Upload
soundgod-victor
View
215
Download
0
Embed Size (px)
Citation preview
8/16/2019 Rat Head Algorithm
1/32
RAT HEAD ALGORITHM
8/16/2019 Rat Head Algorithm
2/32
Abstract
Cyborg intelligence is an emerging kind of intelligence paradigm.
It aims to deeply integrate machine intelligence with biological
intelligence by connecting machines and living beings via neural
interfaces, enhancing strength by combining the biological
cognition capability with the machine computational capability.
Cyborg intelligence is considered to be a new way to augment
living beings with machine intelligence. In this paper, we build rat
cyborgs to demonstrate how they can expedite the maze escape
task with integration of machine intelligence. We compare the
performance of maze solving by computer, by individual rats, and
by computer-aided rats i.e. rat cyborgs!. "hey were asked to #nd
their way from a constant entrance to a constant exit in fourteen
diverse mazes. $erformance of maze solving was measured by
steps, coverage rates, and time spent. "he experimental results
with six rats and their intelligence-augmented rat cyborgs show
that rat cyborgs have the best performance in escaping from
mazes.
8/16/2019 Rat Head Algorithm
3/32
I%"&'()C"I'%
Within the past two decades, bio-robots have been realized on
di*erent kinds of creatures, such as cockroaches, moths, beetles,
and rats. "hey are expected to be superior to traditional
mechanical robots in mobility, perceptivity, adaptability, and
energy consumption. +mong them, rat robots are becoming
popular for their good maneuverability. "o make a rat robot, a pair
of micro electrodes are implanted into the medial forebrain
bundle ! of the rat/s brain, and the other two pairs are
implanted into the whisker barrel #elds of left and right
somatosensory cortices. +fter the rat recovers from the surgery, a
wireless micro-stimulator is mounted on the back of the rat to
deliver electric stimuli into the brain through the implanted
electrodes. "his allows a user, using a computer, to deliver
stimulus pulses to any of the implanted brain sites remotely.
0timulation in can excite the rat robot by increasing the level
of dopamine in its brain, and stimulation in the left or right 0I
makes the rat robot feel as if its whiskers are touching a barrier.
efore a rat robot is used for navigation, a training process is
usually needed to reinforce the desired behaviors i.e. moving
8/16/2019 Rat Head Algorithm
4/32
ahead, turning left and turning right!. In this process, the
stimulation acts as a reward as well as a cue to move ahead, and
the left and right 0I stimulation act as the cues to turn left and
turn right respectively. In order to get the reward, the rat robot
will learn to do the correct behaviors corresponding to the cues.
+fter su1cient navigation training, the rat robot will move ahead
in response to the orward cue, turn left in response to the 2eft
cue, and turn right in response to the &ight cue. In this paper, the
rat robot is referred to as rat cyborg.
'ne of the reasons that researchers take interest in developing
rat cyborgs is that rats have outstanding spatial localization
abilities. "hey can #nd a way in an environment by orienting
themselves in relation to a wide variety of cues, including distal
cues, typically provided by vision, audition, and olfaction, as well
as proximal cues, typically provided by tactile, kinesthetic, and
inertial systems. &ats even have a well-developed magnetic
compass sense for spatial orientation. &ats do not build detailed
geometrical representations of the environment. "hey rely on the
learnt associations between external perception and the pose
8/16/2019 Rat Head Algorithm
5/32
belief created from the self-motion cues. 0tudies of spatial
orientation posit that rats use an 3inner 4$05 in hippocampus and
entorhinal cortex to create a cognitive map of the environment.
urthermore, &at 02+, a navigation model which is inspired by
rat/s brain, can perform simultaneous localization and mapping in
real time on a mechanical robot.
