Skip to main content

Über dieses Buch

This book constitutes the proceedings of the Third International Conference on Biomimetic and Biohybrid Systems, Living Machines 2014, held in Milan, Italy, in July/August 2014.

The 31 full papers and 27 extended abstracts included in this volume were carefully reviewed and selected from 62 submissions. The topics covered are brain based systems, active sensing, soft robotics, learning, memory, control architectures, self-regulation, movement and locomotion, sensory systems and perception.



Full Papers

Monolithic Design and Fabrication of a 2-DOF Bio-Inspired Leg Transmission

We present the design of a new two degree-of-freedom transmission intended for micro / meso-scale crawling robots which is compatible both with laminate manufacturing techniques and monolithic,“pop-up” assembly methods. This is enabled through a new design suite called “popupCAD”, a computer-aided design tool which anticipates laminate manufacturing methods with a suite of operations which simplify the existing design workflow. The design has been prototyped at three times the anticpated scale to better understand the assembly and motion kinematics, and simulated to establish the basic relationships between the actuator and end-effector transmission ratios.

Daniel M. Aukes, Önur Ozcan, Robert J. Wood

Optimization of the Anticipatory Reflexes of a Computational Model of the Cerebellum

The cerebellum is involved in avoidance learning tasks, where anticipatory actions are developed to protect against aversive stimuli. In the execution and acquisition of discrete actions we can distinguish errors of omission and commission due to a failure to execute a required defensive Conditioned Response (CR) to avoid an aversive Unconditioned Stimulus (US), and the energy expenditure of triggering an unnecessary CR in the absence of a US respectively. Hence, a motor learning cost function must consider both these components of performance and energy expenditure. Unlike remaining noxious stimuli, unnecessary actions are not directly sensed by the cerebellum. It has been suggested that the Nucleo-Olivary Inhibition (NOI) serves to internally rely information about these needless protective actions. Here we argue that the function of the NOI can be interpreted in broader terms as a signal that is used to learn optimal actions in terms of cost. We work with a computational model of the cerebellum to address: (i) how can the optimum balance between remaining aversive stimuli and preventing effort be found, and (ii) how can the cerebellum use the overall cost information to establish this optimum balance through the adjusting of the gain of the NOI. In this paper we derive the value of the NOI that minimizes the overall cost and propose a learning rule for the cerebellum through which this value is reached. We test this rule in a collision avoidance task performed by a simulated robot.

Santiago Brandi, Ivan Herreros, Paul F. M. J. Verschure

Evolving Optimal Swimming in Different Fluids: A Study Inspired by batoid Fishes

For their efficient and elegant locomotion,


fishes (e.g. the manta ray) have been widely studied in biology, and also taken as a source of inspiration by engineers and roboticists willing to replicate their propulsion mechanism in order to build efficient swimming machines. In this work, a new model of an under-actuated compliant wing is proposed, exhibiting both the oscillatory and undulatory behaviors underlying


propulsion mechanism. The proposed model allowed an investigation of the co-evolution of morphology and control, exploiting dynamics emergent from the interaction between the environment and the mechanical properties of the soft materials. Having condensed such aspects in a mathematical model, we studied the adaptability of a


-like morphology to different environments. As for biology, our main contribution is an exploration of the parameters linking swimming mechanics, morphology and environment. This can contribute to a deeper understanding of the factors that led various species of the


group to phylogenetically adapt to different environments. From a robotics standpoint, this work offers an additional example remarking the importance of morphological computation and embodied intelligence. A direct application can be an under-water soft robot capable of adapting morphology and control to reach the maximum swimming efficiency.

Vito Cacucciolo, Francesco Corucci, Matteo Cianchetti, Cecilia Laschi

Bipedal Walking of an Octopus-Inspired Robot

In this paper a model is presented which describes an octopus-inspired robot capable of two kinds of locomotion: crawling and bipedal walking. Focus will be placed on the latter type of locomotion to demonstrate, through model simulations and experimental trials, that the robot’s speed increases by about 3 times compared to crawling. This finding is coherent with the performances of the biological counterpart when adopting this gait. Specific features of underwater legged locomotion are then derived from the model, which prompt the possibility of controlling locomotion by using simple control and by exploiting slight morphological adaptations.

Marcello Calisti, Francesco Corucci, Andrea Arienti, Cecilia Laschi

Action Selection within Short Time Windows

In this paper, we study the expansion of a reactive network that is based on a decentralized, heterarchical architecture able to control a hexapod robot. Within this network, problems may occur if more than one task should be addressed within a short time window. Such situations have been studied in so called dual-task experiments in the field of psychology. We take these results as inspiration to develop a structure that can be integrated into our framework. The model focuses on essential aspects of the basic phenomena observed in human subjects as forward masking, backward masking and PRP. While there are detailed models available concentrating on specific paradigms, those do not cover all three types of paradigms. We are not heading for a detailed simulation of the psychological findings, but show how these effects, on a qualitative level, can be implemented in our framework.

Holk Cruse, Malte Schilling

Modelling Legged Robot Multi-Body Dynamics Using Hierarchical Virtual Prototype Design

Legged robots represent the bio-inspired family of robotic devices which has to perform the most complex dynamic tasks. It is essential for them to walk in unstructured terrains, carry heavy loads, climb hills and run up to a certain speed. A complete understanding of these performances and their optimization should involve both the control and the mechanics which has been ignored by robotic researchers for years. The solution we propose is a tradeoff between control and mechanics based on the Virtual Prototype Design Method. We build a simplified numerical model of a quadruped leg based on a hierarchial architecture. The proposed model is validated by comparing the numerical solution and the physical results coming from an extended campaign of experimental tests.

