CURVES OF GENERAL PURSUIT

By ARTHUR BERNHART

O UR discussion of pursuit curves has considered: [AJ Achilles'

one-dimensional pursuit of Zeno's tortoise; [B] Bouguer's lignes de
poursuite for a linear track; and [CJ Circular track as studied by Hath-
away and others. We conclude with a glimpse at [D J Differential
equations valid for arbitrary track and variable speeds, and [E]
Extraneous problems sometimes confused with the pure pursuit prob-
lem.
[Dt J Pursuit with Variable Speed. As early as 1732 Pierre-Louis
Moreau de Maupertuis had derived the differential equations for arbi-
trary pursuit ("Sur les courbes de poursuite," Memoires de l'Academie
Royale des Sciences, p. 15-16) showing that they applied not only to
Bouguer'sparticular case (Ibid., p. 1-14) but also where the track of the
pursued is an arbitrary curve and where the ratio of speeds is not neces-
sarilyconstant.

---
Q'
p
~S
--
~ I ---------
f
--,(1<1:

G)_
Fig. 1

The point Q traverses an arbitrary track Q(w) with speed n = dw/dt

and is pursued by the point P along the curve P(s) with speed m ":'
ds/dt. The ratio of corresponding arc lengths ds/dw = min IS ar.bl-
trary, but the velocity of pursuit dP /dt has always the same direction
as the separation vector PQ.
Received by the editor, April 21, 1958.
IS9
100 CURVES OF GENERAL PURSUIT

Using rectangular coordinates, P (x,y) and Q (tt,v), these dif-

ferential conditions may be written
~ ~ ~ ds mdw (1)
u-x v-y r nr
where r' = (x - u)' +(y - v)'. For uniform pursuit (min ~ con-
stant) Maupertuis suggests an equivalent geometric formulation: The
curve QQ' is given; find the curve PP' such that any two of its tan-
gents PQ and P'Q' intercept an arc QQ' proportional to the arc PP'.
[D2] Curvature. We have seen how Dubois-Ayme and Saint-
Laurent aroused interest in the radius of curvature. A contribution
by Maurice d'Ocagne, "On the center of curvature of curves of pur-
suit," [Bulletin de fa Societe Mathematique de France, v. 11 (1883),p.
134--135] is valid for general pursuit. The new result involves a simple
geometric construction:
Layoff QT = PQnlm tangent to the track so that if the velocities
remained constant at their instantaneous values, then Q moving at
speed n would reach T by the time P moving at speed m had reached
the initial position of Q. Then PQT is the triangle of velocities drawn
to the scale PQ = m, QT = n, and such that PT represents the result-
ant velocity.

C :,
'-'
,,-,
,,
T
,
,

P~:,
--"--..l,---'
" , a
, , I
, I
, I
'~, " I
" \ I
-,'I., M
Fig. 2
 ARTHUR BERNHART 191

At Q draw QM perpendicular to PQ, and at P draw PM perpendicu-

lar to QT, intersecting at M. The perpendicular to PT from M meets
the perpendicular to PQ from P at some point C, which is the center of
curvature of the pursuit curve P(s).
For, ds:dw = ds/dt:dw/dt = m:n ~ PQ:QT. Since triangles
PQT and CP M are similar, the ratios PQ: QT and CP :PM are equal.
Thus ds:dw = CP:PM, and ds/CP = dw/PM. Now the latter ratio
may be written
dw _ dw sin TQP dw'
PM - PM sin MPC PQ'
wheredw' is the component of dw which is transverse to PQ. But the
tangent PQ to pes) turns through this angle dw' /PQ = ds/ p, where p
isthe radius of curvature. This shows that p = CP and C is the center
of curvature.
This geometric theorem of Ocagne provides a neat formula for the
radius of curvature p. Let r = PQ and let </> be the angle between the
twovelocities. From similar triangles
CPIPM = PQ/QT = min,
hence p = PM min. But angle QMP is </> so that sin </> PQ/PM,
and PM = r /sin </>. This yields
mr
p = . . (2)
nsm </>