'n the other hand, machines or computers! are e1cient in
numerical computation, information retrieval, statistical
reasoning, and have almost unlimited storage. (i*erent to rats,
machines have their own methods to explore and learn about the
environment in spatial navigation problems. "hey can capture
many categories of information from the environment through
various sensors, such as range sensors, visual sensors, vibration
sensors, acoustic sensors, and location sensors. "he information
is saved and converted into a discrete and digital form. "hen the
processed information will be synthesized to map the
environment by di*erent paradigms. 6ventually, machines can
choose various searching algorithms in path planning, for
example, 7ood-#ll method, (i8kstra/s algorithm, +9 search
8/16/2019 Rat Head Algorithm
6/32
algorithm, rapidly exploring random tree, probabilistic roadmap.
achines run the sense process, map process, and decision
process in real time and in parallel. + typical demonstration of
machine/s navigation abilities is the micromouse competition, in
which a small rat-like mechanical robot explores a :; can biological intelligence be augmented with the
help of machine intelligence? "o explore the =uestion, a maze
solving task was introduced, in which three kinds of sub8ects
computer, rats and rat cyborgs! were asked to #nd their ways
from a predetermined starting position to a predetermined target
position. It is a challenge to embed the machine/s maze solving
capability into their navigation. In our experiments, the computer
traversed the mazes based on an improved wall follower
approach, six rats traversed :@ mazes one by one all by
themselves, and then traversed the :@ mazes again in the same
order with the assistance of the computer. $erformance was
8/16/2019 Rat Head Algorithm
7/32
measured by steps, coverage rates, and time spent, allowing for
comparisons.
8/16/2019 Rat Head Algorithm
8/32
aterials and ethods
0ub8ects
0ix rats adult 0prague (awley ratsA BD EFD g! took part in the
experiments over the course of one month. +ll of the six rats
named (G:B, 0H:F, 0H:, H:B, D: and DE! had already
gone through a su1cient navigation training process, which
means they would turn left in response to the 2eft stimulus, turn
right in response to the &ight stimulus, and move ahead in
response to the orward stimulus. %ote that in this study, 2eft and
&ight stimuli are used to instruct the rat cyborgs to turn left and
right, as well as to prevent them from moving into dead roads.
orward stimulus is not used to instruct them to move ahead, but
to reward them. In our experiments, the six rats and the six rat
cyborgs are the same rats. + rat solving mazes by itself is called a
rat, while the same rat solving mazes with the assistance of
computer via its backpack stimulator is called a rat cyborg.
Maze Solving by Computer.
0imilar to rats, the computer had to #nd a path from the start to
the target with no knowledge of the mazes. "here are a number of
8/16/2019 Rat Head Algorithm
9/32
di*erent maze solving algorithms, such as random mouse, wall
follower, $ledge, and "remaux/s algorithm. In this study, on the
basis of dead road detection and uni=ue road detection, two wall
follower rules left-hand and right-hand! were adopted to traverse
the :@ mazes. (ead roads refer to roads that cannot lead to the
target. "here are two kinds of dead roads> visited dead road and
non-visited dead road. "he dead road detection algorithm is listed
in Algorithm A. + uni=ue road refers to the only way to the
target. "he detailed uni=ue road detection algorithm is listed
in Algorithm B. "he complete maze solving algorithm is listed in
+lgorithm :. or left-hand rule, the traversal se=uence is
clockwise leftJfrontJrightJback!, and for right-hand rule, the
traversal se=uence is anticlockwise rightJfrontJleftJback!. ig :.
0hows two maze solving processes by our algorithm. 0teps and
coverage rates of maze solving in the :@ mazes were recorded.
We took the average steps and average coverage rates of the left-
hand rule and right-hand rule as the performance measures.
8/16/2019 Rat Head Algorithm
10/32
Fig 1. Maze solvig b! o"r algorithm.
"he yellow cell is the current position of the explorer. lue cells
indicate a uni=ue road to the target. Cyan cells are cells which
have not been explored, and walls of these cells now are unknown
to the explorer. $ink slash 3K5! denotes the non-visited dead
road, &ed Cross 3 aze solving by our algorithm left-hand!.