Mariapaola D’Imperio, Ferdinando Cannella, Fei Chen, Daniele Catelani, Claudio Semini, Darwin G. Caldwell

How Cockroaches Employ Wall-Following for Exploration

Animals such as cockroaches depend on exploration of unknown environments, and the complexity of their strategies may inspire robotic approaches. We have previously shown that cockroach behavior with respect to shelters and the walls of an otherwise empty arena can be captured with a stochastic state-based algorithm. We call this algorithm RAMBLER, Randomized Algorithm Mimicking Biased Lone Exploration in Roaches. In this work, we verified and extended this model by adding a barrier to our cockroach experiments. From these experiments, we have generalized RAMBLER to address an arbitrarily large maze. For biology, this is a model of the decision-making process in the cockroach brain. For robotics, this is a strategy that may improve exploration for goals in certain environments. Generally, the cockroach behavior seems to recommend variability in the absence of planning, and following paths defined by the walls.

Kathryn A. Daltorio, Brian T. Mirletz, Andrea Sterenstein, Jui Chun Cheng, Adam Watson, Malavika Kesavan, John A. Bender, Roy E. Ritzmann, Roger D. Quinn

Machines Learning - Towards a New Synthetic Autobiographical Memory

Autobiographical memory is the organisation of episodes and contextual information from an individual’s experiences into a coherent narrative, which is key to our sense of self. Formation and recall of autobiographical memories is essential for effective, adaptive behaviour in the world, providing contextual information necessary for planning actions and memory functions such as event reconstruction. A synthetic autobiographical memory system would endow intelligent robotic agents with many essential components of cognition through active compression and storage of historical sensorimotor data in an easily addressable manner. Current approaches neither fulfil these functional requirements, nor build upon recent understanding of predictive coding, deep learning, nor the neurobiology of memory. This position paper highlights desiderata for a modern implementation of synthetic autobiographical memory based on human episodic memory, and proposes that a recently developed model of hippocampal memory could be extended as a generalised model of autobiographical memory. Initial implementation will be targeted at social interaction, where current synthetic autobiographical memory systems have had success.

Mathew H. Evans, Charles W. Fox, Tony J. Prescott

A Phase-Shifting Double-Wheg-Module for Realization of Wheg-Driven Robots

Following mechatronic design methodology this paper introduces a phase-shifting double-wheg-module which forms an alternative approach for wheg-driven robots. During construction focus was placed on a smooth locomotion of the wheg-mechanism over flat terrain (low alternation of the CoM in vertical y-direction) as well as the ability to overcome obstacles. Simulations using the multi-body simulation tool ADAMS View


were executed in order to prove estimations done. Using the results of simulation and calculation a first prototype was designed, manufactured and tested by experiment.

Max Fremerey, Sebastian Köhring, Omar Nassar, Manuel Schöne, Karl Weinmeister, Felix Becker, Goran S. Đorđević, Hartmut Witte

Design Principles for Cooperative Robots with Uncertainty-Aware and Resource-Wise Adaptive Behavior

In this paper we describe several principles for designing and implementing bio-inspired robotic collaborative search strategies. The design approach is particularly oriented for algorithms that can tackle search problems that deal with uncertainty, such as locating odor sources that have spatial and temporal variance. These kind of problems can be solved more efficiently by a reasonable amount of collaborative robots, and thus we propose a low-cost platform based on the open-source philosophy. The platform allows to evaluate different collective strategies that emerge from the interaction among robots that are aware of the uncertainty and make a wise use of all available sensors and resources. This includes an adaptive use of sensor signals and actuators, and a good communication strategy. We introduce GNBot, a flexible open-source robotic platform, and a virtual communication network topology approach to validate uncertainty-aware and resource wise bio-inspired search strategies.

Carlos García-Saura, Francisco de Borja Rodríguez, Pablo Varona

Insect-Inspired Tactile Contour Sampling Using Vibration-Based Robotic Antennae

Compared to vision, active tactile sensing enables animals and robots to perform unambiguous object localization, segmentation and shape recognition. Recently, we proposed a bio-inspired, CPG-based, active antennal control model, so-called Contour-net, which captures essential characteristics of antennal behavior in climbing stick insects. In simulation, this model provides a robust and effective way to trace contours and classify various 3D shapes. Here, we propose a physical robotic implementation of Contour-net using vibration-based active antennae. We show that combining tactile contour tracing with vibration-based distance estimation yields fairly accurate localization of contact events in 3D space.

Thierry Hoinville, Nalin Harischandra, André F. Krause, Volker Dürr

A Predictive Model for Closed-Loop Collision Avoidance in a Fly-Robotic Interface

Here we propose a control design for a calibrated fly-brain-robotic interface. The interface uses the spiking activity of an identified visual interneuron in the fly brain, the H1-cell, to control the trajectory of a 2-wheeled robot such that it avoids collision with objects in the environment. Control signals will be based on a comparison between predicted responses – derived from the known robot dynamics and the H1-cell responses to visual motion in an isotropic distance distribution – and the actually observed spike rate measured during movements of the robot. The suggested design combines two fundamental concepts in biological sensorimotor control to extract task-specific information: active sensing and the use of efference copies (forward models). In future studies we will use the fly-robot interface to investigate multisensory integration.

Jiaqi V. Huang, Holger G. Krapp

Neuromechanical Simulation of an Inter-leg Controller for Tetrapod Coordination

A biologically inspired control system has been developed for coordinating a tetrapod walking gait in the sagittal plane. The controller is built with biologically based neurons and synapses, and connections are based on data from literature where available. It is applied to a simplified, planar biomechanical model of a rat with 14 joints with an antagonistic pair of Hill muscle models per joint. The controller generates tension in the muscles through activation of simulated motoneurons. Though significant portions of the controller are based on cat research, this model is capable of reproducing hind leg behavior observed in walking rats. Additionally, the applied inter-leg coordination pathways between fore and hind legs are capable of creating and maintaining coordination in this rat model. Ablation tests of the different connections involved in coordination indicate the role of each connection in providing coordination with low variability.