Equation (2) has been obtained for the special case of linear track
[B9,eq. (23)] but Ocagne's theorem shows that it holds for arbitrary
track and for variable speeds. Ocagne mentions particularly that the
curvature of pes) does not depend on the curvature of Q(w) but only on
the relative velocities.
[D3] Barycentric Pursuit. The study of centers of gravity led
Ernest Cesaro to another type of pursuit. He finds ["Properties of a
Pursuit Curve," Nouvelles Annates de Mathematiques, (1883) p. 85-89]
that the center of gravity of an arc with one end fixed pursues the mov-
ing end. This provides a complete answer to an earlier question he
posedin Nouvelles Correspondence [v. 5 (1879) p. 110].
Employing the notation of [Dl] the centroid P(x,y) of the arc w ~ t
generated by Q(u,v) is determined by the first moments tx = f udt
and ty fvdt. On differentiating, tdx = (u - x)dt and tdy. = ~v
y)dt, we find that P pursues Q according to eqs. (1). The Simplifica-
tion w = t implies that Q moves with unit speed, n = 1. If m is con-
stant then the pursuit is uniform, but for barycentric pursuit Cesaro
192 CURVES OF GENERAL PURSUIT

derives the condition

ds r
m= (3)
dt t
Equation (3) follows from (1) and the relation ds? = dx' +
dy' which
defines the differential of arc. Differentiating the relation whichde-
fines r = PQ, Cesaro finds the radial velocity

at
dr
= -m + cos et>. (4)

Since his derivation of (4) is based on the general pursuit relations (1)
without the restriction (3), it holds for arhitrary pursuit. The same
formula was obtained [B5, eq. (10)] for uniform pursuit. Cesaro ex-
plicitly recognizes the generality of (4) in a later paper [seeD4].
But (3) permits another computation for the radial velocity. From
r = mt we have dr/dt = d(mt)/dt = m +
tdm/dt. Combining this
result with (4) Cesaro writes
dm
cos et> 2m + tdI (5)

Can barycentric pursuit also be uniform? Holding m constant in (5)

we obtain the necessary condition cos et> = 2m, which shows that ¢
must remain constant. Discarding the trivial case of linear track, ¢ =
0, Cesaro examines the case of a logarithmic spiral.
Introducing polar coordinates R and A let the point Q track the
spiral J

u = R cos A, v = R sin A,
where

log R = log C +A cot et>/2. (6)

The radius R assumes the value C when the polar angle A vanishes.

Fig. 3
 ARTHUR BERNHART 193

Measuring the arc OQ along (6) from the pole 0, the rectangular co-
ordinates of its centroid P(x,y) are given by
x = mR cos (A - </»,
y = mR sin (A - </», (7)
wherem = II, cos q,. Equations (7) show that the ratio OP 10Q = m
and the angle POQ = </> are both constant. Hence triangle POQ re-
tains the same shape, so that P and Q move on similar spirals with a
common pole. Therefore the track (6) and pursuit curve (7) are
congruent logarithmic spirals, differing by the rotation </> +
2 tan q, log
m.

Q
c

o T" T
Fig. 4

Let Tbe the foot of the perpendicular from the pole 0 to theline.QT
which is tangent at Q to the track. Let Q* and T* be the midpoints
of OQ and OT, respectively. Then, the centroid P of the arc ~Q h~s
on the median line QT* of right triangle OTQ. For, the loganthmlc
derivative d(log R)ldA of (6) is cot </>/2,and, as this is cot OQT, we have
tan OQT = 2 tan </> = 2 tan PQT. (8)
194 CURVES OF GENERAL PURSUIT