: #hile the current cell C is not the target cell $o
8/16/2019 Rat Head Algorithm
11/32
B i% there is a uni=ue cell ) which is accessible in the four
ad8acent cells the
E move to )A
@ e$
F else i% the left cell of C is accessible and not a dead cell the
; move to the left cellA
e$
M else i% the front cell of C is accessible and not a dead
cell the
move to the front cellA
:D e$
:: else i% the right cell of C is accessible and not a dead
cell the
:B move to the right cellA
:E e$
:@ else i% the back cell of C is accessible and not a dead
cell the
:F move to the back cellA
:; e$
: e$
"o ensure that our algorithm can re7ect the computer/s maze
solving capacity, we compare its performance in B@ mazes maze
8/16/2019 Rat Head Algorithm
12/32
: to maze B@! with that of three classical maze solving algorithms
i.e. wall follower, $ledge and "remaux/s algorithm!. "he
experimental results are presented. We can see from the left
column that both the steps and coverage rates of our algorithm in
most mazes are less than those of the other three algorithms.
rom the right column, the average steps and average coverage
rates of our algorithm are less than those of the other three
algorithms. "he results indicate that our algorithm outperforms
the other three algorithms in maze solving. "hus we take the
performance of our algorithm as computer/s performance in maze
solving, and compare it with that of rats and rat cyborgs.
8/16/2019 Rat Head Algorithm
13/32
Maze Solving by Rats.
+fter the maze solving training, the entire maze was washed and
dried, and then the six rats proceeded to this formal maze solving
procedure. In this procedure, the rats solved :@ mazes one by one
using their own spatial learning abilities. (etails of this procedure
are similar to the maze solving training procedure. "he
experimental time in each day was M-:B a.m. and B-; p.m. 6ach
rat ran each maze only one time. In each maze, at #rst the rat
was placed in the constant starting cell. +fter the rat arrived at
the target cell and drank a drop of water D.:F ml! in a dish, #ve
consecutive rewards were sent. 'nce all of the six rats
#nished a maze, they proceeded to solve the next maze. 6ach rat
was o*ered ; ml water at the end of one day/s experiment. 0teps,
coverage rates and time spent of maze solving by the rats in the
:@ mazes were analysed. +dditionally, we observed that the
strategy rats took to solve the mazes was not easily understood
in other words, it was not perfect!. "hey revisited the visited
cells, explored the dead roads which could be easily detected by
the computer, and even returned to the starting cell when they
8/16/2019 Rat Head Algorithm
14/32
were confused. "his o*ered an opportunity to the computer to
help rats in maze solving.
Maze Solving by Rat Cyborgs.
+ rat cyborg is a computer-aided rat, which is a combination of a
rat and machines. +fter the procedure of maze solving by rats, the
entire maze was washed and dried, and then the same six rats
traversed the :@ mazes in the same order with the computer/s
help. (etails of this procedure are similar to the procedure of
maze solving by rats, except that the rats have the help of the
computer to solve the mazes. 0hows two maze solving processes
by rat cyborgs. With the motivation to help rats solve the mazes,
the computer tracked the rats, analyzed the explored maze
information, and decided when and how to intervene. "he
computer aided the rats under three rules> :! if there was a path
to the uni=ue road, the computer would #nd the shortest path,
then 2eft and &ight commands would be sent to navigate the rat
to the uni=ue roadA B! if the rat was going to enter a dead cell,
2eft or &ight commands would be sent to prevent such a moveA
E! if the rat was in a loop
8/16/2019 Rat Head Algorithm
15/32
(etected by Algorithm &, the computer would #nd the shortest
path to the current destination, then 2eft and &ight commands
would be sent to navigate the rat to follow the path. In this
experiment, the computer automatically analyzed the dead roads,
uni=ue roads, loop roads and shortest roads in real time, and
generated the guiding information to overlay on the live video,
which is playing on the screen of the computer. +t the same time,
an expert was watching the guiding information shown on the
screen and sent the 2eft and &ight commands to direct the rat
under the three rules. "he sent 2eft and &ight commands were
also shown simultaneously in the video. $lease note that the
expert was not directly involved in the maze solving, hisNher
operations to send the 2eft and &ight commands under the
guidance of the computer is 8ust a simple replacement of the
computer/s execution action!, not an augmentation or a
diminution of the intelligence, since all the path detection are
performed by the computer. +part from the above three rules,
rats solved the mazes of their own free will. We did not treat rats
as machines which were totally controlled by the computer. 0teps,
8/16/2019 Rat Head Algorithm
16/32
coverage rates and time spent of maze solving by the rat cyborgs
in the :@ mazes were analysed.
ig B. aze solving by rat cyborgs.
