Time-Variant Characteristics And Sensory Information Of Biomimetic Dynamic Sonar Emitter Inspired By Bats

The ultrasound emission system of bats,such as Old World leaf-nosed bats(family Hipposideridae)and the related horseshoe bats(Rhinolophidae),is characterized by unique dynamics,i.e.,the complicated baffle shapes('noseleaves')surrounding the emitter(nostrils)deform actively within the pulse duration as a result of muscular actuation.Studies from numerical simulation and biomimetics have shown that the dynamics can alter the shape of emission beam substantially,introduce time-variant features to the emission characteristics,and hence change the acoustic emission characteristics to be a function of time,direction and frequency.Moreover,the added time dimension could support the encoding of additional,useful sensory information.However,the mechanism of the time-variant sensory effects is still unknown.As of now,nothing comparable to this dynamics has been used to any related engineering application(e.g.sonar or radar).Almost all the emitters used in robotic models are static.The exceptions are the dynamic sonar emitters inspired by a single species(horseshoe bats)with no comparative model.Therefore,in this dissertation,a new biomimetic dynamic sonar emitter was designed and established from the perspective of biomimetics by combining the dynamic biosonar emission system of Pratt's roundleaf bats(Hipposideros pratti)with the engineering techniques,to study the time-variant features and the influences on the encoding capacity of sensory information introduced by the dynamic baffles.A new biomimetic dynamic sonar emitter inspired by the noseleaf shape of Pratt's roundleaf bats was designed to recreated the noseleaf motion pattern observed in real bats,which enriches the bat species in biomimetics and thus provide a comparative model for current studies.Based the dynamic sonar emitter,quantitative and qualitative analysis for the time-variant features and its influence factors can provide experimental gist and methods for optimizing the parameters of engineering sonar emitter.Intrinsic dimensionality of the time-variant acoustic emission characteristics was estimated,which proves the time dimension added by noseleaf dynamics can have an independent influence on the encoding of sensory information.Information theoretic methods,such as differential entropy and relative entropy,were introduced to analyze and compare the encoding capacity of sensory information among time,direction,and frequency dimensions.Algorithm for estimating the mutual information based on differential entropy was introduced to analyze the additional sensory information in different beampatterns added by the noseleaf dynamics during the emission of time-variant signals,which can remedy the drawbacks of the algorithm based on discrete entropy.The meaning of this dissertation is twofold.It will help biologists to better understand the impact of noseleaf dynamics on the bats' 'field of view' within a pulse emission,so as to promote the study on the dynamic compilation mechanism of sensory information used by bats.On the other hand,it will help engineers to better understand the feasibility of dynamic engineering sonar emitter,thus promoting the development of novel engineering sensing system.Main contents and conclusions of this dissertation are as follows:Firstly,a new biomimetic dynamic sonar emitter was designed and built based on the noseleaf shapes and motion pattern of Pratt's roundleaf bats.To study the noseleaf dynamics from the biological perspective,an experimental platform consisting of high-speed video camera array,microphone array,and data acquisition system,was established to collect the noseleaf geometric deformations and the synchronized emitted ultrasonic pulses of Pratt's roundleaf bats.Trajectories of noseleaf motion and the corresponding acoustic emission characteristics were reconstructed.Impacts of noseleaf dynamics on the acoustic emission characteristics were assessed using correlation analysis.It was found that Pratt's roundleaf bats have conspicuous dynamics in the coronet,posterior leaf,anterior leaf and two nostril flips of the noseleaves presented as closing-opening motion.These dynamics influenced the emission characteristics with large noseleaf aperture size having narrower beam with opening.Whereas,there were many local changes for small noseleaf aperture size,which need to be further studied.Then,a biomimetic dynamic sonar emitter based on these biological experimental data,was designed and assembled by a flexible biomimetic noseleaf model,actuation mechanism,control system,and ultrasonic emission system,to reproduce the noseleaf dynamics observed in real bats and then use to study the time-variant characteristics and the encoding of sensory informationSecondly,based on the biomimetic dynamic sonar emitter,the impact of the dynamic noseleaf deformation on the time-variant emission