Alexander Hunt, Manuela Schmidt, Martin Fischer, Roger D. Quinn

A Minimum Attention Control Law for Ball Catching

We present an attention-minimizing LQR-based feedback control law for vision-based ball catching. Taking Brockett’s control attention functional as our performance criterion, and under the simplifying assumption that the optimal control law is the sum of a linear time-varying feedback term and a time-varying feedforward term, we show that our control law is stable, and easily obtained as the solution to a finite-dimensional optimization problem over the set of symmetric positive-definite matrices. We perform numerical experiments for robotic ball-catching and compare our control law with a discretized version obtained as the solution to a mathematical programming problem. Like human ball catching, our results also exhibit the familiar transition from open-loop to closed-loop control during the catching movement, and also show improved robustness to spatiotemporal quantization. Our approach is applicable to more general control settings in which multi-tasking must be performed under limited computation and communication resources.

Cheongjae Jang, Jee-eun Lee, Sohee Lee, Frank C. Park

Simulated Neural Dynamics Produces Adaptive Stepping and Stable Transitions in a Robotic Leg

Animals exhibit flexible and adaptive behavior. They can change between modes of locomotion or modify the details of a step to better suit their environment. Insects have massively distributed control architectures in which each joint has its own central pattern generator (CPG), which is coordinated with its neighbors only through sensory information. Different modes of walking (forward, turning, etc.) can be produced by changing which CPGs are affected by which sensory information, called a reflex reversal. The presented robotic leg is controlled by a computational neuroscience model of part of the nervous system of the cockroach

Blaberus discoidalis

. It steps adaptively to correct for unexpected obstacles and can reverse reflexes to produce turning motions.

Matthew A. Klein, Nicholas S. Szczecinski, Roy E. Ritzmann, Roger D. Quinn

Blending in with the Shoal: Robotic Fish Swarms for Investigating Strategies of Group Formation in Guppies

Robotic fish that dynamically interact with live fish shoals dramatically augment the toolset of behavioral biologists. We have developed a system of biomimetic fish for the investigation of collective behavior in Guppies and similarly small fish. This contribution presents full implementation details of the system and promising experimental results. Over long durations our robots are able to integrate themselves into shoals or recruit the group to exposed locations that are usually avoided. This system is the first open-source project for both software and hardware components and is supposed to facilitate research in the emerging field of bio-hybrid societies.

Tim Landgraf, Hai Nguyen, Joseph Schröer, Angelika Szengel, Romain J. G. Clément, David Bierbach, Jens Krause

Capturing Stochastic Insect Movements with Liquid State Machines

A Liquid State Machine (LSM) is trained to model the stochastic behavior of a cockroach exploring an unknown environment. The LSM is a recurrent neural network of leaky-integrate-and-fire neurons interconnected by synapses with intrinsic dynamics and outputs to an Artificial Neural Network (ANN). The LSM is trained by a reinforcement approach to produce a probability distribution over a discrete control space which is then sampled by the controller to determine the next course of action. The LSM is able to capture several observed phenomenon of cockroach exploratory behavior including resting under shelters and wall following.

Alexander Lonsberry, Kathryn Daltorio, Roger D. Quinn

Acquisition of Synergistic Motor Responses through Cerebellar Learning in a Robotic Postural Task

Coordination of synergistic movements is a crucial aspect of goal oriented motor behavior in postural control. It has been proposed that the cerebellum could be involved in the acquisition of adaptive fine-tuned motor responses. However, it remains unclear whether motor patterns and action sequences can be learned as a result of recurrent connections among multiple cerebellar microcircuits. Within this study we hypothesize that such link could be found in the Nucleo-Pontine projection and we investigate the behavioral advantages of cerebellar driven synergistic motor responses in a robotic postural task. We devise a scenario where a double-joint cart-pole robot has to learn to stand and balance interconnected segments by issuing multiple actions in order to minimize the deviation from a state of equilibrium. Our results show that a cerebellum based architecture can efficiently learn to reduce errors through well-timed motor coordination. We also suggest that such strategy could reduce energy cost by progressively synchronizing multiple joints movements.

Giovanni Maffei, Marti Sanchez-Fibla, Ivan Herreros, Paul F. M. J. Verschure

I-CLIPS Brain: A Hybrid Cognitive System for Social Robots

Sensing and interpreting the interlocutor’s social behaviours is a core challenge in the development of social robots. Social robots require both an innovative sensory apparatus able to perceive the “social and emotional world” in which they act and a cognitive system able to manage this incoming sensory information and plan an organized and pondered response. In order to allow scientists to design cognitive models for this new generation of social machines, it is necessary to develop control architectures that can be easily used also by researchers without technical skills of programming such as psychologists and neuroscientists. In this work an innovative hybrid deliberative/reactive cognitive architecture for controlling a social humanoid robot is presented. Design and implementation of the overall architecture take inspiration from the human nervous system. In particular, the cognitive system is based on the Damasio’s thesis. The architecture has been preliminary tested with the FACE robot. A social behaviour has been modeled to make FACE able to properly follow a human subject during a basic social interaction task and perform facial expressions as a reaction to the social context.

Daniele Mazzei, Lorenzo Cominelli, Nicole Lazzeri, Abolfazl Zaraki, Danilo De Rossi

Change of Network Dynamics in a Neuro-robotic System

In the past, tetanic stimulation has been used in several different instances to induce changes in the firing patterns of neural networks

in vitro

. In this paper, we ran a new experimental campaign to verify if this protocol induced lasting changes and if those changes were predictable. We found out that our stimulation protocol led to different results in cortical and hippocampal preparations: in the first case, stronger connections were weakened, resulting in a reduction of bursting activity and late evoked response; in the case of hippocampal preparations, single strong connections underwent strong changes but, on average, remained unchanged. In both preparations, the geometry of induced changes remains largely uncorrelated with the actual site of stimulation delivery.

