VirtualEnaction

A Platform for Systemic Neuroscience Simulation

Nicolas Denoyelle

1,2

, Florian Pouget

1,2

, Thierry Vieville

and Fr

eric Alexandre

1,2,3

Inria Bordeaux Sud-Ouest, 200 Avenue de la Vieille Tour, 33405 Talence, France

LaBRI, Universit

e de Bordeaux, Institut Polytechnique de Bordeaux, CNRS, UMR 5800, Talence, France

Institut des Maladies Neurod

eratives, Universit

e de Bordeaux, CNRS, UMR 5293, Bordeaux, France

Keywords:

Simulation, Computational Neuroscience, Virtual Reality.

Abstract:

Considering the experimental study of systemic models of the brain as a whole (in contrast to models of one

brain area or aspect), there is a real need for tools designed to realistically simulate these models and to exper-

iment them. We explain here why a robotic setup is not necessarily the best choice, and what are the general

requirements for such a bench-marking platform. A step further, we describe an effective solution, freely avail-

able on line and already in use to validate functional models of the brain. This solution is a digital platform

where the brainy-bot implementing the model to study is embedded in a simpliﬁed but realistic controlled

environment. From visual, tactile and olfactory input, to body, arm and eye motor command, in addition to

vital somesthetic cues, complex survival behaviors can be experimented. The platform is also complemented

with algorithmic high-level cognitive modules, making the job of building biologically plausible bots easier.

1 INTRODUCTION

The brain is a fascinating complex structure and de-

signing global models of such a structure is particu-

larly difﬁcult. This is speciﬁcally true with regard to

the fact that the brain is a complex system in interac-

tion with the body and the environment. Two impor-

tant consequences can be driven. On the one hand,

modeling the brain includes not only understanding

how each subsystem (visual, motor, emotional, etc.)

works but also how these subsystems interact as a

whole, to yield emerging behaviors, i.e. effects that

result from interactions between subsystems. On the

other hand, studying and validating functional models

of brain structures at a macroscopic scale cannot be

performed with restrained artiﬁcial static paradigms

but requires experiments in complex environments,

with realistic sensory-motor tasks to perform, includ-

ing high-level interactive behaviors (e.g. survival

strategy in the presence of prays/predators) and long-

term protocols (since both statistical studies and bio-

plausible learning mechanisms require long epochs).

Such paradigms are to be related to biological ex-

periments conducted on animals. These statements

are not only characterizing brain models, they also

give strong requirements on the tools that must be

designed to simulate these models and to experiment

them.

Designing such tools is also an excellent way to

address at the same time the two main objectives of

such brain models at the macroscopic scale. One the

one hand, they are intended to serve neuroscientists

as a new platform of experimentation, on which they

can apply their classical protocols of observation and

analysis of animals at the behavioral as well as elec-

trophysiological levels. It is consequently important

that neuroscientists can observe the inner activity of

the models, as they use to do for example with elec-

trodes (but we can imagine that this observation in

digital models might be more easy than in the real

brain). It is also important that they can deﬁne clas-

sical behavioral protocols like they do in animals (eg.

fear conditioning) in order to observe the resulting be-

havior and the corresponding brain activation. Deﬁn-

ing such protocols implies that the structure of the ex-

ternal world (e.g., maze, food magazine) as well as its

intrinsic rules (eg. tone followed by an electric shock)

be easy to design.

On the other hand, they are also intended to serve

computer scientists as a way to design artiﬁcial au-

tonomous systems, driven by the brain models. In

this case, it is important that the supposed properties

of the models (e.g., capacity to learn, robustness to

noise or changing rules) be assessed by rigorous eval-

155

Denoyelle N., Pouget F., Vieville T. and Alexandre F..

VirtualEnaction - A Platform for Systemic Neuroscience Simulation.