But, tan T*QT/tan OQT = T*T/OT = '/" by construction. Accord-

ingly tan T*QT = tan ¢ = tan PQT, which means that P, Q, and T'
are collinear. (Incidentally, P is between T* and Q whenever cos' ~
exceeds '/3') Since OP = MOQ = '/,OQ cos ¢ = OQ* cos ¢, it follows
that OPQ* is a right angle, and P lies on the circle with diameter OQ'.
Recapitulating, angle POQ = ¢ = angle PQT, and angle Q*PO =
Ij,1r. Given triangle OTQ we have three loci any two of which willde-
termineP.
Cesaro concludes his first paper on barycentric pursuit by givingthe
location of the centers of curvature C and D of the P and Q spirals,re-
spectively. The point D is diametrically opposite Q on the circle OPQ,
while C is the midpoint of the chord PD.
[D3] Generalized Barycentric Pursuit. Three years after his first
barycentric study of curves Cesaro considered the generalization in-
volving a variable linear mass density A. [Nouv. Ann. Math., v. 5
(1886), "Sur les Lignes de Poursuite," p. 65-83; "Les lignes barycen-
triques,' p. 511-520.] The centroid integrals for a plane curve Q(t)
become
x f Adt = f uAdt, and
y J' Adt = J' vAdt. (9i)
Their differential counterparts are
Ldx = (u - x)Adt, and
Ldy = (v - y)Adt, (9d)
in which we have introduced the abbreviation L = f Adt for the total
mass of the arc. Equations (9) show that P(x,y) pursues Q(u,v) ac-
cording to the conditions

dx .sz: ds »dt (10)

u-x v-y r L'
Comparison of (10) with (1) yields the relative speed m of generalized
barycentric pursuit,

ds Ar
m = dt = L . (11)
Substituting r Lm/A in (4) we obtain the generalization of (5),
namely

cos </> 2m + L '!..

dt ('!!)
A .
(12)
 ARTHUR BERNHART 195

From eq. (11) we have dL/L = mdt/r = ds/r. If r is given as a func-

tion of s, then L may be computed by an integration, thus

log L = f ~. (13)

Considering infinitesimal displacements PP' = ds and QQ' = dt and

arbitrary pursuit speed m = ds/dt, Cesaro writes the law of sines for
triangle PQQ' in the form
dt r r+dr+ds
ds/p = sin (¢ - ds/p) sin (.- - ¢) .
The first equality yields the radius of curvature (2), and the second
equality yields cos ¢ = d(r +
s)/dt which is equivalent to the familiar
formula (4) for the radial velocity. But he is specially interested in the
angle ¢ between the velocity vectors. This angle oj attack changes at
the rate
d¢ 1 m !._sin¢
(14)
dt a p a r

wherethe new variable a represents the radius of curvature of the track

Q(t), just as p gives the radius of curvature of the pursuit curve P(s).
Equation (14) follows directly from kinematical considerations. The
velocity dQ/dt turns at the rate n/a, with n = 1; the velocity dP /dt
turns at the rate m/ p; and the difference of these angular velocities
givesd¢/dt immediately.
On choosing L = t, h = dL/dt = 1, these general equations reduce
to the particular type treated in [D2J, since (11) becomes (3); but
with L an arbitrary function of arc length t, Cesaro's pursuit (10) is
correspondingly arbitrary. Thus any pursuit, satisfying (1) by defini-
tion, may be interpreted via (11) as generalized barycentric pursuit.
For example, if d(log L)/dt is inversely proportional to Y, then the
pursuit is uniform, m being the proportionality constant.
[D4] Attack Curves. One type of pursuit which Cesaro examines
particularly is attack from a constant angle ¢, where d¢/dt vanishes,
hence

p = ma and r = a sin ¢. (15)

Equation (15) means geometrically that P lies on an attack circle which
is independent of the particular attack angle. It is tangent to the
track at Q, and the diameter QD of length a terminates at the center
of curvature D. Cesaro says this result is due to Reamur.
 196 CURVES OF GENERAL PURSUIT