Irene Nava, Jacopo Tessadori, Michela Chiappalone

Enhanced Locomotion of a Spherical Robot Based on the Sea-urchin Characteristics

The objective of this research is to present the design of a robot able to act in highly unstructured environments. To do so, a spherical robot loosely based on the sea-urchin characteristics is introduced; to synthesize such characteristics a set of actuators were incorporated into a pendulum-based spherical robot in order to handle obstacles and hollows successfully, as demonstrated in the simulation results that are shown and discussed.

Jorge Ocampo-Jiménez, Angelica Muñoz-Meléndez, Gustavo Rodríguez-Gómez

Design of a Control Architecture for Habit Learning in Robots

Researches in psychology and neuroscience have identified multiple decision systems in mammals, enabling control of behavior to shift with training and familiarity of the environment from a goal-directed system to a habitual system. The former relies on the explicit estimation of future consequences of actions through planning towards a particular goal, which makes decision time longer but produces rapid adaptation to changes in the environment. The latter learns to associate values to particular stimulus-response associations, leading to quick reactive decision- making but slow relearning in response to environmental changes. Computational neuroscience models have formalized this as a coordination of model-based and model-free reinforcement learning. From this inspiration we hypothesize that it could enable robots to learn habits, detect when these habits are appropriate and thus avoid long and costly computations of the planning system. We illustrate this in a simple repetitive cube-pushing task on a conveyor belt, where a speed-accuracy trade-off is required. We show that the two systems have complementary advantages in these tasks, which can be combined for performance improvement.

Erwan Renaudo, Benoît Girard, Raja Chatila, Mehdi Khamassi

Dynamic Model of a Jet-Propelled Soft Robot Inspired by the Octopus Mantle

This article addresses the study of cephalopods locomotion. As a first step, we here propose an analytical model of the cephalopod mantle. The approach is based on the geometrically exact shell theory used in nonlinear structural dynamics and exploits the symmetric shape of the mantle. Once all the mathematical background is reminded we propose a first use of this model by using the constitutive laws of the shell as control laws able to reproduce the axisymmetric contractions observed in cephalopods. Further numerical exploitation of these theoretical results are to date in progress along with works in fluid mechanics.

Federico Renda, Frederic Boyer, Cecilia Laschi

Hippocampal Based Model Reveals the Distinct Roles of Dentate Gyrus and CA3 during Robotic Spatial Navigation

Animals are exemplary explorers and achieve great navigational performances in dynamic environments. Their robotic counterparts still have difficulties in self-localization and environment mapping tasks. Place cells, a type of cell firing at specific positions in the environment, are found in multiple areas of the hippocampal formation. Although, the functional role of these areas with a similar type of cell behavior is still not clearly distinguished. Biomimetic models of navigation have been tested in the context of computer simulations or small and controlled arenas. In this paper, we present a computational model of the hippocampal formation for robotic spatial representation within large environments. Necessary components for the formation of a cognitive map [1], such as grid and place cells, were obtained through attractor dynamics. Prediction of future hippocampal inputs was performed through self-organization. Obtained data suggests that the integration of the described components is sufficient for robotic space representation. In addition, our results suggest that dentate gyrus (DG), the hippocampal input area, integrates signals from different dorsal-ventral scales of grid cells and that spatial and sensory input are not necessarily associated in this region. Moreover, we present a mechanism for prediction of future hippocampal events based on associative learning.

Diogo Santos Pata, Alex Escuredo, Stéphane Lallée, Paul F. M. J. Verschure

Trajectory Control Strategy for Anthropomorphic Robotic Finger

This paper proposes a trajectory control strategy for a tendon-driven robotic finger based on the musculoskeletal system of the human finger. First, we analyzed the relationship between the stereotypical trajectory of the human finger and joint torques generated by the muscles, and hypothesized that the motion of the human finger can be divided into two categories: one following a predetermined trajectory and the other changing the trajectory, which is mainly caused by the action of intrinsic muscles. We applied this control method to an anthropomorphic tendon-driven robotic finger and observed the change in motion caused by adjustments in the actuator’s pattern, which corresponds to human intrinsic muscles.

Shouhei Shirafuji, Shuhei Ikemoto, Koh Hosoda

Neuromechanical Mantis Model Replicates Animal Postures via Biological Neural Models

A neuromechanical model of a mantis was developed to explore the neural basis of some elements of hunting behavior, which is very flexible and context-dependent, for robotic control. In order to capture the complexity and flexibility of insect behavior, we have leveraged our previous work [1] and constructed a dynamical model of a mantis with a control system built from dynamical neuron models, which simulate the flow of ions through cell membranes. We believe that this level of detail will provide more insight into what makes the animal successful than a finite state machine (FSM) or a recurrent neural network (RNN). Each of the model’s walking legs has six degrees of freedom. Each joint is actuated by an antagonistic pair of muscles, controlled by a custom designed variable-stiffness joint controller based on insect neurobiology. The resulting low-level control system serves as the groundwork for a more complete behavioral model of the animal.

Nicholas S. Szczecinski, Joshua P. Martin, Roy E. Ritzmann, Roger D. Quinn

A Natural Movement Database for Management, Documentation, Visualization, Mining and Modeling of Locomotion Experiments

In recent years, experimental data on natural, un-restrained locomotion of animals has strongly increased in complexity and quantity. This is due to novel motion-capture techniques, but also to the combination of several methods such as electromyography or force measurements. Since much of these data are of great value for the development, modeling and benchmarking of technical locomotion systems, suitable data management, documentation and visualization are essential. Here, we use an example of comparative kinematics of climbing insects to propose a data format that is equally suitable for scientific analysis and sharing through web repositories. Two data models are used: a relational model (SQL) for efficient data management and mining, and the Resource Description Framework (RDF), releasing data according to the Linked Data principles and connecting it to other datasets on the web. Finally, two visualization options are presented, using either a photo-realistic rendering or a plain but versatile cylinder-based 3D-model.