DOI: 10.5220/0005166701550163

In Proceedings of the 2nd International Congress on Neurotechnology, Electronics and Informatics (-2014), pages 155-163

ISBN:

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

uation procedures, as it is deﬁned for example in the

domain of machine learning. In this case also an easy

access must be proposed both to the inner circuitry

of the models and to the speciﬁcation of the external

world.

With in mind this double goal of offering conve-

nient tools to both scientiﬁc communities, we report

in this paper the speciﬁcations that we have elaborated

and present the corresponding software platform that

we call VirtualEnaction.

2 PROBLEM POSITION

Concerning the nature of such a simulator, real

robotic systems are often used and answer particu-

larly well to the second requirement about a realistic

environment. However, building viable robotic sys-

tems and making them evolve in realistic environ-

ments (e.g. natural sites) for long periods of time

(e.g. several days) is just too expensive in term of

cost and manpower in many circumstances and par-

ticularly during early phases of development. Fur-

thermore, the goal of such simulation is not only to

make a demo, but also, and more importantly, to study

and quantify the behavior of functional models of the

brain. As a consequence we not only need a com-

plex, long-term, realistic experimental setup, but we

also need a controllable and measurable setup where

stimuli can be tuned and responses can be measured.

In fact, real environment complexity and parameters

are intrinsically difﬁcult when not impossible to con-

trol. This is the reason why we propose to use a digi-

tal simulator implementing realistic survival and other

biological scenarios

A step further, available macroscopic models of

brain functions are not designed for ”performance”

but to properly implement phenomenological con-

cepts that have been investigated in some cogni-

tive or behavioral framework. They would therefore

have ”no chance” in a real world. Note that recent

computer science mechanisms designed without any

constraint regarding biological plausibility but only

towards ﬁnal performances are nowadays probably

more efﬁcient but explain nothing.

As a consequence we also need a setup which can

provide a “simpliﬁed environment”, in order systemic

models of the brain at the state of the art not to fail

immediately. We must also take into account the fact

that (i) such models are rather slow to simulate (un-

less huge computer power is available), and that (ii)

they are not supposed to focus on precise issues re-

garding low-level sensory input or motor output but

on integrated cognitive functions and the resulting be-

haviors.

This, in addition to technical constraints, yields

three key characteristics:

1. No real-time but a look-and-move paradigm : The

main restriction we propose to accept here is to

have the simulator running at a “slower” time (i.e.

using several seconds to simulate one real-time

second) and also to consider discrete time sam-

pling. This seems obvious as far as digital sim-

ulation is concerned, but in terms of underlying

framework, this has several consequences (e.g.,

giving up the possibility for a human observer to

interact with the simulation, restraining to clock-

based (and not event-based) dynamical models,

etc.) (Taouali et al., 2011).

2. No real robotic control but only motor command :

Since in the nervous system motor control seems

to be a hierarchical system with high-level motor

commands, while their closed loop execution is

delegated to the peripheral motor system (Uithol

et al., 2012), we may accept to only simulate ges-

ture and displacement at a rather symbolic level

such as “stand-up” or “turn 90

◦

rightward”. This

indeed cancels the possibility to study sharp phe-

nomena of motor interactions with the environ-

ment but allows us to concentrate on high-level

control such as action selection mechanism and

motor planiﬁcation.

3. Complex but not necessarily natural visual envi-

ronment: The third main restriction we propose

to accept is to consider a complex visual environ-

ment (with visual textures, several objects in mo-

tion, etc.) but not to invest in the simulation of a

realistic natural scene simulation. The reason of

this choice is that natural image vision is an issue

already well studied (Hyv

arinen, 2009). The gen-

eral conclusion is that biological visual systems

are tuned to natural image statistics, decomposed

by the early visual front-end in such a way that

higher-level visual input only relates on cues or-

thogonal (in a wide sense) to natural image statis-

tics. In other words, the job regarding this aspect

is done by early-vision layers and we may con-

sider more stylistic visual cues at a higher-level.

