Welcome to Scribd, the world's digital library. Read, publish, and share books and documents. See more
Download
Standard view
Full view
of .
Look up keyword
Like this
0Activity
0 of .
Results for:
No results containing your search query
P. 1
Behavioral Dynamics of Intercepting

Behavioral Dynamics of Intercepting

Ratings: (0)|Views: 8|Likes:
Warren, W.H., & Fajen, B.R. (2007). Behavioral dynamics of intercepting a moving target. Experimental Brain Research, 180, 303–319.

From matters of survival like chasing prey, to games like football, the problem of intercepting a target that moves in the horizontal plane is ubiquitous in human and animal locomotion. Recent data show that walking humans turn onto a straight path that leads a moving target by a constant angle, with some transients in the target-heading angle. We test four control strategies against the human data: (1) pursuit, or nulling the target-heading angle, (2) computing the required interception angle; (3) constant target heading angle, or nulling change in the target-heading angle; and (4) constant bearing, or nulling change in the bearing direction of the target; which is equivalent to nulling change in the target-heading angle while factoring out the turning rate. We show that human interception behavior is best accounted for by the constant bearing model, and that it is robust to noise in its input and parameters. The models are also evaluated for their performance with stationary targets, and implications for the informational basis and neural substrate of steering control are considered. The results extend a dynamical systems model of human locomotor behavior from static to changing environments.
Warren, W.H., & Fajen, B.R. (2007). Behavioral dynamics of intercepting a moving target. Experimental Brain Research, 180, 303–319.

From matters of survival like chasing prey, to games like football, the problem of intercepting a target that moves in the horizontal plane is ubiquitous in human and animal locomotion. Recent data show that walking humans turn onto a straight path that leads a moving target by a constant angle, with some transients in the target-heading angle. We test four control strategies against the human data: (1) pursuit, or nulling the target-heading angle, (2) computing the required interception angle; (3) constant target heading angle, or nulling change in the target-heading angle; and (4) constant bearing, or nulling change in the bearing direction of the target; which is equivalent to nulling change in the target-heading angle while factoring out the turning rate. We show that human interception behavior is best accounted for by the constant bearing model, and that it is robust to noise in its input and parameters. The models are also evaluated for their performance with stationary targets, and implications for the informational basis and neural substrate of steering control are considered. The results extend a dynamical systems model of human locomotor behavior from static to changing environments.

More info:

Categories:Types, Research, Science
Published by: Laboratory in the Wild on Jul 09, 2012
Copyright:Attribution Non-commercial

Availability:

Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less

07/09/2012

pdf

text

original

 
RESEARCH ARTICLE
Behavioral dynamics of intercepting a moving target
Brett R. Fajen
Æ
William H. Warren
Received: 25 January 2006/Accepted: 5 January 2007/Published online: 2 February 2007
Ó
Springer-Verlag 2007
Abstract
From matters of survival like chasing prey,to games like football, the problem of intercepting atarget that moves in the horizontal plane is ubiquitousin human and animal locomotion. Recent data showthat walking humans turn onto a straight path thatleads a moving target by a constant angle, with sometransients in the target-heading angle. We test fourcontrol strategies against the human data: (1)
pursuit 
,or nulling the target-heading angle
b
, (2)
computing
therequired interception angle
^
b
;
(3)
constant target-heading angle
, or nulling change in the target-headingangle
_
b
;
and (4)
constant bearing
, or nulling change inthe bearing direction of the target
_
w
;
which is equiva-lent to nulling change in the target-heading angle whilefactoring out the turning rate
ð
_
b
À
_
/
Þ
:
We show thathuman interception behavior is best accounted for bythe constant bearing model, and that it is robust tonoise in its input and parameters. The models are alsoevaluated for their performance with stationary tar-gets, and implications for the informational basis andneural substrate of steering control are considered. Theresults extend a dynamical systems model of humanlocomotor behavior from static to changing environ-ments.
Keywords
Interception
Á
Visual control
Á
Locomotion
Á
Dynamical systems modeling
Introduction
As we walk about the world each day, our locomotorbehavior is effortlessly coordinated with a complex,dynamic environment. Consider, as an extreme exam-ple, the American game of football, in which a playermust run toward stationary goals, avoid stationary andmoving obstacles, intercept and tackle moving targets,and evade pursuit. Similar challenges are faced on adaily basis by animals in the wild, people walkingthrough public spaces, and increasingly by autonomousmobile robots. In order to understand these funda-mental abilities, we seek to model the
behavioral dynamics
of locomotion—the time-evolution of behavior as an agent interacts with its environment.Our approach to this problem begins by modeling aset of elementary locomotor behaviors as simpledynamical systems, including steering to a stationarygoal, avoiding a stationary obstacle, intercepting amoving target, and avoiding a moving obstacle. In-spired by the work of Scho ¨ ner et al. (1995) on roboticcontrol, we recently developed a model of humansteering and obstacle avoidance in static environmentsthat generates the time-course of heading direction(Fajen and Warren2003). One component of themodel accurately simulates the paths people take whenwalking to a stationary goal, while a second componentreproduces human detours around an obstacle andpredicts routes through novel configurations of obsta-cles. In the present study, we seek to extend the modelto the case of intercepting a moving target. Specifically,
B. R. Fajen (
&
)Department of Cognitive Science,Rensselaer Polytechnic Institute, Carnegie Building 308,110 8th Street, Troy, NY 12180-3590, USAe-mail: fajenb@rpi.eduW. H. WarrenDepartment of Cognitive and Linguistic Sciences,Brown University, Providence, USA
 123
Exp Brain Res (2007) 180:303–319DOI 10.1007/s00221-007-0859-6
 