characteristics was evaluated by virtue of gradient.Effects of frequency,noseleaf shape configuration,and motion pattern on the time-variant characteristics were analyzed.The results have shown that the noseleaf dynamics did introduce time variances to the emission characteristics,e.g.changes in the direction and width of lobes over time,the occurrence of new sidelobes.The time-variant features depended on frequency,noseleaf shape configuration.as well as motion pattern.Among the three parameters,frequency resulted in greater gradient magnitudes in the beampattern,followed by noseleaf shape configuration and motion pattern,respectively.However,the influences of frequency and noseleaf motion pattern on the gradient orientations were greater than shape configuration.The largest impact of noseleaf dynamics on time-variant characteristics was the obvious changes of gradient orientations.The entropy in gradient orientations was found to be larger for the dynamic than for the static condition across all the 72 combinations of frequency,noseleaf shape configuration,and motion pattern.Moreover,the difference could reach up to 9.4 times over the entropy in the respective static condition.However,the difference in average gradient magnitudes reached up to 53%over the value in the respective static reference.The biomimetic noseleaf model here generated much greater time variances than the highly simplified concave baffle inspired by horseshoe bats.This demonstrates that these time-variant properties are not solely due to the increase or decrease in aperture size but depend on the geometric shape details.Thirdly,a three-dimensional acoustic emission characteristics with dense sampling over time,direction,and frequency were obtained by the biomimetic dynamic sonar emitter.Intrinsic dimensionality of the emission characteristics was estimated using Isomap algorithm.The results have shown that the emitted acoustic amplitudes changed differently with time,direction,as well as frequency.The intrinsic dimensionality suggested by the algorithm was equal(or slightly larger)than the physical dimensions(time,direction,frequency),i.e.,the algorithm was not able to reduce the three physical dimensions.This indicates that the time dimension added by the noseleaf dynamics was independent with direction and frequency dimension,and hence could generate independent effects to the encoding of sensory information.Fourthly,differential entropy and relative entropy were used to characterize and assess the encoding capacity of sensory information along time,direction,and frequency dimension of the time-variant acoustic characteristics.The results have shown that the sensory information encoded by all the three dimensions distributed almost uniformly over the other two dimensions.Especially,the sensory information encoded in time and frequency operated equally over a large range of directions.This indicates that the biosonar emission system of bats was similar to human hearing,which serves to sense the entire surroundings of a person rather than a narrow foveal region.The capacity for encoding of sensory information along time dimension(mean:2.55 bits)was found to be clearly less than the capacity associated with direction and frequency(mean:4.5 bits and 4.24 bits,respectively).However,the high-value tail of the distribution over time(3.2?4.5 bits)overlapped the value ranges of differential entropy along frequency and direction(3.2?5 bits).The differential entropy over direction and frequency could be predicted by the standard deviations with correlation coefficient r=0.96 and 0.86 respectively.In contrast,the relation between differential entropy and standard deviation scattered over a large range for time dimension(r=0.78).The average entropy for frequency relative to time(3±0.7 bits)were always larger than those for time relative to frequency(1.4±0.3 bits)across all the 1040 direction samples.The differences between the averages were 1.6±0.5 bits,which indicate the coding capacity along the time and frequency dimensions does not differ substantially.Fifthly,the sensory information shared between pairs of beampattem sets that belong to different points in time spaced along the entire duration of a closing motion,i.e.,mutual information,was estimated using two estimators that is based on discrete entropy and differential entropy respectively.Even though the latter estimator was more accurate,the results from both estimators were quite similar and shown that the mutual information was a function of the distance between the points in time where the two beampattern sets were taken.In general,mutual information decreased as the distance increased.Nevertheless,the mutual information always ecayed down to 70%within 15 ms.This demonstrates that the noseleaf dynamics could provide bats with a variety of nearly independent 'field of view',i.e.,more sensory information would be accumulated in the course of the entire pulse.