André Frank Krause

Tactile efficiency of insect antennae with two hinge joints

Antennae are the main organs of the arthropod tactile sense. In contrast to other senses that are capable of retrieving spatial information, e.g. vision, spatial sampling of tactile information requires active movement of the sense organ. For a quantitative analysis of basic principles of active tactile sensing, we use a generic model of arbitrary antennae with two hinge joints (revolute joints). This kind of antenna is typical for Orthoptera and Phasmatodea, i.e. insect orders that contain model species for the study of antennal movements, including cricket, locust and stick insect. First, we analyse the significance of morphological properties on workspace and sampling acuity. It is shown how joint axis orientation determines areas out of reach while affecting acuity in the areas within reach. Second, we assume a parametric set of movement strategies, based on empirical data on the stick insect Carausius morosus, and investigate the role of each strategy parameter on tactile sampling performance. A stochastic environment is used to measure sampling density, and a viscous friction model is assumed to introduce energy consumption and, thus, a measure of tactile efficiency. Up to a saturation level, sampling density is proportional to the range or frequency of joint angle modulation. The effect of phase shift is strong if joint angle modulation frequencies are equal, but diminishes for other frequency ratios. Speed of forward progression influences the optimal choice of movement strategy. Finally, for an analysis of environmental effects on tactile performance, we show how efficiency depends on predominant edge direction. For example, with slanted and non-orthogonal joint axis orientations, as present in the stick insect, the optimal sampling strategy is less sensitive to a change from horizontal to vertical edge predominance than with orthogonal and non-slanted joint axes, as present in a cricket.

Neuroethological Concepts and their Transfer to Walking Machines

A systems approach to animal motor behavior reveals concepts that can be useful for the pragmatic design of walking machines. This is because the relation of animal behavior to its underlying nervous control algorithms bears many parallels to the relation of machine function to electronic control. Here, three major neuroethological concepts of motor behavior are described in terms of a conceptual framework based on artificial neural networks (ANN). Central patterns of activity and postural reflexes are both interpreted as a result of feedback loops, with the distinction of loops via an internal model from loops via the physical environment (body, external world). This view allows continuous transitions between predictive (centrally driven) and reactive (reflex driven) motor systems. Motor primitives, behavioral modules that are elicited by distinct commands, are also considered. ANNs capture these three major concepts in terms of a formal description, in which the interactions and mutual interdependences of the various output parameters are comprised by the weight matrix of the net. Based upon behavioral observations of insect walking, we further demonstrate how a decentralized network of separate modules, each one described by an ANN, can account for adaptive behavior. Complex coordination patterns of several manipulators are controlled by imposing simple interaction rules between limbs, and by exploiting the interaction of the body with its physical environment. Finally, we discuss the technical use of leg-like active tactile sensors for obstacle detection, and we show how specific design of such active sensors may increase efficiency of walking on rough terrain. Applied to active sensors, an example of parallel, self-organizing forward models on the basis of extended Kohonen maps is presented to emphasize the potential of adaptive forward models in motor control.

Central drive and proprioceptive control of antennal movements in the walking stick insect.