we evaluate four possible control strategies for targetinterception: (1)
pursuit 
, (2)
computing
the requiredinterception angle, (3)
constant target-heading angle
,and (4)
constant bearing
, and test these hypothesesagainst our empirical data (Fajen and Warren2004).Our analysis also bears on the perceptual informa-tion used to guide locomotion. It is well established thathumans can perceive their direction of heading fromoptic flow with an accuracy of about 1
°
under a varietyof environmental and self-motion conditions (seeWarren2004, for a review). People can also judge theirdirectionofwalkingbasedonproprioceptiveandmotorinformation about the locomotor axis, but are an orderof magnitude less accurate (Telford et al.1995; seeIsrae ¨ l and Warren2005, for a review). Thus, walking toa stationary target might be based on several possiblevariables, including
optic flow
, such that one aligns theheading specified by optic flow with the visual target(Gibson1950; Warren et al.2001);
egocentric direction
,such that one aligns the locomotor axis with the ego-centric direction of the target (Rushton et al.1998);
centering
, such that one fixates the target and aligns thebody midline with direction of gaze on the basis of proprioception
1
(Hollands et al.2002); or
target drift 
,such that one cancels the motion of the target in thefield of view, which is equivalent to nulling change inthe target-heading angle
2
(Llewellyn1971; Rushtonet al.2002; Wilkie and Wann2003,2005). In the case of steering to a stationary goal, the evi-dence indicates that participants rely on both optic flowand egocentric direction. Optic flow tends to dominatewhen there is sufficient visual surface structure in thescene (Harris and Carre´ 2001; Li and Warren2002; Turano et al.2005; Warren et al.2001; Wilkie and Wann2002,2003; Wood et al.2000),whereas egocentric direction dominates when visual structure is reduced(Rushton et al.1998; Harris and Bonas2002). But par- adoxically, in the case of intercepting a moving target,Fajen and Warren (2004; see also Chardenon et al.2004) found no influence of optic flow and concludedthatparticipants reliedsolelyon the egocentric directionof the target. This inconsistency might be resolved if itturned out that there are different underlying controlstrategies for stationary and moving targets. Indeed, thepresent findingssuggest thatsteering to a stationarygoalis based on the nulling the target-heading angle,whereas intercepting a moving target is based on nullingchange in the direction of the target in space.Information about self-motion appears to be ex-tracted by cortical networks in the dorsal pathway thatintegrate multisensory signals and perform sensori-motor transformations related to locomotor control(Duffy 2005; Raffi and Siegel2004). Single-unitrecordings in macaque indicate that cells in area MSTdare selective for large-field optic flow patterns andintegrate eye pursuit and vestibular signals; neurons inarea VIP have a heading tuning that is invariant overeye movements (Zhang et al.2004) and integrate tac-tile and vestibular signals; area 7a cells are sensitive tothe pattern and speed of optic flow and exhibit eyeposition tuning; area STPa neurons respond to bothoptic flow and object boundaries, and may be sensitiveto object-relative heading; flow-sensitive cells in areaPEc project to premotor cortex; and even motor cor-tical neurons in M1 are selective for expansionpatterns. Functional neuroimaging studies indicateflow-specific activity in homologous regions of thehuman brain, including the hMT+ complex (primateMT–MST), VIP, DISPM/L (primate 7a), and STG(primate STP) (Morrone et al.2000; Peuskens et al.2001; Vaina and Soloviev2004). In particular, area STG is activated by heading with respect to landmarks(Vaina and Soloviev2004), and a region in the superiorparietal lobule (SPL) that includes LIP responds toheading errors relative to a roadway in an activesteering task (Field et al.2006). Taken together, thesefindings suggest a neural substrate for the extraction of target-heading angle and for transformations relevantto the control of steering. A functional model of locomotor control would offer candidate hypothesesabout these sensorimotor mappings whose neuralimplementation could then be investigated.In what follows, we briefly describe the steeringdynamics model for a stationary goal (Fajen andWarren2003), and then extend the model to the caseof intercepting a moving target. We then comparesimulations of each control strategy against our previ-ous human data on target interception.
Steering to a stationary goal
We define
heading
(
/
) as the direction of locomotionwith respect to an allocentric reference axis,
3
and
1
This assumes that the midline is coincident with the locomotoraxis, but exceptions include a ‘‘crabbing’’ gait for terrestrialanimals, a crosswind for aerial animals, and a crosscurrent foraquatic animals.
2
This assumes that the observer is not rotating. It follows fromthe basic law of optic flow (Nakayama and Loomis1974) that thetarget’s angular velocity is proportional to the sine of the target-heading angle and inversely proportional to distance.
3
The use of an allocentric reference axis to define heading is forconvenience of analysis. The perceptual input to the agent is thetarget-heading angle
b
=
/
w
m
.304 Exp Brain Res (2007) 180:303319
 123
 