Leslie M. Theunissen, Michael Hertrich, Cord Wiljes, Eduard Zell, Christian Behler, André F. Krause, Holger H. Bekemeier, Philipp Cimiano, Mario Botsch, Volker Dürr

Benchmarking Human-Like Posture and Locomotion of Humanoid Robots: A Preliminary Scheme

The difficulty in defining standard benchmarks for human likeness is a well-known problem in bipedal robotics. This paper proposes the conceptual design of a novel benchmarking scheme for bipedal robots based on existing criteria and benchmarks related to the sensorimotor mechanisms involved in human walking and posture. The proposed scheme aims to be sufficiently generic to permit its application to a wide range of bipedal platforms, and sufficiently specific to rigorously test the sensorimotor skills found in humans.

The achievement of global consensus on the definition of human likeness has a crucial importance not only in the field of humanoid robotics, but also in neuroscience and clinical settings. The EU project H2R is specifically dealing with this problem. A preliminary solution is here given to encourage the international discussion on this topic within the scientific community.

Diego Torricelli, Rahman S. M. Mizanoor, Jose Gonzalez, Vittorio Lippi, Georg Hettich, Lorenz Asslaender, Maarten Weckx, Bram Vanderborght, Strahinja Dosen, Massimo Sartori, Jie Zhao, Steffen Schüetz, Qi Liu, Thomas Mergner, Dirk Lefeber, Dario Farina, Karsten Berns, Jose Louis Pons

The Influence of Behavioral Complexity on Robot Perception

Since robots’ capabilities increase, they will soon be present in our daily lives and will be required to interact with humans in a natural way. Furthermore, robots will need to be removed from controlled environments and tested in public places where untrained people will be able to freely interact with them. Such needs raise a number of issues: what kind of behaviors are considered important in promoting interaction and how these behaviors affect people’s perception regarding the robot in terms of anthropomorphism, likeability, animacy and perceived intelligence. In this paper, we propose a motivational and emotional system that drives the robot’s behavior and test it against six interaction scenarios of varying complexity. In addition, we evaluate our system in two different environments: a controlled (laboratory) environment and a public space. Results suggest that the perception of the robot significantly changes depending on the complexity of the interaction but does not change depending on the environment.

Vasiliki Vouloutsi, Klaudia Grechuta, Stéphane Lallée, Paul F. M. J. Verschure

Design and Control of a Tunable Compliance Actuator

TCERA (Tunable Compliance Energy Return Actuator) is a robotic actuator inspired by properties and behavior of the human knee joint, in that it utilizes antagonistic contraction to vary torsional stiffness and joint angle. The actuator is an electrically activated artificial muscle which uses two constant air-mass pneumatic springs configured antagonistically about the knee joint. The positions of the actuator insertion points are controlled in order to change the effective torsional stiffness about the knee axis. Additionally, the rigid member to which the insertion points are attached is able to rotate in order to alter the static equilibrium of the knee system. Furthermore, the system has been simulated and an artificial neural network (ANN) has been trained to determine the control bar angle and attachment location required based on the desired angle, stiffness, and predicted torque state of the joint. Using this controller we have achieved actuator position and stiffness control both with and without load.

Victoria Webster, Ronald Leibach, Alexander Hunt, Richard Bachmann, Roger D. Quinn

An Experimental Eye-Tracking Study for the Design of a Context-Dependent Social Robot Blinking Model

Human gaze and blinking behaviours have been recently considered, to empower humanlike robots to convey a realistic behaviour in a social human-robot interaction. This paper reports the findings of our investigation on human eye-blinking behaviour in relation to human gaze behaviour, in a human-human interaction. These findings then can be used to design a humanlike eye-blinking model for a social humanlike robot. In an experimental eye-tracking study, we showed to 11 participants, a 7-minute video of social interactions of two people, and collected their eye-blinking and gaze behaviours with an eye-tracker. Analysing the collected data, we measured information such as participants’ blinking rate, maximum and minimum blinking duration, number of frequent (multiple) blinking, as well as the participants’ gaze directions on environment. The results revealed that participants’ blinking rate in a social interaction are qualitatively correlated to the gaze behaviour, as higher number of gaze shift increased the blinking rate. Based on the findings of this study, we can propose a context-dependent blinking model as an important component of the robot’s gaze control system that can empower our robot to mimic human blinking behaviour in a multiparty social interaction.

Abolfazl Zaraki, Maryam Banitalebi Dehkordi, Daniele Mazzei, Danilo De Rossi

Extended Abstracts

Motor Learning and Body Size within an Insect Brain Computational Model

Nowadays modeling insect brains is also an important source of inspiration to develop learning architectures and control algorithms for applications on autonomous walking robots. Within the insect brain two important neuropiles received a lot of attention: the mushroom bodies (MBs) and the central complex (CX). Recent research activities considered the MBs as a unique architecture where different behavioural functions can be found. MBs are well known in bees and flies for their role in performing associative learning and memory in odor conditioning experiments [4]. They are also involved in the processing of multiple sensory modalities including visual tasks [3], different forms of learning in choice behavior [5] and also in motor learning [6]. The CX is mainly considered as a center for the initiation of behaviors; it is responsible for visual navigation, spatial memory and visual feature extraction [7,8]

Paolo Arena, Luca Patané, Roland Strauss

Effects of Gaze Synchronization in Human-Robot Interaction

In recent years, humanoid robots have been designed to resemble humans and interact with them like human beings interact with each other in a social setting. Joint attention plays an essential role in human-human social interaction. In this study, based on these findings, we want to implement gaze synchronization to a new game scenario, as a part of the communication process between humans and robots. We describe a method, which will be used in order to study the effects of gaze synchronization as a communication channel on human performance and on trustworthiness during a cooperative cognitive task between humans and a robot.

Stavroula Bampatzia, Vasiliki Vouloutsi, Klaudia Grechuta, Stéphane Lallée, Paul F. M. J. Verschure

Individual Differences and Biohybrid Societies

Contemporary robot design is influenced both by task domain (e.g., industrial manipulation versus social interaction) as well as by classification differences in humans (e.g., therapy patients versus museum visitors). As the breadth of robot use increases, we ask how will people respond to the ever increasing number of intelligent artefacts in their environment. Using the Paro robot as our case study we propose an analysis of individual differences in HRI to highlight the consequences individual characteristics have on robot performance. We discuss to what extent human-human interactions are a useful model of HRI.