Depending on the study, we may also wish to

work on either a pixelic or a symbolic represen-

tation of the visual scene. See (Teftef et al., 2013)

for details of how the early-visual system relates

both representations.

NEUROTECHNIX2014-InternationalCongressonNeurotechnology,ElectronicsandInformatics

156

3 SYSTEM DESCRIPTION

We consider that a “brainy-bot”, i.e. the implementa-

tion of a global model of the brain functionality, inter-

acts with its environment with the simple goal to sur-

vive. Our objective is to simulate the sensory-motor

interactions of this bot with respect to its environment.

Examples of such surroundings are shown in Fig. 1.

Figure 1: Two examples of digital experimental environ-

ments for systemic neuroscience. Up: A minimal environ-

ment corresponding to a standard maze reinforcement learn-

ing task (source: one of our virtual enaction built). Down:

A complex environment in which survival capabilities are

to be checked (source: landscape encountered when play-

ing with the standard game).

Survival is precisely deﬁned as maintaining vital

variable values in correct ranges, as formalized in,

e.g., (Friston, 2012). In our context, health, food, wa-

ter, energy, and oxygen are the vital state variables.

The bot has access to these values. These variables

decrease or increase with time since the bot body is

supposed to consume the related resource depending

on its activity, or change in the presence of an exter-

nal event (e.g. energy during a predator attack), and

restore resources. Restoring resources is obtained ei-

ther by ingesting items or taking rest (i.e., make the

choice to stop action, with the beneﬁt of vital resource

increase and the drawback of not acting on the en-

vironement, this might be a short-term policy choice

if vital variables are low and is a middle-term policy

choice otherwise).

The environment structure is very simple and

made of “blocks”. Each object in this environment

(including the ﬂoor, relief, ..) is a collection of blocks.

Each block is deﬁned by its 3D position, orienta-

tion and sizes, roughness, hardness, temperature, and

color. Some blocks correspond to eatable or drinkable

resources. Other entities correspond to objects that in-

teract with the bot (e.g. predators that attack the bot

or lava to avoid). At this level, survival corresponds

to avoid or kill the predators and ﬁnd resources to eat

or drink.

The bot anatomy is functionally made of a body,

an arm and an head/eye. It carries a bag with objects

found in its environment. The body can move at a

given speed and in a given direction, and also rotate

at each step to a given angle. It can also jump up to a

given relative height, or knee down to take a rest. The

head/eye can gaze in a given yaw/pitch direction. The

arm can perform predeﬁned symbolic gestures : take

an object in hand out of its bag, put the object in hand

into its bag, either drop or throw the object in hand,

ingest the block in hand (food or water), grasp the

object in front of him. The arm can also attack (quan-

tiﬁed by a force value and with or without an object in

hand) the object in front of him. This is the complete

description of the bot motor command output in the

present context.

The bot sensory input corresponds to cues related

to the blocks which are around it. The touch cues al-

low the bot to estimate the roughness, hardness and

temperature of the object in hand. The olfactory cues

allow to estimate the smell type and intensity of ob-

jects close to it (computed by integrating average val-

ues over the blocks characteristics). At the bot level,

pixelic vision provides an image of the visual ﬁeld

view (i.e., calculating the blocks texture and color

projection on the virtual retina). Finally, the proprio-

ceptive cues correspond to gaze direction estimation.

In order to quantify the bot behavior, the interface

provides an additional access to the bot absolute posi-

tion and orientation in space. Symbolic vision is also

available, as a list of blocks visible in its visual ﬁeld,

which access to the block characteristics.