Among the attack curves [developpoides J associated with a given

tract QU) the eiolute [developpee] corresponds to the case q, = If,.·
Otherwise said, the center of curvature attacks any curve at right
angles. The relation r = a sin </> yields r = a; while the differential
equation dr/ dt = cos </> - m reduces to ~t (r + s) = O. Integration
yields s = t:
-r +constant, or simply s* = -r, if we measure the arc
s* from the point where r = O. Again, the corresponding radius p* is
given by p* = am = ads/dt = -ada/dt. Given the functional rela-
tion between the track length t and its curvature a, the equations
s* = -a, and
da
p* = -a dt (16*)

express parametrically the relation between the intrinsic coordinates,

s* and, of its evolute.
The mass L * and linear density X* which interpret any evolute as a
pursuit curve (9) may be written
L* ~ I/a, and X* = pO/a'. (17*)
For eq. (13) yields log L* = -da/a = -log a, so that the massL* is
'. dL* -I diJ
inversely proportional to the radius a. Then X* = -;Ii = -;.- di
which combines with (16*) to give (17*).
More generally, let </> be constant but not necessarily a right angle.
Then eqs. (16*) are replaced by
s = t cos </> - a sin </>, and
p = a cos </> + p* sin </>, (16)
where p* is the value of p for </> = '/''1r as given in (16*). We shall call
the P locus defined by (16) the attack curve at the angle q, of the Q
locus.
Equation (16) for p means geometrically that the center of curvature
C for each attack curve lies on the circle of diameter DE, where D and
E are the centers of curvature of the track QU) and its evolute pesO),
re~p~ctively. For PD = a cos </> and DC = p* sin </>. Cesaro credits
Mtn1c~ and Habich with discovering this result independently.
Lettmg D(</»Q symbolize the transformation (16), the accompanyiog
figure shows that D(</»D('/''1r)Q = C = D('/''1r)D(</»Q. In words, the
attack curv~ of the evolute is the evolute of the attack curve. This is
only a particular case of a theorem by Lancret, which states that
 ARTHUR BERNHART 197

D(<pJ)D(q,,)Q = D(<p,)D(</>J)Q; so that, in a succession of attacks at

varying angles, the operators D(</» commute.
If the track curvature 1/a is always tan </>It, then s and the attack
speedm vanish, and P is a fixed point, This occurs when the track is
a logarithmic spiral crossing lines issuing from pole P at the constant
angle<p.

Fig. 5

Let the attack circles DPQ and DeE of the track and its evolu,te in-
tersect again in point H. This poing (point de rebroussement) IS the
foot of the altitude to the hypotenuse in right triangle QDE, whose legs
are the radii of curvature in position. The poi';'ts Q. a,;,dI-!
separate
the points on the attack circle into two arcs which distinguish attack
 198 CURVES OF GENERAL PURSUIT

proper from flight. For, the pursuit speed m = pia becomes negative
when the curvatures have opposite sense, indicating flight. At the
transition point H the speed m and curvature p both vanish. Among
other interesting results Cesaro proves (p. 73) that the tangents to the
loci of Q and H form the sides of an isosceles triangle with base QH.
Accordingly, both are tangent to the attack circle.
We conclude this section by setting down the generalization of eqs.
(17*), namely
L=eE,ot·/a, and A=pL/a'sinq" (17)
where E indicates the total angular deviation of the track. For,
evaluating rand ds by eqs. (15) and (16), formula (13) may be written
log L + log a = cot q, f dt/a, and f dt/a = E. Again, from (11)we
have AIL = mfr. But m ~ p/ a and r = a sin q, by the conditions (15)
for attack. For the particular case of flank attack, q, = ';'''' the eqs.
(17) reduce to the form (17*) for an evolute.
The study of barycentric pursuit led Cesaro to the concept of sum-
mability of divergent series. He makes the connection of ideas ex-
plicit in his paper "Nouvelles remarques sur ... la Theorie des series"
[Nouv. Ann. Math., v. 9 (1890), p. 364J.
[DS] The Tractrix. In order to study various types of pursuit
curves it is convenient to choose relative polar coordinates (r, q,), where
r is the separation distance PQ and q, is the angle between the velocity
of pursuit dP /dt and the track velocity dQ/dt. The previous section
was concerned with pursuit which kept q, constant. It is intuitively
simpler to demand that r be constant, which we shall do in this section.
With r constant and the track linear the pursuit curve becomes the
familiar tractrix, such as would be obtained if Q dragged P by a ropeof
len~h r. If the positive y-axis is taken as the track, Q being on the
x-axis at (r, 0) as P passes the origin, the tractrix has the equation