bearing
(
w
 g
) as the direction of a target with respect tothe same axis, at a distance
d
 g
(see Fig.1a; the sub-script
g
designates that the target is a stationary goal).The
target-heading angle
(
b
=
/
w
 g
) is the differencebetween the two. To turn onto a straight path towardthe goal, the agent must null this heading error andstabilize the heading in the direction of the goal, suchthat the target-heading angle (
b
=
/
w
 g
) and theturning rate
ð
_
/
Þ
both go to zero before the goal isreached. There is thus an
attractor 
of heading at
ð
/
;
_
/
Þ ¼ ð
w
 g
;
0
Þ
:
The model consists of a system of differentialequations that generates a trajectory through the statespace
ð
/
;
_
/
Þ
—that is, a sequence of headings andturning rates—for a given speed of travel (Fajen andWarren2003; Fajen et al.2003). To produce the req- uisite turns toward the goal, imagine that the headingdirection is attached to the goal direction by a dampedspring. Analogously, the model describes the angularacceleration of heading as a function of the currentgoal direction (
w
 g
),
/
¼ À
b
_
/
À
k
 g
ð
/
À
w
 g
Þð
e
À
c
1
d
 g
þ
c
2
Þ ð
1
Þ
where
b
,
k
 g
,
c
1
, and
c
2
are parameters. As the agentmoves through the environment to the next (
 x
,
z
) po-sition, the goal angle and distance change, and the nextheading and turning rate are determined from Eq. 1.The ‘‘damping’’ term
À
b
_
/
acts as a frictional force thatresists turning and is proportional to the turning rate,which tends to keep the path straight and prevents theheading from oscillating about the goal; parameter
b
expresses the ratio of damping to the body’smoment of inertia, in units of s
–1
. The ‘‘stiffness’’ term
k
 g
(
/
w
 g
) reflects the empirical finding that angularacceleration increases linearly with the target-headingangle over some range (for humans, Fajen and Warren2003, as well as flies, Reichardt and Poggio1976); parameter
k
 g
expresses the ratio of stiffness to momentof inertia, in units of s
–2
. The stiffness is modulated bythe distance of the goal, reflecting the finding thatturning acceleration increases exponentially withnearer goals
ð
e
À
c
1
d
 g
þ
c
2
Þ
(Fajen and Warren2003).The constant
c
1
determines the decay rate with dis-tance in units of m
–1
, and
c
2
determines a minimumvalue so acceleration does not go to zero at large dis-tances, and is dimensionless. The model fits the meanhuman time series of target-heading angle for a sta-tionary goal extremely well, with a mean
2
= 0.98 forparameter values of 
b
= 3.25,
k
 g
= 7.50,
c
1
= 0.40, and
c
2
= 0.40 across all conditions.A second component of the model accurately de-scribes obstacle avoidance by defining a similar springforce that repels the heading away from the directionof a stationary obstacle (Fajen and Warren2003; Finket al. in press). A similar distance term is needed be-cause humans turn to avoid near obstacles before theyrespond to more distant obstacles or goals. The direc-tion of travel is thus determined by the net resultant of all forces acting on the agent at any moment. A routethrough a complex scene unfolds as the agent tracksthe local heading attractor, whose location is deter-mined by the sum of the goal and obstacle components.Our modeling strategy is to fix the parameter values foreach component using a basic data set, and then usethe model to predict human paths with novel configu-rations of objects.We should point out that the goal component(Eq. 1) is a dynamical system that corresponds to astandard proportional-derivative controller, but with anonlinear stiffness term that depends on distance,whereas the obstacle component is a dynamical systemthat does not correspond to a PD controller. The per-tinent psychological question is what variables areregulated by this relatively simple system to achieveadaptive locomotor behavior.
Strategies for intercepting a moving target
There are a number of ways the steering dynamicsmodelmightbeextendedtoaccountforhumanpathstoa moving target, in the case of a constant velocity targeton the ground plane. In this section, we introduce four
Fig. 1 a
Definition of variables for the model.
b
Intercepting amoving target will be successful if the agent (i) matches thetransverse speed of the target, so
v
,
a
=
v
,
m
, and (ii) approachesthe target, so
v
,
a
>
v
,
m
. The resulting path has a constant target-heading angle
b
and a constant bearing direction
w
m
Exp Brain Res (2007) 180:303319 305
 123

You're Reading a Free Preview

Download
scribd
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->