HomePage

Angels

ANGuilliform robot with ELectric Sense

AIM

The aim of the ANGELS project is to investigate interactions between body morphology and behaviour by designing and building a prototype for a reconfigurable anguilliform swimming robotic system. The robot will be able to function either as an eel-like whole entity, or may split into smaller agents (and vice-versa). These two different morphological forms will use a bio-inspired “electric sense”, both for recognition of objects and obstacles, and for communication between agents, and will explore and exploit the different swimming and electro-sensing strategies used by gymnotid and mormyrid electric fish. Such a robotic system is intended to be used for recognition of objects in environments where vision –for perception– and propellers –for locomotion– are not suitable due to murky water, industrial waste, sea weeds, etc.

CONCEPT

Anguilliform swimming is a mode of locomotion used by elongated fish (eels, lampreys, moray eels etc.) in which a travelling undulation is propagated along the body from head to tail. Such a swimming mode can be used in a large variety of fluidic environments. Similarly to elongated fish, the anguilliform robots to be designed in this project will have an elongated “snake-like” structure of variable length. With a body design of this sort, morphology is closely linked to behavior. Bending or twisting of the elongated body shape will change the spatial relationship between electric emitters and receptors and this can be exploited with advantage for sensory scanning in a variety of behaviors. The novel “electric sense” with which the ANGELS robot will be equiped is used by several families of fishes either passively (exploiting external electric fields) or actively (producing their own field), for navigation and communication in murky water and at night when vision is of little use. In its principle, the active electric sense is based on a mechanism known as “electrolocation”: by polarization of a region of one body with respect to another, the fish generates an electric field in its surrounding world, and measures the induced trans-cutaneous currents that flow back through electro-sensors (also named “receivers”) distributed over the skin. From differences between expected currents, supposing a clear environment, and those actually measured, the fish can infer shapes, position and electrical properties of nearby objects (von der Emde & Fetz 2007). The sense perceives in all directions, thus creating a bubble-like area of perception, as wide as the fish’s length.

Electric fishes provide ideal paradigms for studying the interactions between morphology, perception, and locomotion since:

while swimming, the active surface of their body is the “electric retina” through which they perceive the world;
they actively modulate their body geometry to shape the electric field in order to better perceive the details of their environment;
different species present a wide variety of spatial organizations of “electric emitters” and “receptors” along their body, combined with a wide variety in the frequency and form of electric carriers (pulsed, undulated);
the electric carrier signal serves for both electrolocation and for communication between individuals;
the electric field actually creates an immaterial prolongation of the material body sensing the surrounding environment

This last point introduces a completely new concept in that the “electric body” gives the electric field a kind of materiality, much as predicted by Physicists exploring “Fields Theory”. Beside demonstration purposes, the robotic prototype will act as a theoretical paradigm and an experimental platform to study the interactions between morphology, perception and locomotion through all the possibilities offered by:

Close interactions between perception and action, where action is:
1. emitting electric fields (e.g. the frequency of pulses for pulsed emissions),
2. changing body shape (e.g. bending around an object),
3. exploratory swimming movements.
Morphological reconfiguration of the robotic system, where reconfiguration can be of two types:
1. changing the role of the distributed electrodes, i.e. switching them from being emitters to receivers or vice versa, effectively changing the location and the shape of the electric field along the body,
2. attaching and detaching sub-agents, such as to be able to switch between a connected morphology (one single long robot) and a disconnected morphology (group of smaller robots).
Multi-agent perception and communication, where the possibility to switch between a connected and disconnected morphology can be useful to
1. improve perception. Temporarily switching to a distributed group of smaller robots can for instance be advantageously used to increase the perception range through multi-agent communication, reinforce electric fields at some specific point (shaping the field and “zooming” on a detail of the object by creating a multi-polar lens for instance), delocalize emitting and receiving functions and so exploiting the possibilities offered by passive sensing, and to multiply points of view – therefore introducing ubiquity,
2. improve locomotion. Depending on the situation it might be more advantageous to swim as a single long robot (e.g. for efficiency in free water) or as a group of smaller robots (e.g. for maximizing the exploration range or for manoeuvrability in confined regions),
3. improve cognition. By programming some artificial behaviours inspired from the fishes, in particular sociological rules from which could emerge efficient collective perception.

The ultimate goal of the project will be to tackle the following type of scenario, as illustrated in Figure 1. An object “of interest” is placed in a swimming pool among others “of no interest”. The anguilloform robot is dropped in the water. It splits into four smaller robots that swim in various directions, randomly searching for the object of interest. Once one finds it, it sends a strong electric communication signal that the others use passively for guidance. Once reunited, the four smaller robots swim around the object to gather the best possible electric picture of it. For doing this, they combine electrical sensing, swimming, body deformations and topological and electrical reconfiguration (depending for instance on the size of the object). When finished, they re-attach themselves in a single long robot, and return to their starting point with a certain knowledge of the recognized object. In this type of scenario we will use the modalities of fish electric sensing which allow them to recognize shape, but also a certain kind of electric color conferred by a combination of conductance and capacitance. For instance, the “objects of interest” will have a priori a certain electrical color and, the robot will be required to recognize their shapes. The type of scenario proposed above is voluntarily simple. In fact, the given task is more a bench-mark allowing one to quantitatively evaluate the performances of the different possibilities offered by the ANGELS robot. This simplicity is also required to allow biologists to make comparisons and to draw parallels with their ethological observations. Beyond this type of scenario, there exist many concrete applications which could benefit from the richness of solutions offered by ANGELS, for instance, the exploration of industrial sewers containing dangerous materials.