An adaptation of the minecraft open game soft-

ware yields the proper answer to this wish-list and the

so called virtualenaction is an open-source free-

license implementation of these speciﬁcations. Each

user buys a end-user low-cost mojang license (< 20$)

for minecraft, while virtualenaction is free of

use under a CeCILL-C license. Fully-documented

scripts facilitate the installation of the software bun-

dle under Linux OS. The bot is implemented in ei-

ther C/C++, or possibly in Python (via an existing

swig wrapper) or other computer languages. It uses

a simpleAPI, as described in Fig. 2. Furthermore,

in order to both observe in slow-down real-time the

VirtualEnaction-APlatformforSystemicNeuroscienceSimulation

157

Figure 2: Left: The software interface is trivial: each imple-

mentation of a brainy-bot provides an initialization routine

initBot() and a stop function stopCondition(), while

at each time-step the brainDo() method is called. Right:

All status, input and output functionality are available via

a simple API. For instance, hand or vital input, body and

head displacement, gestures of resource injection and attack

against predator are shown.

bot behavior and interact with the digital experiment,

a graphic user interface is available as described in

Fig. 3.

More details on the computer implementation is

given in Appendix 5.

4 NEUROSCIENCE

APPLICATION

Let us now discuss how this setup constitutes a step

towards integrative neuroscience digital experimenta-

tion.

First of all, let us compare this project with com-

plementary connected projects. The AnimatLab is a

software platform allowing to simulate embodiment

(bio-mechanical simulation of a body) allowing in-

vestigate the relation between brain and body (Cofer

et al., 2010). Furthermore, it proposes a neural net-

work architecture for the implementation of cognitive

function. On the contrary, the present framework has

a rather limited description of the embodiment, but a

much larger set of possible interactions with the en-

vironment. A step further, not only artiﬁcial based

neural network models are usable in VirtualEnaction,

whereas the interface with any existing neuroscience

simulation tool (e.g., python based neural simulators,

see (Brette et al., 2007; Davison et al., 2008)) is

straightforward. This feature is essential, since we

must simulate the system at different modeling scale,

as developed now. The Morse is a generic simula-

tor for academic robotics, with realistic 3D simulation

of small to large environments, allowing complete in-

tegration with any simulation tools. It outperforms

concurrent systems like Webot. The interest of Vir-

tualEnaction with respect to Morse is twofold: Since

we target integrative cognitive tasks of survival which

is exactly what happens with the Minecraft environ-

ment, using this specialized product is far simpler

and somehow more demonstrative. In term of per-

formances, as being less sophisticated (using a sim-

pliﬁed 3D rendering, while Morse has all 3D capa-

bilities) and being agnostic in terms of programming

languages (i.e., allowing fast C/C++ implementation

of user modules, whereas Morse is limited to Python

scripts) the VirtualEnaction platform is a priori ex-

pected to be more efﬁcient in terms of CPU usage.

However, with a larger humanpower all what has been

developed within VirtualEnaction could have been de-

veloped in Morse.

The main application regarding neuroscience is to

test cognitive computational models in realistic con-

ditions. Very simply, a behavioral experiment is per-

formed on an animal model or on human. Usually

with a training phase, the measurement phase and the

data analysis. In order to formalize the obtained re-

sult a computational model is proposed that explain

the data, and may also have prediction regarding other

falsiﬁable future experiments. The present software

and methodological tool allow us to propose to en-

hance this very general paradigm in the following di-

rections:

• Test the model prediction for several others ex-

perimental conditions or model parameter ranges :

The idea is to reproduce the experimental setup in

this virtual environment (e.g., a delayed reward

task, an exploration paradigm) and connect the

computational model to this paradigm. As for

usual computational modeling, the chosen biolog-

ical measurements (e.g. neural activity, task suc-

cess performance) are simulated when running the

model. Such model is indeed expected to repro-

duce qualitatively the ground truth, for a given set

of parameters value. A step further, it is very im-

portant to numerically verify what happens when

modifying any quantitative or qualitative parame-

ter value. If the numerical sensibility is so strong

that the results cannot be reproduced for some tiny

parameter variation, the model is meaningless be-

cause biological values make sense as a numerical

range, not a single number. If the numerical sen-

sibility is so weak that any parameter value pro-

duces the expected result, this parameter is mean-

ingless and a simpler model very likely explains

the same data set. The key-point here is that such

usual model predictive veriﬁcation is not only go-

ing to be possible, given a ﬁxed data set, but for

any data set obtained running experiment in the

virtual environment. In other words, the computa-

tional model variants are going to be always tested

in-situ. With no practical bounds on the experi-

mental variants (e.g., number of trials, sensory in-

NEUROTECHNIX2014-InternationalCongressonNeurotechnology,ElectronicsandInformatics