In terrestrial locomotion, active touch sensing is an important source of near-range information. Walking stick insects show active tactile exploration behaviour by continuously sampling the ambient space with their antennae. Here, we identify central and proprioceptive contributions to the control of this behaviour. First, we investigate the potential role of synaptic drive to central neural networks using pilocarpine, an agonist of muscarinic acetylcholine receptors. In an in situ preparation, pilocarpine induced rhythmic antennal movements with a persisting pattern of inter-joint coordination, matching that seen in intact walking animals, albeit with lower cycle frequency. After de-cerebration, stick insects were still able to walk but no longer moved their antennae during walking. Here, pilocarpine still induced antennal movement, suggesting that synaptic drive to central neural networks involved in antennal movement generation occurred in the brain and not in the suboesophageal ganglion. During intact walking, these networks are likely to receive activation by ascending input. Second, we show persistent coupling of both antennal joints during intact walking, with the distal scape-pedicel joint (SP) always leading the proximal head-scape joint (HS). Ablation of joint proprioceptors had no effect on this overall pattern of inter-joint coordination but could affect the magnitude of the phase-lag. Third, we revise the description of antennal hair fields and show that complete ablation of all seven hair fields strongly affects antennal movements. Ablating dorsal hair fields mainly affected the working-ranges of antennal joints: Ablation of the dorso-medial pedicellar hair plate caused a ventral shift of the SP working-range. Ablation of the dorsal scapal hair plate considerably expanded the dorsal HS working-range, and, in combination with ablation of pedicellar hair fields, increased the SP working-range, too. We conclude that the working-ranges of both joints are under proprioceptive control of dorsal antennal hair fields. Thus, both synaptic drive to central neural networks and proprioceptive feedback are involved in the control of active tactile exploration behaviour in stick insects.

An insect-inspired bionic sensor for tactile localization and material classification with state-dependent modulation

Insects carry a pair of antennae on their head: multimodal sensory organs that serve a wide range of sensory-guided behaviors. During locomotion, antennae are involved in near-range orientation, for example in detecting, localizing, probing, and negotiating obstacles. Here we present a bionic, active tactile sensing system inspired by insect antennae. It comprises an actuated elastic rod equipped with a terminal acceleration sensor. The measurement principle is based on the analysis of damped harmonic oscillations registered upon contact with an object. The dominant frequency of the oscillation is extracted to determine the distance of the contact point along the probe and basal angular encoders allow tactile localization in a polar coordinate system. Finally, the damping behavior of the registered signal is exploited to determine the most likely material. The tactile sensor is tested in four approaches with increasing neural plausibility: first, we show that peak extraction from the Fourier spectrum is sufficient for tactile localization with position errors below 1%. Also, the damping property of the extracted frequency is used for material classification. Second, we show that the Fourier spectrum can be analysed by an Artificial Neural Network (ANN) which can be trained to decode contact distance and to classify contact materials. Thirdly, we show how efficiency can be improved by band-pass filtering the Fourier spectrum by application of non-negative matrix factorization. This reduces the input dimension by 95% while reducing classification performance by 8% only. Finally, we replace the FFT by an array of spiking neurons with gradually differing resonance properties, such that their spike rate is a function of the input frequency. We show that this network can be applied to detect tactile contact events of a wheeled robot, and how detrimental effects of robot velocity on antennal dynamics can be suppressed by state-dependent modulation of the input signals.

Direct Control of an Active Tactile Sensor Using Echo State Networks

Tactile sensors (antennae) play an important role in the animal kingdom. They are also very useful as sensors in robotic scenarios, where vision systems may fail. Active tactile movements increase the sampling performance. Here we directly control movements of the antenna of a simulated hexapod using an echo state network (ESN). ESNs can store multiple motor patterns as attractors in a single network and generate novel patterns by combining and blending already learned patterns using bifurcation inputs.

Active tactile sampling by an insect in a step-climbing paradigm.

Many insects actively explore their near-range environment with their antennae. Stick insects (Carausius morosus) rhythmically move their antennae during walking and respond to antennal touch by repetitive tactile sampling of the object. Despite its relevance for spatial orientation, neither the spatial sampling patterns nor the kinematics of antennation behavior in insects are understood. Here we investigate unrestrained bilateral sampling movements during climbing of steps. The main objectives are: (1) How does the antennal contact pattern relate to particular object features? (2) How are the antennal joints coordinated during bilateral tactile sampling? We conducted motion capture experiments on freely climbing insects, using steps of different height. Tactile sampling was analyzed at the level of antennal joint angles. Moreover, we analyzed contact patterns on the surfaces of both the obstacle and the antenna itself. Before the first contact, both antennae move in a broad, mostly elliptical exploratory pattern. After touching the obstacle, the pattern switches to a narrower and faster movement, caused by higher cycle frequencies and lower cycle amplitudes in all joints. Contact events were divided into wall- and edge-contacts. Wall contacts occurred mostly with the distal third of the flagellum, which is flexible, whereas edge contacts often occurred proximally, where the flagellum is stiff. The movement of both antennae was found to be coordinated, exhibiting bilateral coupling of functionally analogous joints [e.g., left head-scape (HS) joint with right scape-pedicel (SP) joint] throughout tactile sampling. In comparison, bilateral coupling between homologous joints (e.g., both HS joints) was significantly weaker. Moreover, inter-joint coupling was significantly weaker during the contact episode than before. In summary, stick insects show contact-induced changes in frequency, amplitude and inter-joint coordination during tactile sampling of climbed obstacles.