The red point is the body position, the black point is the head position, and the blue line is the heading direction of the rat
cyborg. Our rat cyborg tracking method (see Algorithm 2) is
implemented by OpenCV. lue cells indicate a uni!ue road to the
target. "ed cells are cells #hich ha$e not been e%plored, and
#alls of these cells no# are unkno#n to the rat cyborg. The pink
slash (&') denotes the non$isited dead road, the red cross (&*)
denotes the $isited dead road, and the yello# &g denotes
entrances to the une%plored area. Consecuti$e blue points
indicate the shortest path to the current destination. The shortest
path is detected by A+ algorithm
8/16/2019 Rat Head Algorithm
17/32
Algorithm '( Rat c!borg trac)ig algorithm.
: 0ave a background image to aA
B #hile the rat cyborg has not reached the target cell - $o
E pick an image b from the cameraA
@ use c$Absi/ to subtracta fromb, write the di*erence to cA
F use c$Threshold to get a binary imaged fromcA
; use c$0indContours to #nd all of the contoursCo indA
search the largest contour Cma% inCoA
M calculate the centroid of pixels in Cma% , write it to the body
position "bA
use c$-ood0eaturesToTrack to detect corners inCma% #nd a
@D
8/16/2019 Rat Head Algorithm
18/32
rat cyborgs. inally, correlation between the steps and correlation
between the coverage rates are analyzed.
*TE+*
0teps are the sum of times visiting each cell. + cell can be visited
multiple times. 0teps of maze solving by the six ratsNrat cyborgs
are presented with those by the computer in ig E. We can see
from the left column that the steps of rat cyborgs in most mazes
are less than those of rats, and may be e=uivalent to those of the
computer.
8/16/2019 Rat Head Algorithm
19/32
&overage Rates
Coverage rate is the number of visited cells. 6very visited cell is
counted only one time, so the maximum value of the coverage
rate is :DD. "he larger the coverage rate is, the less e1cient the
maze solving is. It can evaluate the performance of the explorer in
maze solving from another point of view. Coverage rates of maze
solving by the six ratsNrat cyborgs are presented with those by the
8/16/2019 Rat Head Algorithm
20/32
computer in ig @. We can see from the left column that the
coverage rates of rat cyborgs in most mazes are less than those
of the computer and those of the rats. rom the right column, the
average coverage rates of each rat cyborg are less than that of
each corresponding rat> (G:B from FM.:O;.F to FB.B:O;.M!,
0H:F from FF.E;O;.B; to F:.:O;.M@!, 0H: from FE.FO.D;
to F:.;@O;.EF!, H:B from ;D.E;OF.BF to [email protected]!, D:
from ;F.DOF.M; to FB.:O;.B!, and DE from ;@.M;O.B to
F:.DO;.B!A and the average coverage rates of each rat cyborg
are also less than that of the computer ;D.FDOF.:E!.
8/16/2019 Rat Head Algorithm
21/32
8/16/2019 Rat Head Algorithm
22/32
Fig ,. &overage rates o% maze solvig b! com-"ter rats
a$ rat c!borgs.
"he coverage rates of maze solving by (G:B, 0H:F, 0H:, H:B,
D: and DE are shown from the top to the bottom, respectively.
8/16/2019 Rat Head Algorithm
23/32
Time *-et
"ime spent measures the period between a ratNrat cyborg starts
to move at 0 and the ratNrat cyborg reaches 4. esides steps and
coverage rates, we further compare the time spent of the rats in
the :@ mazes with those of the rat cyborgs. "ime spent of maze
solving by the six ratsNrat cyborgs are presented in ig . +s we
can see from the left column that the time spent of rat cyborgs in
most mazes are less than those of rats. rom the right column,
the average time spent of each rat cyborg is less than that of
each corresponding rat> (G:B from :;M.DO@@.F to
:B.E;OEB.@!, 0H:F from E@.@EO;M.;M to :BF.DDOE:.@M!,
0H: from EBM.FDO@D.: to BFE.@EOE:.;!, H:B from
FMM.@EO:D;.MF to EEE.DO;:.FM!, D: from :F;.:@OEE.:E to
D.FDOM.MB!, and DE from :@@.E;OE;.EF to @F.:@O:E.;!. "he
two-tailed paired t-test shows that there are four rats i.e. 0H:F,
H:B, D: and DE! that performed better with the assistance of
the computer. "hese results suggest that in maze solving, based
on the time spent, the performance of rat cyborgs is better than
that of individual rats.