Emily C. Collins, Tony J. Prescott

Programming Living Machines: The Case Study of Escherichia Coli

In 1952, Turing outlined computational processes in the morphogenesis [8], thus thinking of the biological evolution of an organism as a consequence of the computation that it can perform. Following Turing’s idea on morphogenesis, many biological processes have been recently analysed from a computational standpoint. In 1995, Bray [2] argued that

a single protein is a computational or information carrying element

, being able to convert input signals into an output signal. Evolution had already been associated with computation many years before, by von Neumann and Burks [9], who constructed a self-replicating cellular automaton with the aim of developing synthetic models of a living organism. Starting from this concept, in this work we propose a relation between computation and metabolism.

Jole Costanza, Luca Zammataro, Giuseppe Nicosia

Force Contribution of Single Leg Joints in a Walking Hexapod

We study leg joint torques of a large insect (

Carausius morosus

) to infer the functions of individual joints in closed kinematic chains during unrestrained walking. Leg joints were found to differentially contribute to multiple locomotor functions of a leg, such as body weight support and propulsion. We conclude that quantifying joint torques in freely behaving hexapods may provide a powerful tool in unraveling the feedback control strategies underlying motor flexibility.

Chris J. Dallmann, Josef Schmitz

High Resolution Tactile Sensors for Curved Robotic Fingertips

Tactile sensing is a key element for various animals that interact with the environment and surrounding objects. Touch provides information about contact forces, torques and pressure distribution and by the means of exploration it provides object properties such as geometry, stiffness and texture[5]. For humans, extracting high level information from touch provides a better understanding of the objects manipulated while for insects it is essential for locomotion[3]. While robot designers have been using vision systems to provide the robot with information about its surroundings, this is not always trivial to obtain, dealing with limited accuracy, occlusions and calibration problems. In terms of sensors for static stimuli, such as pressure, there are a range of technologies that can be used to manufacture transducers with various results[5]. A simple approach is to use fingertips with a 6-DOF force-torque sensor for estimating contact conditions[1], but this only allows a single point of contact and is costly. In terms of fingertip and foot tip prototypes, tactile sensors are used for multi modal sensing, similar to biology, for pressure and dynamic stimuli. In this respect Hosoda et al. [4] propose an anthropomorphic fingertip which has randomly distributed straingauges and PVDF (polyvinylidene fluoride) transducers. In [7] a biomimetic tactile array is proposed that shows a low hysteresis and good sensitivity for skin like deformations.

Alin Drimus, Vince Jankovics, Matija Gorsic, Stefan Mátéfi-Tempfli

Adhesive Stress Distribution Measurement on a Gecko

Gecko adhesion has inspired climbing robots and synthetic adhesive grippers. Distributing loads between patches of adhesive is important for maximum performance in gecko-inspired devices, but it is unknown how the gecko distributes loads over its toes. We report

in vivo

measurements of stress distributions on gecko toes. The results are significantly non-uniform.

Eric V. Eason, Elliot W. Hawkes, Marc Windheim, David L. Christensen, Thomas Libby, Mark R. Cutkosky

Design of an Articulation Mechanism for an Infant-like Vocal Robot “Lingua”

Spoken language is one of the important means for humans to communicate with others. In developmental psychology, it is suggested that an infant develops it through verbal interaction with caregivers by observation experiments [1]. However, what kind of underlying mechanism works for that and how caregiver’s behavior affects on this process has not been fully investigated yet since it is very difficult to control the infant vocalization. On the other hand, there are several constructive approaches to understand the mechanisms by using infant robots with abilities equivalent to those of human infants, as a controllable platform [2].

Nobutsuna Endo, Tomohiro Kojima, Yuki Sasamoto, Hisashi Ishihara, Takato Horii, Minoru Asada

Optimising Robot Personalities for Symbiotic Interaction

The Expressive Agents for Symbiotic Education and Learning (EASEL) project will explore human-robot symbiotic interaction (HRSI) with the aim of developing an understanding of


over long term tutoring interactions. The EASEL system will be built upon an established and neurobiologically grounded architecture -

Distributed Adaptive Control (DAC)

. Here we present the design of an initial experiment in which our facially expressive humanoid robot will interact with children at a public exhibition. We discuss the range of measurements we will employ to explore the effects our robot’s expressive ability has on interaction with children during HRSI, with the aim of contributing optimal robot personality parameters to the final EASEL model.

Samuel Fernando, Emily C. Collins, Armin Duff, Roger K. Moore, Paul F. M. J. Verschure, Tony J. Prescott

A Bio-inspired Wing Driver for the Study of Insect-Scale Flight Aerodynamics

Insect flight studies have advanced our understanding of flight biomechanics and inspire micro-aerial vehicle (MAV) technologies. A challenge of centimeter or millimeter scale flight is that small forces are produced from relatively complex wing motions. We describe the design and fabrication of a millimeter-sized wing flapping mechanism to simultaneously control pitch and stroke of insect and MAV wings. Using micro-fabrication techniques we construct this wing driver and observe that wing motion matches the natural degrees of freedom of insect wings. We actuate wing stroke-position and pitch in open-loop at frequencies relevant to Dipteran and Hymenopteran flight (100-200Hz) and describe the advancements and limitations of this system.

Nick Gravish, Stacey Combes, Robert J. Wood

Characterizing the Substrate Contact of Carpal Vibrissae of Rats during Locomotion

Excitation of sensors triggered by carpal vibrissae has an influence on the kinematics of legs during locomotion of rats. Via motion studies, anatomic and mechanical characterization of vibrissae – especially in the contact period with the substrate – we try to gain a better understanding of the adaptability of those special sensory organs. This knowledge might lead to new approaches for passive sensor systems in robot locomotion or other tactile tasks.