158

Figure 3: The software’s graphic user interface. In order to observe and control the experiment, a view of the bot vision,

status, input and output is available via interactive panels. It is also possible to “cheat” by controlling these variables values,

in order to debug an experiment and understand in details what happens in such a complex system interaction.

put precision, task complexity).

• Design new experimental paradigms to confront

the computational model to falsiﬁable conditions :

Building on an animal model an experiment,

training the animals for weeks, restarting from

the beginning, if it appears that there is an unex-

pected trap (e.g., the task in too simple or unfeasi-

ble, or does not allow to discriminate between two

concurrent models) may be a huge work. Start-

ing to design the experiment in a virtual envi-

ronment completely changes the method: the hy-

pothesized computational model is ﬁrst tested in

silico (i.e., neither in vivo, nor in vitro, but in a

software environment) and only confronted to the

biological reality in a second stage. The work-

plan is inverted with respect to usual neuroscience

studies, but in computational engineering (e.g.,

designing new airplanes) this is exactly the way

it goes until a few decades. It is however not

new in neuroscience, at the scale of mesoscopic

brain map study (see e.g., (Chemla et al., 2007) or

(Brette et al., 2007)). The key point here, is that

such approach is now possible at the behavioral

level, considering sensory-motor interactions with

a simple or complex environment. Such process

also obviously yields a parsimonious use of ani-

mal models. A step further, it lays down the chal-

lenge to perform realistic experiment with a com-

putational model of the brain behavior, not only

consider toy situations where the plausibility of

the model can not be checked.

The degree of equivalence between simulation

outcome and neuroscience experimental results is key

issue. This is the reason why we have chosen a soft-

ware platform where usual systemic neuroscience ex-

periments can be reproduced “as a whole”, as illus-

trated by several examples throughout this draft. Ba-

sis omissions (e.g., level of modeling detail) have

been justiﬁed or rejected, in order to attain this objec-

tive of simulating classes of behavioral experiments

where the subject is trained to realize a survival or

rewarding task in a known or unexpected environ-

ment. Applications include Pavlovian (e.g., action

of the amygdala) or operant conditioning, reinforce-

ment learning, training and habituation, task oriented

focus of attention, multi-sensory interaction. The

main brain structures involved in such tasks are the

basal ganglia system (including afferent and efferent

structures) regarding selection of action, the different

memory structures (e.g., episodic memory in the hip-

pocampus, or working memory in the prefrontal cor-

tex).

5 DISCUSSION

Beyond these basic features, the input and output can

be easily manipulated in order to enlarge the exper-

imental setup. Up to now, the main aspect is “input

or output degradation”, i.e. adding noise. Originally,

bot perception and action are performed without any

added noise or random mistake. Depending on the ex-

VirtualEnaction-APlatformforSystemicNeuroscienceSimulation

159

periment to be conducted, it is obvious (i.e. inserting

a few lines of code between the platform and the bot),

either to reduce variables precision range (e.g., add

noise to the pixelic image) or to randomly draw the

fact that a gesture may succeed or fail (e.g., introduce

spurious command).

A step further, as already implemented as plug-

ins, since all environment elements are available (not

to the bot, but to the experiment software), it would

be possible to design other cues, or more generally

other interactions with the environment. However, in

collaboration with neuroscience experimentalists, we

have carefully selected what seems useful to explore

biological systemic models, and avoided to provide a

too general tool that do anything.