8/16/2019 Rat Head Algorithm
24/32
Fig /. Time s-et o% maze solvig b! rats a$ rat c!borgs.
"he time spent of maze solving by (G:B, 0H:F, 0H:, H:B, D:
and DE are shown from the top to the bottom, respectively. (ata
are presented as meanOs.e.m. on the right. 9 pPD.DF, 99 pPD.D:,
999 pPD.DD:.
8/16/2019 Rat Head Algorithm
25/32
(iscussion
"he experimental results show that, in terms of steps, coverage
rates and time spent, rat cyborgs performed better than rats in
maze solving. Gowever, in terms of steps, performance of rat
cyborgs did not show remarkable advantages over that of
computer. +ctually, rats did not strive to reach the destination in
the least number of steps. "hey sometimes would revisit visited
cells again and again, and wander between ad8acent cells.
%onetheless, all of the six rat cyborgs outperformed the computer
in terms of the coverage rates. oreover, the rat cyborgs are
agile in di*erent types of terrain.
ecause the six rats and the six rat cyborgs were the same rats,
the possible interference should be carefully prevented.
&apid advance of brain-machine interfaces Is!, the connection
and interaction between the organic components and computing
components of the cyborg intelligent systems are becoming
deeper and better. ased on such kinds of symbiotic bio-machine
systems, a new type of intelligence, which we refer to as cyborg
intelligence, will play an increasingly crucial role. Cyborg
8/16/2019 Rat Head Algorithm
26/32
intelligence is a convergence of machine and biological
intelligence, which is capable of integrating the two
heterogeneous intelligences at multiple levels. It has the potential
to deliver tremendous bene#ts to society, such as in search and
rescue, health care, and entertainment. "he mobility,
perceptibility and cognition capability of the rats were combined
with the sensing and computing power of the machines in the rat
cyborg system. "he experimental results of the rat cyborg system
in maze solving provide a proof-of-principle demonstration for
cyborg intelligence.
8/16/2019 Rat Head Algorithm
27/32
Conclusions and uture Work
In this paper, we build intelligence-augmented rat cyborgs and
present a comparative study of maze solving by computer, by
rats, and by rat cyborgs. Computer aids rats in dead road
detection, uni=ue road detection, loop detection, and shortest
path detection.
In future work, more tasks will be introduced, and the complexity
of tasks will be =uanti#ed. "o avoid excessive intervention with
the rats, the strength of the computer/s assistance will be graded.
In addition, more practical rat cyborgs will be investigated> the
web camera will be replaced by sensors mounted on rats, such as
tiny camera, ultrasonic sensors, infrared sensors, electric
compass, and so on, to perceive the real unknown environment in
real timeA and the computer-aided algorithms can be housed on a
wireless backpack stimulator instead of in the computer.
8/16/2019 Rat Head Algorithm
28/32
&eferences
• ozkurt +, 4ilmour &, 2al +. alloon-assisted 7ight of
radio-controlled insect biobots. I666 "ransactions on iomedical
6ngineering. BDDAF;!>BED@QBED. doi>
:D.::DN"6.BDD.BDBBFF:. pmid>:;BED;
• 6tienne 0, 4uthrie , 4oillandeau , %guyen "G,
'rignac G, 4ross C, et al. 6asy rider> monkeys learn to drive a
wheelchair to navigate through a complex maze. $lo0 'ne.
BD:@AF!>e;BF. doi> :D.:E:N8ournal.pone.DD;BF.
pmid>B@ME::ED
•
eng R, Chen W, Se T, Rhang 0, Rheng T, Wang $, et al.
+ remote control training system for rat navigation in complicated
environment. Uournal of Rhe8iang )niversity 0cience +.
BDDAMB!>EBEQEED. doi> :D.:;E:N8zus.BDD.+DEBE.
• yhn , olden 0, Witter $, oser 6I, oser .
0patial representation in the entorhinal cortex. 0cience.