Thomas Helbig, Danja Voges, Sandra Niederschuh, Manuela Schmidt, Hartmut Witte

Self-organization of a Joint of Cardiomyocyte-Driven Robot

In this presentation, we intend to spontaneously realize a joint by utilizing self-organization of cardiomyocytes. The function of the joint is provided by mechanical structure and cell aggregation. The robot was built by culturing neonatal rat cardiomyocytes on a thin collagen sheet whose shape is like a butterfly. The robot could move around the butterfly’s hinge because of the beats of cardiomyocytes, and the aggregation of the cells is self-organized through motion-based mechanical stimulation. After 1 week cultivation of cardiomyocyte-driven robot, cell aggregation emerged around the hinge, and motion of the joint became efficient.

Naoki Inoue, Masahiro Shimizu, Koh Hosoda

Development of an Insect Size Micro Jumping Robot

An insect size micro jumping mechanism is developed and jumps 40cm. The prototype is fabricated with the composite structures cut by precision UV laser. The robot mechanism is bio-mimetic system that is inspired by the small jumping insect, Flea. A single sheet shape memory alloy coil actuator is used for propulsion and energy storage. The compliant mechanism in the body allows to reduce the number of actuators for triggering. The robot mechanism has 36mg weight, 2 cm length and 2mm height except of wire legs.

Je-Sung Koh, Kyu-jin Cho

Soil Mechanical Impedance Discrimination by a Soft Tactile Sensor for a Bioinspired Robotic Root

During the penetration into the soil, plant roots experience mechanical impedance changes and come into contact with obstacles which they avoid and circumnavigate during their growth. In this work, we present an experimental analysis of a sensorized artificial tip able to detect obstacles and discriminate between different mechanical impedances in artificial and real soils. The conical shaped tip is equipped with a soft capacitive tactile sensor consisting of different elastomeric and conductive layers. Experimental results show that the sensor is robust yet sensitive enough to mechanical impedance changes in the experimented soils.

Chiara Lucarotti, Massimo Totaro, Lucie Viry, Lucia Beccai, Barbara Mazzolai

Fetusoid35: A Robot Research Platform for Neural Development of Both Fetuses and Preterm Infants and for Developmental Care

We have been developing a robot called Fetusoid35 that resembles a human fetus or preterm infant. We suppose that the robot could contribute to developmental science by shedding a new insight on the understanding the developmental process of fetuses and preterm infants. Based on the mechanism, we would expect to improve a developmental care of preterm infants. This extended abstract briefly introduces the design policy of Fetusoid35 with its specifications and the current status.

Hiroki Mori, Daii Akutsu, Minoru Asada

High Speed Switched, Multi-channel Drive for High Voltage Dielectric Actuation of a Biomimetic Sensory Array

Electro-Active Polymers (EAP) have been described as artificial muscles due to their composition andmuscle-like dynamics [1]. Consequently they have attracted a lot of attention from the biomimetic robotics research community and heralded as a potential alternative to conventional electromagnetic, pneumatic or hydraulic actuation technologies [2]. However, in practice there are a number of technical barriers to overcome before they gain widespread acceptance as robotic actuators [3]. Here we focus on overcoming one of those limiting factors for a type of EAP referred to as Dielectric Electro-Active Polymers (DEAP).

Martin J. Pearson, Tareq Assaf

A Combined CPG-Stretch Reflex Study on a Musculoskeletal Pneumatic Quadruped

Quadruped animals combine the versatility of legged locomotion with extra stability from the additional number of limbs. Cats and dogs can walk in different gait patterns, outperforming current robotic technology while exploiting interactions between brain, spine, muscle and environment, which inner workings are not fully understood. We propose a controller which combines a rhythmic sinusoidal pattern (feed-forward) with a muscular stretch reflex (feedback) on a biomimetic musculoskeletal robot.

Andre Rosendo, Xiangxiao Liu, Shogo Nakatsu, Masahiro Shimizu, Koh Hosoda

Swimming Locomotion of Xenopus Laevis Robot

An adaptive swimming locomotion of

Xenopus laevis

are mainly generated through hydrodynamic interaction between its musculoskeletal system and water environments. To understand the mechanism of frog locomotion, therefore, it is a promising approach to copy morphology of the frog and let it swim in the water. We developed a swimming robot that had a similar musculoskeletal structure as a frog driven by living muscles. We realized kick motion that generated propulsion for swimming locomotion by exciting the living muscles. The robot is expected to be a powerful tool to understand the swimming locomotion of

Xenopus laevis


Ryo Sakai, Masahiro Shimizu, Hitoshi Aonuma, Koh Hosoda

Empathy in Humanoid Robots

Humanoid robots should be able to interact with humans in a familiar way since they are going to play a significant role in the future. Thus, it is necessary that Human-Robot Interaction (HRI) is designed in such a way that allows humans to communicate with robots effortlessly and naturally. Emotions play an important role in this interaction since humans feel more predisposed to interact with robots if they are able to create an affective bond with them. In this study, we want to know whether humans are able to empathize with a humanoid robot. Therefore, in the present research, we are going to recreate a Milgram experiment in which we expect participants to empathize with the robot while playing a matching game. Like in Milgram’s experiment, they will have to give fake electrical shocks to the robot thinking that they are punishing it. In that way, an empathic state, which we expect to see in our results, may be induced.

Marina Sardà Gou, Vasiliki Vouloutsi, Klaudia Grechuta, Stéphane Lallée, Paul F. M. J. Verschure

HECTOR, A Bio-Inspired and Compliant Hexapod Robot

The newly built and currently tested hexapod robot HECTOR is introduced. The robot consists of 18 embedded, custom designed and compliant joint drives based on an integrated elastomer coupling.