Figure 1: Example of a recognition scenario using the anguiliform robotic system. The cross indicates the starting point, and the circle is an object of interest. The gray boxes represent obstacles. See text for details.

OBJECTIVES

The project has both engineering and scientific objectives.
The engineering objectives are the following:

To develop a new sensor for object recognition based on a new sense for engineers: the electric sense
To develop new hardware for the electric sense. The hardware will involve a system of distributed electrodes and the electronics necessary (a) for online changing the role of the electrodes (emitters, or receivers) in order to reconfigure in real time the emitter and receptor arrays (spatial electric shaping), (b) for generating pulsed or undulated signals (electric shaping through modulation of frequency and form), (c) amplifying and processing the measured signals.
To develop a new dynamic connector for runtime attachments and detachments. The dynamic connector will allow the robots to mechanically attach one to the other in series, while preserving the electric continuity between modules and to exchange data.
To fit these new devices into a new Anguilliform reconfigurable swimming robot prototype, whose design will be based on the mutualisation of the experience acquired from the prototypes of EPFL, SSSA and ARMINES, on which some of these devices will be implemented for testing and tuning.
To develop multiphysical models and simulators coupling fluid and body dynamics with models of the electric sense, and that can tackle multi-agent scenarios.
To develop algorithms for active perception using the electric sense and its reconfiguration abilities.
To develop swimming controllers that can tackle a changing body structure.
To implement some recognition scenarios as this described above.

The scientific objectives are:

To explore interactions between morphology and behaviour in electric fish.
To explore group behaviour and communication in groups of electric fish.
To investigate how morphological (electric and mechanic) reconfiguration can help perception
To investigate how to design for emergence, namely how to design local control and interaction rules for individual sub-agents such as to obtain robust behaviour at the group level.
To explore how the group of robots should synchronize and switch between connected and disconnected morphologies such as to optimize swimming and perception according to various criteria (e.g. energy efficiency, agility, volume of exploration, size of the perceived objects).
To investigate the opportunities offered by a disconnected morphology through multi-agent perception and communication

Relevance to the FET embodied intelligence call

This project is highly relevant to the FET embodied intelligence call for three reasons: (1) it addresses the relationship and interaction between morphology and behaviour in a robot that embodies bio-inspired intelligent principles of locomotion and sensory exploration, (2) it will produce innovative technology by exploring the artificial implementation of a biological sense that has been very little exploited in robotics to date, and (3) the highly interdisciplinary composition of the consortium will extend the capacity of research in both engineering and biological sciences.

Research topics

The project directly addresses two core topics of the call, namely the interaction between morphology and behaviour, and the design for emergence.
Interaction between morphology and behaviour. The project addresses several key issues related to the interaction between morphology and behaviour such as (1) a new sense (the electric sense) that is strongly dependent on the body shape, (2) a morphology (here extended to that of the ‘electric body’) that changes over time depending on environment through locomotion (mechanical reconfiguration) and perception (electrical reconfiguration), and (3) the possibility to have an entity that is at multiple locations at the same time (i.e. some kind of ubiquity) and that uses the electric sense both for perceiving its surroundings and for communication.

Innovations

Through ANGELS, we expect multiple advances both in technology and in science. This project targets technological and conceptual breakthroughs (new sense, new paradigms), and will lead the consortium to explore completely new territories. Hence ANGELS has the immediate potential for many “medium risk/high gain” innovations.
Compared to the present existing solutions for Autonomous Underwater Vehicles (AUV) exploration (including object recognition), ANGELS first exploits the advantages of the Anguilliform swimming which is the optimal morphology that nature has found for solving the trade-off between swimming efficiency, agility (manoeuvrability) and morphological simplicity (eels appeared quite early in the phylogenesis of species).
Moreover, based on the morphological (electric and mechanical) reconfiguration abilities and multi-agent cooperation (for perception and action), and on all their possible combinations, ANGELS will propose more efficient solutions than those today used for AUV. These solutions will be concretely evaluated and compared in terms of energy consumption, richness of object recognition, or other speed for finding targets and executing a giving task or mission… These technological solutions will achieve a first step towards a new generation of AUV based on the embodied intelligence paradigm for circumventing the difficulties met by today solutions.

Interdisciplinary research

Angels is a fundamentally interdisciplinary project. It involves three communities that have not worked together previously in Europe: those of Roboticians and electric fish Biologists. Furthermore, to our knowledge, fundamental Physics has rarely been combined with Robotics. This interdisciplinary research is the necessary characteristic of a project whose fundamental purpose is to contribute to the design of a new generation of autonomous robots inspired from the intelligence encoded in animals’ morphology by the evolution laws over millions of years. Starting from this basic motivation and observing the very high degree of sophistication and integration of technological solutions that nature invented to solve its problems; it becomes urgent to transfer our most advanced technologies as those of fundamental physics, to a new bio-inspired robotics.