BDD@AEDFF;MM!>:BFMQ:B;@. doi> :D.::B;Nscience.:DD:.
pmid>:FEEEMEB
8/16/2019 Rat Head Algorithm
29/32
:. Golzer &, 0himoyama I. 2ocomotion control of a bio-robotic
system via electric stimulation. In> I666N&0U International
Conference on Intelligent &obots and 0ystems, :. vol. E. pp.
:F:@Q:F:. doi> :D.::DNiros.:.;F;FF
• 2ee 4, Uun 4, Choi G0, Uang G0, ae SC, 0uk V, et al.
'perant conditioning of rat navigation using electrical stimulation
for directional cues and rewards. ehavioural $rocesses.
BD:DAM@E!>:FQBD. pmid>BD@:BF doi>
:D.:D:;N8.beproc.BD:[email protected]
• ather U4, aker &&. agnetic sense of direction in
woodmice for route-based navigation. %ature.
:M:AB:FM::!>:FBQ:FF. doi> :D.:DEMNB::FBaD.
• uheim &, 6dgar %, 0loan V+, $hillips U. agnetic
compass orientation in CF2N;U mice. 2earning ehavior.
BDD;AE@@!>E;;QEE. doi> :D.EFMNDE:EBD:.
• '/Veefe U. $lace units in the hippocampus of the freely
moving rat. 6xperimental %eurology. :;AF::!>MQ:D.
pmid>:B;:;@@ doi> :D.:D:;NDD:@-@MM;;!DDFF-M
8/16/2019 Rat Head Algorithm
30/32
• $hillips U, Soumans $W, uheim &, 0loan V+, 2andler 2,
$ainter 0, et al. &apid learning of magnetic compass direction by
CF2N; mice in a @-armed Xplus/ water maze. $lo0 'ne.
BD:EAMM!>eE::B. doi> :D.:E:N8ournal.pone.DDE::B.
pmid>B@DBE;E
• 0argolini , yhn , Gafting ", c%aughton 2, Witter
$, oser , et al. Con8unctive representation of position,
direction, and velocity in entorhinal cortex. 0cience.
BDD;AE:BF@!>FMQ;B. doi> :D.::B;Nscience.::BFFB.
pmid>:;;FD@
• 0ato G, erry CW, $eeri S, aghoomian 6, Casey 6,
2avella 4, et al. &emote radio control of insect 7ight. rontiers in
Integrative %euroscience. BDDAEB!>:Q@. doi>
:D.::DNmemsys.BDD.@MDFEF
• 0un C, Rheng %, Rhang T, Chen W, Rheng T. +utomatic
navigation for rat-robots with modeling of the human guidance.
Uournal of ionic 6ngineering. BD:EA:D:!>@;QF;. doi>
:D.:D:;Ns:;B-;FB:E!;D:M-F
8/16/2019 Rat Head Algorithm
31/32
• 0utherland &U, Whishaw IY, Volb . + behavioural
analysis of spatial localization following electrolytic, kainate- or
colchicine-induced damage to the hippocampal formation in the
rat. ehavioural rain &esearch. :MEAB!>:EEQ:FE. doi>
:D.:D:;ND:;;-@EBMME!D:MM-B. pmid>;MED;@M
• "alwar 0V, Tu 0, Gawley 60, Weiss 0+, oxon V+,
Chapin UV. &at navigation guided by remote control. %ature.
BDDBA@:;MM@!>EQEM. doi> :D.:DEMN@:DEa. pmid>::M;;F
• "homson 66, Carra &, %icolelis +. $erceiving invisible
light through a somatosensory cortical prosthesis. %ature
Communications. BD:EA@B!>;;QM. doi> :D.:DEMNncommsB@
• Wang S, 2u , Wu R, "ian 2, Tu V, Rheng T, et al. Hisual
cue-guided rat cyborg for automatic navigation. I666
Computational Intelligence. BD:FA:DB!>@BQFB. doi>
:D.::DNCI.BD:F.B@DFE:M.
• Tu (, Gao T, Wang R, (uan S, arsh &, Su 0, et al. +
virtual radial arm maze for the study of multiple memory systems
8/16/2019 Rat Head Algorithm
32/32
in a functional magnetic resonance imaging environment.
International Uournal of Hirtual &eality IUH&!. BD:BA::B!>;EQ;.