Building one “brainy-bot” is a rather huge task

and requires several high-level cognitive functional-

ity. However, though systemic neuroscience requires

to study the system as a whole, it does not imply that

each functionality has to be studied at the same level

of details. Several blocks may be considered as black-

boxes interacting with the part of the system to be ex-

tensively studied. This is the reason why the present

platform is not limited to a survival environment, but

comes also with middle-ware (presently in develop-

ment) related to the basic cognitive functionality in-

volved in such paradigms, as listed in Appendix 5.

Some modules will thus be implemented according

to a rough description, e.g., via an algorithmic ersatz.

The nervous sub-system under study, on the reverse is

going to be implemented at a very ﬁne scale (neural

network mesoscopic models or even spiking neural

networks).

The key features of this digital experimenta-

tion platform include the capability to perform ex-

periments involving both long-term continuous time

paradigms or short-term decision tasks with a few

time-steps. It also allows us to consider either sym-

bolic motor command or sensory input (e.g., ingest

or not food, detect the presence of a stimulus) or

quantitative gestures and complex trajectory genera-

tion (e.g., ﬁnd resources in an unknown environment).

A key point is to be able to mimic and repeat at will

experiments performed in neuroscience laboratory on

animals. Here, the obtained computational models are

not only going to “ﬁt the data” but to be explored far

beyond, yielding the possibility to study long-term

adaptation, statistical robustness, etc. Not only one

instance of a bot can be checked, but several parallel

experiments can be run in order to explore different

parameter ranges, or compare alternative models.

It would also be instructive to better understand to

which extent such bio-inspired architectures actually

required to control a biological system could enhance

artiﬁcial control rules commonly applied in robotics

or game engines. This is a challenging issue, beyond

the present study, but an interesting perspective of the

present work.

As a conclusion, let us mention that this platform

has already been used for preliminary digital exper-

iments about Pavlovian conditioning (Gorojosky and

Alexandre, 2013) involving the functional modeling

of the amygdala and hippocampus, decision making

mechanisms in link with reinforcing signals yielded

by aversive or appetitive stimuli and internal com-

putation (Beati et al., 2013), plus a student work of

the AGREL connexionist categorization model here

confronted to a realistic environment (Carrere and

Alexandre, 2013).

ACKNOWLEDGMENTS

This work was partly supported by the KEOpS ANR

project. Huge thanks to Nicolas Rougier for precious

advises and Maxime Carrere for his feedback. The

NeBICA’14 review was a real chance to improve the

original draft, thanks.

REFERENCES

Beati, T., Carrere, M., and Alexandre, F. (2013). Which re-

inforcing signals in autonomous systems? In Third In-

ternational Symposium on Biology of Decision Mak-

ing, Paris, France.

Brette, R., Rudolph, M., Carnevale, T., Hines, M., Beeman,

D., Bower, J. M., Diesmann, M., Morrison, A., Good-

man, P. H., Harris, F. C., Zirpe, M., Natschl

ager, T.,

Pecevski, D., Ermentrout, B., Djurfeldt, M., Lansner,

A., Rochel, O., Vieville, T., Muller, E., Davison, A. P.,

El Boustani, S., and Destexhe, A. (2007). Simulation

of networks of spiking neurons: a review of tools and

strategies. Journal of computational neuroscience,

23(3):349–398.

Carrere, M. and Alexandre, F. (2013).

Emergence

de cat

egories par interaction entre syst

emes

d’apprentissage. In Preux, P. and Tommasi, M.,

editors, Conf

erence Francophone sur l’Apprentissage

Automatique (CAP), Lille, France.

Chemla, S., Chavane, F., Vieville, T., and Kornprobst, P.

(2007). Biophysical cortical column model for optical

signal analysis. BMC Neuroscience, 8(Suppl 2):P140.

Cofer, D., Cymbalyuk, G., Reid, J., Zhu, Y., Heitler, W. J.,

and Edwards, D. H. (2010). AnimatLab: a 3D

graphics environment for neuromechanical simula-

tions. Journal of neuroscience methods, 187(2):280–

288.