y = r log r + Yr' - x' _ Y r' _ x'. (18)

x
On the scale r = 1 this locus is the involute of the catenary x = cosh
y. The tractrix was studied by Huygens and by Bourie ["PropriHes
de la tractrice," Memoires de l'Academie Royale des Sciences (1712), p.
215-225]. Professor C. A. Scott in Encyclopedia Americana mentions
the use of this curve by A. F. Beltrami for an interpretation of non-
Euchdean geometry ["S: . d' '. .
I'd " aggzo mterpretaeione della geometrw non-
eue 2 ea, (1868); Ann. de l'Ec. Norm., v. 6 (1869) p. 251]. See alsoa
~ap:r by E. ~esaro, "Sur la tractrice," (Mathesis, v. 2, p. 217-219) for
I s uahty With the logarithmic spiral.
 ARTHUR BERNHART 199

During the eighteenth century the tractrix was called "curve of

pursuit," a term which also applied to the courbe du chien studied by
Bouguer and Dubois-Ayme [B7]. This ambiguous terminology has
caused Niels Nielsen to confuse them in an off-the-cuff comment in
Geometres Francais. "We remark in passing that the tractrix is noth-
ingelse but the curve of pursuit, studied later by Bouguer and by Mau-
pertuis; we are able to add that neither Bouguer nor Maupertuis
mentioned their precedessors." Our exposition of pursuit has included
many examples of the truism that problems with strong intuitive appeal
are often solved more than once. But the tractrix, with r = constant,
and the courbe du chien, with m = constant, are two different prob-
lems.
The simple tractrix (18) implies a linear track. The pursuit restric-
tion that r be constant permits the obvious generalization to an arbi-
trary track. The equations m = cos r/J, dr/J/dt = l/a - sin rf>!r,and
p = r cot r/J apply here.
[D6] Aberrancy Curves. The center of curvature D pursues the
point of contact Q of the osculating circle, with an angle of attack
whichis always a right angle. Can the result be extended to the center
A of the osculating conic? Such conics were investigated by Abel
Transon, "Recherches sur la courbure des lignes et des surface" [Jour.
Liouville, v. 6 (1841), p. 191-208J. Using rectangular coordinates the
track Q(u,v) may be approximated in the neighborhood of the origin by
the series

1 1
v = a,u + '/,a,u' + fja,u' + 24a,u4 + ... ,
in which each coefficient an is the value of the corresponding derivative
dnv/dun at u = O. For the coordinates of the center of aberrancy A (x,y)
Transon obtains

x = + 3a.a,/(5a,'
u - 3a,a,), and
y = v + 3a.(a,a, - 3a,')/(5a,' - 3a,a,). (19)
The quotient (v - y)/(u - x) = a, - 3a,'/a, gives the slope of the
axis of aberrancy QA, which deviates from the normal QD by an angle
AQD such that
tan AQD = a, - (1 + a,')a,/3a,' = p* /3a, (20)
where a and p* indicate the radii of curvature for. th~ track and its
evolute, respectively. The radius of aberrancy QA IS gtven by .
 200 CURVES OF GENERAL PURSUIT