Slanted joint axes of the stick insect antenna: an adaptation to tactile acuity

Like many flightless, obligatory walking insects, the stick insect Carausius morosus makes intensive use of active antennal movements for tactile near range exploration and orientation. The antennal joints of C. morosus have a peculiar oblique and non-orthogonal joint axis arrangement. Moreover, this arrangement is known to differ from that in crickets (Ensifera), locusts (Caelifera) and cockroaches (Blattodea), all of which have an orthogonal joint axis arrangement. Our hypothesis was that the situation found in C. morosus represents an important evolutionary trait of the order of stick and leaf insects (Phasmatodea). If this was true, it should be common to other species of the Phasmatodea. The objective of this comparative study was to resolve this question. We have measured the joint axis orientation of the head–scape and scape–pedicel joints along with other parameters that affect the tactile efficiency of the antenna. The obtained result was a complete kinematic description of the antenna. This was used to determine the size and location of kinematic out-of-reach zones, which are indicators of tactile acuity. We show that the oblique and non-orthogonal arrangement is common to eight species from six sub-families indicating that it is a synapomorphic character of the Euphasmatodea. This character can improve tactile acuity compared to the situation in crickets, locusts and cockroaches. Finally, because molecular data of a recent study indicate that the Phasmatodea may have evolved as flightless, obligatory walkers, we argue that the antennal joint axis arrangement of the Euphasmatodea reflects an evolutionary adaptation to tactile near range exploration during terrestrial locomotion.

Bionic Tactile Sensor for Near-Range Search, Localisation and Material Classification

Insects use their antennae (feelers) as near range sensors for orientation, object localisation and communication. Here, we use the stick insect antenna as a paragon for an actively moved tactile sensor. Our bionic sensor uses vibration signals from contact events for obstacle localisation and classification of material properties. It is shown how distance is coded by salient peaks in the frequency spectrum, and how the damping time constants can be exploited to distinguish between eight objects made of a range of materials. Thus, we demonstrate application of bionic principles for non-visual, reliable, near-range object localisation and material classification that is suitable for autonomous exploratory robots.

Feel like an insect: a bio-inspired tactile sensor system

Insects use their antennae (feelers) as near range sensors for orientation, object localization and communication. This paper presents an approach for an active tactile sensor system. This includes a new type of hardware construction as well as a software implementation for interpreting the sensor readings. The discussed tactile sensor is able to detect an obstacle and its location in 3D space. Furthermore the material properties of the obstacles are classified by use of neural networks.

No-Prop-fast - A High-Speed Multilayer Neural Network Learning Algorithm: MNIST Benchmark and Eye-Tracking Data Classification

While the No-Prop (no back propagation) algorithm uses the delta rule to train the output layer of a feed-forward network, No-Prop-fast employs fast linear regression learning using the Hopf-Wiener solution. Ten times faster learning speeds can be achieved on large datasets like the MNIST benchmark, compared to one of the fastest backpropagation algorithm known. Additionally, the plain feed-forward network No-prop-fast can distinguish gaze movements on cartoons with and without text, as well as age-specific attention shifts between text and picture areas with minimal pre-processing.