Davison, A. P., Br

uderle, D., Eppler, J., Kremkow, J.,

Muller, E., Pecevski, D., Perrinet, L., and Yger, P.

(2008). PyNN: A Common Interface for Neuronal

NEUROTECHNIX2014-InternationalCongressonNeurotechnology,ElectronicsandInformatics

160

Network Simulators. Frontiers in neuroinformatics,

Denoyelle, N. and Pouget, F. (2014). Virtualenaction,

user documentation. Technical report, virtualenac-

tion.gforge.inria.fr.

Friston, K. (2012). A Free Energy Principle for Biological

Systems. Entropy, 14(11):2100–2121.

Gorojosky, R. and Alexandre, F. (2013). Models of

Hippocampus for pavlovian learning. Rapport de

recherche RR-8377, INRIA.

Hyv

arinen, A. (2009). Natural image statistics a probabilis-

tic approach to early computational vision. Hardcover.

Pouget, F. and Denoyelle, N. (2014). Virtualenaction, de-

veloper documentation. Technical report, virtualenac-

tion.gforge.inria.fr.

Taouali, W., Vi

eville, T., Rougier, N. P., and Alexandre, F.

(2011). No clock to rule them all. Journal of physiol-

ogy, Paris, 105(1-3):83–90.

Teftef, E., Escobar, M.-J., Astudillo, A., Carvajal, C., Ces-

sac, B., Palacios, A., Vi

eville, T., and Alexandre, F.

(2013). Modeling non-standard retinal in/out function

using computer vision variational methods. Rapport

de recherche RR-8217, INRIA.

Uithol, S., van Rooij, I., Bekkering, H., and Haselager, P.

(2012). Hierarchies in action and motor control. Jour-

nal of cognitive neuroscience, 24(5):1077–1086.

eville, T. and Crahay, S. (2004). Using an Hebbian learn-

ing rule for multi-class SVM classiﬁers. Journal of

Computational Neuroscience.

eville, T. and Vadot, C. (2006). An improved biologically

plausible trajectory generator. Technical Report 4539-

2, INRIA.

APPENDIX

A BRAINY-BOT GENERIC

FUNCTIONALITY

Even when restraining to functional modeling, it is

not possible to simulate all sub-systems of the brain

at the same level of details. In order to help repre-

senting the part of the system that may be simulated

without biologically plausible models in a given con-

text, a few sets of functionality are proposed. Let us

brieﬂy present the main algorithmic cues.

- Episodic memory of input/output: Any system has

not only to take into account the present input and

output in order to generate a proper behavior, but also

store and consider the recent past information. In the

brain, this function is mainly located in the hippocam-

pus. In our context, thanks to the design choices, such

information corresponds to a simple ﬁle of symbolic

information, namely a hierarchical itemization of pa-

rameter values (i.e. a XML data structure). As a con-

sequence, the memorization, transmission, compres-

sion (in the sense of eliminating negligible values or

values related to older or smaller elements) is easy

to well-deﬁne. It has also the consequence to pro-

pose a generic internal representation of the sensory-

motor information at a symbolic level, without both-

ering about how this information is encoded on neural

maps. It is an interesting perspective of the present

work and an ongoing work to further investigate this

issue, while basic algorithmic modules for such data

management are already available.

- Trajectory generation: At a functional level a “be-

havior” (i.e., a complex gesture) can be speciﬁed as

ﬁnding a path from an initial state (e.g., being hun-

gry while food is known to be present elsewhere)

to a ﬁnal state (e.g., having the food ingested) tak-

ing constraints into account (e.g., avoiding or mov-

ing aside obstacles on the way). Such issue may be

a topic, or not (e.g., when studying the selection of

action we may prefer not to bother with the planiﬁca-

tion and execution of such actions). In the latter case,

generic speciﬁcation of such problem and universal

algorithms to solve them exist, in relation with har-

monic control which is a biologically plausible frame-

work (i.e., with fully distributed computation based

on diffusion mechanism) (Vi

eville and Vadot, 2006).