r =
3a, Va,' + 9a,,' (21)
3a,a, - Sa,'
This radius becomes infinite whenever the osculating conic is a parab-
ola.
Choosing axes so the slope a, vanishes, and choosing the scaleso
a, = 2, the conic has the equation
72y = 72x' + 12a,xy + (3a, - 2a,')y'.
Using Transon's formulas (19) it is easy to verify that the center of
aberrancy does pursue the point of osculation. For, dyldx = (v - y)/
(u - x), in agreement with eqs. (1) of section [DI]. But the angleof
attack, which is the complement of angle AQD, generally varies as
Q(u,v) moves along its track, being constant only on a logarithmic
spiral.
[D7] Miscellaneous Pursuit Problems. The eqs. (I) are so general
that any two plane curves may be interpreted as track and pursuit,
so long as P and Q are properly coupled. From each point Q on the
first curve, we draw any tangent QP to the other. If there happens
to be several tangents we obtain several points P which pursue Q along
the same course. Conversely, at each point P on one curve we may
consider the tangent line and its intersection Q with the other curve.
If there are several intersections each generates the same track. If the
loci of both P and Q are given the problem would be to determine the
coupling, and the corresponding ratio of speeds. .
Normally the Q track is given and we seek to find the P locus by in-
t~grating (I) subject to some auxiliary condition which implies a defini-
tion of the speed ratio. Thus Bouguer [B] and Hathaway [C] sup-
pose explicitly that the speeds are uniform, with linear and circular
tracks, respectively. The tractrix [D5] with fixed separation, and tbe
attack curves [D4] with fixed angle of approach imply speeds m = cos
1> and m = pia, respectively. The barycenter [D3], the center of
curvatu~e [D3], and the center of aberrancy [D6 J each generate a curve
of pursuit.
T~e examples ~f pursuit discussed here are by no means compre
hensivo. Cantor in Geschuhte der Mathematik (v, 4, p. 506) mentions
t~e early mterest of Lambert (1769) and Euler (1775). Beginning
With a 'pap,er by C. Sturm, "Extension du problema des courbes des
poursUl~e, the Annales de Mathematique [v. 13, (1822-23)] carnes
su~estlOns by Saint Laurent, Querret, Tedenat, and others.
"S e woul? like to mention two papers by V.' Nobile. In the first,
ullo studio mtrinseco della curva di caccia," [Palermo Rend., v. 20
 ARTHUR BERNHART 201

(1905), p. 73-82] he obtains the differential equations in intrinsic co-

ordinates, s and p, and integrates them for a number of cases. In the
second,"Sui problema della curva di caccia," [Batt. Giorn., v. 46 (1908),
p. 135-143] he introduces an auxiliary curve for the geometric deter-
mination of the pursuit curve.
The drama of pursuit may be played on a three-dimensional stage.
But if the track is linear the curve of pursuit lies in the plane deter-
mined by this straight line and the initial position of the pursuer.
Thus C. C. Puckette in "The Curve of Pursuit" [Math. Gazette, v. 37
(1953), p. 256-260] shows that the space curve formulas of Frenet
reduceto the differential eqs. [B] for a plane curve. Bouguer's 1732
solutionfor constant speeds is more recently packaged with tables and
graphs for the Office of Scientific Research and Development ["Pur-
suit Curve Characteristics," O. S. R. D. No. 3721 (February, 1944)].
In the pursuit of a bomber flying horizontally with uniform velocity
by a fighter the kinematical problem is still two dimensional if the pur-
suit speed rn, distance r, or attack angle 1> is constant. But if m varies
with vertical altitude the plane formulation requires that m be given
as a function of the distance from P to the linear track. Again, if the
speed of the bomber is comparable with that of the bullets from the
pursuing fighter P, it is necessary to aim the guns at a point Q' ahead
of the moving target Q. Lead pursuit satisfies the tangential require-
ment of Maupertuis [Dl] if we replace Q by its anticipated position
Q'. But then the track speed dQ'ldt depends in a complicated way on
the instantaneous triangle PQQ'. It seems simpler to deal with P and
Q and to regard the angle QPQ' as a perturbation from pure pursuit.
The psychology of pursuit assumes that the motion of the quarry Q
along its track is independent of the proximity of the pursuer P. A
newspecies of problem is posed if the track of Q is to be determined in
order to achieve some object, such as to enable Q to escape to some
point of refuge Q*, or to decoy P to a neutral zone, say PQ* > d, or
into a trap at P*. Such problems have more than academic interest,
but seem to have been neglected. .
In this category is the concluding essay "Lion and Man" given by
J. E. Littlewood in A Mathematician's Miscellany (London. 1953).
"A lion and a man in a closed arena have equal maximum speeds.
What tactics should the lion employ to be sure of his meal?" .The
question was invented by R. Rado. If the lion chooses p.ursmt, Little-
wood remarks that capture takes an infinite time provided the ~an
runs in a circle. The lion P can defeat the circular strategy by keepmg
on the radius OQ from some fixed point 0 to the instanta",leous poslt~on
Q of the man. But if the man alters his track to a certam broken Ime
 202 CURVES OF GENERAL PURSUIT