Axel Schneider, Jan Paskarbeit, Malte Schilling, Josef Schmitz

Gesture Recognition Using Temporal Population Coding and a Conceptual Space

The processing of visual information requires a robust way of encoding stimuli that is able to extract key features for further stimulus assessment while staying invariant to disturbances. We propose a gesture recognition system based on a temporal population code to transform gestures into elements in a conceptual space.

Jan Niklas Schneider, Stéphane Lallée, Paul F. M. J. Verschure

Roving Robots Gain from an Orientation Algorithm of Fruit Flies and Predict a Fly Decision-Making Algorithm

Simple organisms like bacteria are directly influenced by momentary changes in concentration or strength of sensory signals. In noisy sensory gradients frequent zigzagging reduces the performance of the cell or organism.

Drosophila melanogaster

flies significantly deviate from a direct response to sensory input when orienting in gradients. A dynamical model has been derived which reproduces fly behaviour. Here we report on an emergent property of the model. Implemented in a robot, the algorithm is sustaining decisions between visual targets. The behaviour was consequently found in wild-type flies, which stay with a once-chosen visual target for considerable longer times than mutant flies with a specific brain defect. This allowed the localisation of the integrator. Flies were tested in a virtual-reality arena with two alternatingly visible target objects under different visibility regimes. The finding exemplifies how basic research and technical application can mutually benefit from close collaboration.

Roland Strauss, Stefanie Flethe, José Antonio Villacorta, Valeri A. Makarov, Manuel G. Velarde, Luca Patané, Paolo Arena

The Si elegans Project – The Challenges and Prospects of Emulating Caenorhabditis elegans

Caenorhabditis elegans

features one of the simplest nervous systems in nature, yet its biological information processing still evades our complete understanding. The position of its 302 neurons and almost its entire connectome has been mapped. However, there is only sparse knowledge on how its nervous system codes for its rich behavioral repertoire. The EU-funded

Si elegans

project aims at reverse-engineering

C. elegans

‘ nervous system function by its emulation. 302 in parallel interconnected field-programmable gate array (FPGA) neurons will interact through their sensory and motor neurons with a biophysically accurate soft-body representation of the nematode in a virtual behavioral arena. Each FPGA will feature its own reprogrammable neural response model that researchers world-wide will be able to modify to test their neuroscientific hypotheses. In a closed-feedback loop, any sensory experience of the virtual nematode in its virtual environment will be processed by sensory and subsequently interconnected neurons to result in motor commands at neuromuscular junctions at the hardware-software interface to actuate virtual muscles of the virtual nematode. Postural changes in the virtual world will lead to a new sensory experience and thus close the loop. In this contribution we present the overall concepts with special focus on the virtual embodiment of the nematode. For further information and recent news please visit

Axel Blau, Frank Callaly, Seamus Cawley, Aedan Coffey, Alessandro De Mauro, Gorka Epelde, Lorenzo Ferrara, Finn Krewer, Carlo Liberale, Pedro Machado, Gregory Maclair, Thomas-Martin McGinnity, Fearghal Morgan, Andoni Mujika, Alessandro Petrushin, Gautier Robin, John Wade

A Self-organising Animat Body Map

Self-organising maps can recreate many of the essential features of the known functional organisation of primary cortical areas in the mammalian brain. According to such models, cortical maps represent the spatial-temporal structure of sensory and/or motor input patterns registered during the early development of an animal, and this structure is determined by interactions between the neural control architecture, the body morphology, and the environmental context in which the animal develops. We present a minimal model of pseudo-physical interactions between an animat body and its environment, which includes each of these elements, and show how cortical map self-organisation is affected by manipulations to each element in turn. Initial simulation results suggest that maps robustly self-organise to reveal a homuncular organisation, where nearby body parts tend to be represented by adjacent neurons.

Hiroki Urashima, Stuart P. Wilson

A Novel Bio-inspired Tactile Tumour Detection Concept for Capsule Endoscopy

Examination of the gastrointestinal(GI) tract has traditionally been performed using endoscopy tools that allow a surgeon to see the inside of the lining of the digestive tract. Endoscopes are rigid or flexible tubes that use fibre-optics or cameras to visualise tissues in natural orifices. This can be an uncomfortable and very invasive procedure for the patient.

Benjamin Winstone, Chris Melhuish, Sanja Dogramadzi, Tony Pipe, Mark Callaway

Electro-communicating Dummy Fish Initiate Group Behavior in the Weakly Electric Fish Mormyrus rume

The mechanisms that underlie collective behavior in groups of fish have been the subject of numerous quantitative and theoretical studies. We use a robotic platform to investigate social interactions in weakly electric fish by exploiting their unique electro-sensory modality for animal-robot communication. Our results demonstrate that weakly electric fish interact with a mobile dummy fish based on species-specific electrical playback signals and are therefore proposed to be a promising model organism for establishing a mixed-society of real and artificial fish.

Martin Worm, Tim Landgraf, Hai Nguyen, Gerhard von der Emde

A Concept of Exoskeleton Mechanism for Skill Enhancement

While common implementation of the exoskeleton mechanism is “powered suit” with onboard power plant, it suffers low power-to-weight ratio and often fails to inefficacy problem. On the other hand, quasi-passive mechanism can exploit human efficiency further, which can be regarded as a “wearable bicycle”. In this approach, we are developing an exoskeleton mechanism as a hybrid system for improving users’ skills.

Tomoyuki Yamamoto, Hiroshi Ishiguro

Erratum: Individual Differences and Biohybrid Societies

In the original publication of the paper the author “Abigail Millings” was erroneously omitted from the author list. It should read as:

Emily C. Collins, Abigail Millings, and Tony J. Prescott

Department of Psychology, The University of Sheffield, UK


Furthermore, the following acknowledgement text was not included by mistake:

This work was supported by the European Union Seventh Framework Programme (FP7-ICT-2013-10) under grant agreement n° 611971 (EASEL: Expressive Agents for Symbiotic Education and Learning).

Emily C. Collins, Abigail Millings, Tony J. Prescott


Weitere Informationen

Premium Partner