- Generic categorization: Another generic key cogni-

tive feature is the capability to “extract” symbolic in-

formation from a bundle of quantitative or qualitative

values. This includes sensory events detection (e.g.,

detect the presence of predator from sensory cues),

object labeling (e.g., an element as a resource to in-

gest). Though such issue is indeed a topic on its own,

we may wish in some context to have it available as

a black box. A biologically plausible support vector

machine mechanism is available to this end, with ver-

satile uses (Vi

eville and Crahay, 2004). Let us also

mention that informing the bot about its absolute po-

sition and orientation or about symbolic information

of the scene is a way to shortcut its sensory modules

and provide integrated cognitive information.

B COMPUTER

IMPLEMENTATION

This experimental virtualenaction setup is fully

documented both at the user (Denoyelle and Pouget,

2014) and developer (Pouget and Denoyelle, 2014)

levels. The user view of the computer implemen-

tation is oversimple as illustrated in Fig. 4, though

the underlying computer development is not that triv-

ial as sketched out in Fig. 5. It runs a bukkit

server in the 1.4.5 version and is designed as a set

of plugin for this server, with minimal changes in

the original code. This design choice was optimal

VirtualEnaction-APlatformforSystemicNeuroscienceSimulation

161

Figure 4: The platform’s architecture. At the user level, only the notion of (i) game server where the environment and its

mobile objects interactions are computed, (ii) game client where the game-play is rendered and (iii) brainy-bot where the

brain model simulation is issued have to be taken into account.

Figure 5: The platform underlying protocol. At the system level, the synchronization between the existing game events and

the different system component is a non trivial middle-ware, taking into account both “real players” (thus playing in real time)

and “bot players” (with slow reaction delays). The developed middle-ware can cope with several bots and humans (which

will have to be patient, since the game runs at a reduced rate). Furthermore, the design allows the platform to be distributed

on a cluster (server, client and bots running on different machines).

in terms of middle-ware development effort and ﬁ-

nal code stability. At the server level, two plugins

have been designed. The VirtualEnaction plugin

allows us to replace usual player mouse and keyboard

interactions with the game-play by a programmable

API. The Characteristics plugin allows us to im-

plement blocks and entities additional characteristics

such as temperature, hardness or smell.

The minecraft server and client have been modi-

ﬁed in order to be able to run in this programmatic

mode, and to get a full access to all game gauges, be-

cause we not only must run the bot but also observe in

details its behavior and measure all interactions with

the environment. As a consequence, the actual source

code corresponds to a fork of a frozen version of the

minecraft game, while everything is documented in

order to easily migrate to some more recent version.

Another key point is the modiﬁcation of the en-

vironment. Using the standard game, it is possible

to build an environment, include other entities, and

so on. This is simply realized via the game interface.

Futhermore, scripted scenarios can also be introduced

(e.g., some resource appears if and only if a block is

put in a given position), as additional plugins. In ad-

dition, conﬁguration ﬁles on the server side, allows us

to act on available resources or survival gauges, and

build versatile survival situations. It is thus a fully

editable setup. As being an open-source collaborative

software, it is available on the Inriaforge with more

than 15 co-developers or project followers.

A step further, in order this virtual experimenta-

tion platform to be extensible in the perspective of

NEUROTECHNIX2014-InternationalCongressonNeurotechnology,ElectronicsandInformatics

162

new setups, the botplug that allows the bot to be

plugged in the client/server system is also designed

with a notion of plugin. For instance, a tactile sense

(not described here), and a variant of the pixelic vision

to be connected to different image sequence compu-

tation libraries are made available as plugins.

It is clear that several extensions are going to be

developed in a near future thanks to this modular ar-

chitecture choice.

VirtualEnaction-APlatformforSystemicNeuroscienceSimulation

163