spiral, A. S. Besicovitch discovered that the lion again dies of starva-

tion if he stubbornly keeps to the radius. Littlewood argues that the
man can avoid capture whatever the lion does. He urges the man at
the nth stage to travel a route Q.Q.+1 perpendicular to the displace-
ment vector p.Q. from the position of the lion P; and man Q. at the
beginning of that stage, continuing in that direction until Q.H is beyond
its closest approach O. to a fixed point 0 by the amount I. = O.Q.tl =
1In-al'. He says that for a suitable choice of the constant II this in-
finitely long path will stay inside a circle with center 0 inside the arena,
We are not convinced that this disposes of the general case, however,
for the lion might be so disagreeable as to run parallel to the man, so
that the man must either travel along a straight course or alter his
strategy. This seems to be a typical theory of games situation. If
either player knows the other's fixed strategy he may choose his own
to better advantage. At any rate the lion's meal, like happiness, is
not to be gained by pursuit.
The history of pursuit curves has attained international stature.
Thus Andre Clarinval wrote "Esquisse historique de la courbe de
poursuite" for UNESCO [Archives internationales d'Histoire des Sci-
ences, v. 10, 38 (1957), p. 25-37]. He devotes four pages to the
solution of Bouguer, but a typographical error mars the integral (p.
28). He condemns the commendable solution of Dubois-Ayme [B7]
as "replete with errors," and substitutes the "corrected" formula of
Saint-Laurent (again with typographical errors, some old, some new).
This historical sketch is chronological, and includes several references
t.o the case of circular track [C], with precise quotations of various pub-
lished problems, and with brief but accurate reviews of the longer pa-
pers devoted to their solution. (American readers may smile at hisas-
sumption that F. E. Hackett was named Johns Hopkins.) Clarinval
ac~uses .Samuel Jones (Appendix to the gentleman's Diary, 1839; re-
p~mted in Math. Gazette, v. 15) of "deliberately ignoring" French con-
tributors, and he concludes with a number of references that cross na-
tio~al barriers. Among these are R. Hoseman, Fiat Review of German
S~tence, 19~~-194~, Applied Mathematics, part I, p. 269; Luke Chia-
LIU yuan, Horning and Navigational Courses of Automatic Target-
Seekmg Devices," J. App. Phys., v. 19 (1948), p. 1122-1128; Ralph
Hoyt Bacon, "The pursuit course," J. App. Phys., v. 21 (1950), p.
1065-1066; and Pedrazzini, Period, Mat. Ital., n02 (1949), p. 99-103.
[E! Extraneous Problems. We say "P pursues Q" if and only if,
t~e dIsplacement vector PQ is tangent to the path of P in accordance
WIth .eqs . (1) . In non- t ec hni cal usage "pursuit" .' may have other
meanIlUgs, which are extraneous to our definition We list a few ex-
amp es. .
 ARTHUR BERNHART 203

