamic sonar soft-robot as the research object,the Energy,Euclidean distance,and Normalized mutual information of the signal were analyzed.It is revealed the differences in the time dimension of the signal and the different coordinated motion patterns lead to different acoustic characteristics due to different coordinated motion patterns.The deep learning and neural network visualization methods were used to elucidated that the coordinated motions between the noseleaf and pinna of hipposiderid bats have different functional correlation among biosonar systems.These results demonstrated that the coordinated motions between the noseleaf and the pinna of the bat have different properties in the biological sonar system. |
PDF Full Text Request