The Workprogramme

In overview, the project work programme is as follows. We begin from the assumption that controllable variations in morphology are one set of variables to be exploited in the bat's active perception process. Consequently, a main focus of this project is the development of tools for obtaining detailed knowledge of the head, nose and ear morphology of living bats engaged in ecologically relevant tasks. Combining this morphological data with knowledge of bat vocalisations and movements --- the second set of controllable parameters at the bat's disposal --- during their task-related activity (i.e., search, prey detection, localization and capture) we reconstruct the bat's acoustic experience as it flies through the task. From this data set, we attempt to identify the salient acoustic information the bat uses to coordinate its movements and vocalisations and to what extent the bat controls these to optimise its ability to perform the sensory discrimination required to succeed. The outcome of this last step is a set of innovative computational models explicitly specifying hypothesised mechanisms for determining flight behaviour, acoustic emissions and choices of morphological deformation in relation to the task and the bat's ongoing acoustic experience of its enironment. The final step is to assess the quality and usefulness of these models: first, by an engineer's characterisation of their performance and limits, as examples of engineered embodied active sonar systems; and second, with reference back to failure modes and generalisation abilities of their living prototypes, as working models of the bat's behaviour control processes.

Since the methodology is biomimetic engineering, there is a critical requirement for accurate and appropriate biological data. This will be supplied in part from the expertise and experience of the consortium biologists but mostly from new bat experiments carefully designed and focussed to provide the engineering detail needed for the project.

To optimise the outcome of the biomimetic engineering phase, that phase will be run twice with the second round benefitting from the insights gained in the first modelling attempt (an example of a spiral design methodology).

Selected Bat Species

We have selected four species which exhibit either gleaning or trawling behaviour. Both types face rather similar sensorial tasks: they have to detect, localize and capture prey from surfaces. However, the complexity of the tasks differs, ranging from a highly cluttered, three-dimensional environment (vegetation) to a simpler, mostly two-dimensional background (water surface) which can be simulated using a mobile robot platform. Furthermore, the four species are characterized by striking differences in morphology of features that are associated with sound emission and reception, in particular nose, ear and face. The decision to work with four species of bat represents a compromise between practicality and scientific and engineering knowledge gained.

Micronycteris microtis has a simple face which functions as a `baseline model' for the morphological part of the project. It also has amazing discriminatory abilities, hovering to find stationary insects amongst the high clutter of vegetation. This bat forms the basis of our gleaning model.

Either Myotis daubentonii or Noctilio leporinus will form the prototype for the trawling bat model. Both are essential to the project for practical reasons --- the convenient availability of the European Myotis daubentonii allows testing of the visual shape acquisition methods using only `local' travel, while only Noctilio leporinus is large and strong enough to carry the telemike recording equipment needed to validate the acoustic simulation methods for reconstructing emitted bat calls.

The fourth species, Macrophyllum macrophyllum, is closely related to the gleaner Micronycteris microtis yet is itself a trawling bat. It thus provides a bridge between the gleaning and trawling bat models.

From a scientific point of view, the chosen bats offer a number of significant inter-species comparisons to help elucidate the interaction of morphology, acoustic behaviour and motor behaviour in the hunting task. The three trawling bat species have quite different morphological adaptations and echolocation calls, which we expect will help identify the key interactions to be modelled and will also help maintain generality of the engineering knowledge gained.

Note that two models are to be constructed, one for Micronycteris microtis, the only bat proven able to find motionless food in dense acoustic clutter; and one for a selected trawling bat --- Myotis daubentonii, Noctilio leporinus or (possibly) Macrophyllum macrophyllum. These models are constructed in parallel, by the two engineering teams. There is considerable common ground since the teams use the same acoustic tools, the same standard models of early auditory processing, and share the same system integration tools for the two platforms. We expect that, despite the very different tasks being modelled, each will to a considerable extent inform the modelling of the other and the common engineering goal of building robust versatile active perception systems; hence the shared workpackage structure.

The choice of Myotis daubentonii or Noctilio leporinus as the principal prototype species for the trawling bat model will depend on the outcome of the acoustic and shape data capture work described in the detailed work plan below.

[More detail]