[EI] Navigation. Does one swimmer P pursue another Qwhen his

course is toward Q though his heading is somewhat upstream? If P
swimsthrough the water medium at speed e, and the current flows with
speed1at an angle rjJ with the desired course PQ, then P must head off
courseby a correction angle e in order to make good his course.
Applying the law of sines to the triangle of velocities, we obtain
e/sin rjJ = l/sin f = g/sin (rjJ+ f), where g is the resultant speed.
+
Therefore e = arcsin (f sin rjJ/ f) and g = 1sin (rjJ f)/sin f. Inciden-
tally, the magnitude of eis unchanged if either the current or the course
is reversed. These concepts apply to air navigation where e is air-
speed.j' is wind speed, and g is ground speed. In case the course is at
right angles to the current, f = arcsin (f / e) and g = Ve' - 1'. Such
formulas are crucial in interpreting the interferometer ether drift ex-
periments of Michelson and Morley. In general, P does not pursue Q
with speed e. Using a ground fixed reference frame, P pursues Q with
speedg; but with axes moving with the current, P does not pursue Q.
Literally speaking, P pursues Q along-the-ground but not through-
the-water!

Fig. 6

Exploiting the ambiguity in what is meant by sw!nu."ing ::directly

toward" an object is a pamphlet by L. T. Houghton entItled A.Com-
man Sense Solution of a Pursuit Problem That Has Been Considered
Unsolvable by Many Eminent Mathematicians." [This r~ference,
among others, is supplied by R. C. Archibald and H. P. Manmng, Am.
Math. Mo., v. 28 (1921), p. 93.] .
A similar problem was proposed by Thomas de S~mt-Laurent
[Annales de Gergonne, v. 13 (1822), p. 289]. A dog SWlll1S a canal
headed toward his master who walks along the other bank. In the
 204 CURVES OF GENERAL PURSUIT

same volume Querret and Tedenat showed that the problem may be
interpreted as pursuit (p. 392) by choosing axes which move withtbe
current.
Howard Eves and E. P. Starke selected two pursuit problemsas
among the 400 best problems appearing in the Mathematical Monthly
from 1918 to 1950 for inclusion in the Otto Dunkel Memorial Problem
Book [Am. Math. Mo., v. 64 (1957)]. One by H. E. Tester (#3942)is
Bouguer's linear track, with m = 2n. The other (#3696) by J. B.
Reynolds is more involved. "A dog directly opposite his master on the
banks of a stream, flowing with uniform speed, swims at a still-water
speed of two miles per hour heading directly toward his master at all
times. The man notes that the dog does not stop drifting down
stream until he is two-thirds across measured perpendicularly to the
banks, and that it takes five minutes longer to make the trip than if
the water had been still. How wide is the stream?"
Artemas Martin proposed the following problem in Educational
Times. A boy walked across a horizontal turntable while it was in
motion at a uniform rate of speed, keeping all the time in the same
vertical plane. The boy's velocity is uniform with respect to the table,
and equal to m times the velocity of a point in the circumferenceofthe
table. James McMahon showed [Math. Questions from E. T., v.51
(1889), p. 158] that the curve described on the table of radius a is an-
other circle of radius ma, and that the boy walks a distance s = 2ma
arc sin Va' - d'/ma where d is his nearest approach to the center of the
table. The turntable problem is also found in the Mathematics Visitor
[v. 1 (1878), p. 37].
[~2] Dynamic Pursuit. The term pursuit is often applied to
motions where the physical forces must be considered. In such dy-
namic. situations the velocity dP /dt no longer has the direction PQ.
Thus In Keplerian motion it is the acceleration of the planet (not its
velocity) which points toward the sun.
George H. Handelman wrote his doctoral dissertation on "Aerody-
namic Pursuit Curves for Overhead Attacks" [Jour. Franklin Institute,
v. 247 (1949), p. 205--221], and amplified the discussion in reports to
the Office of Scientific Research and Development with W. Prager
[106.1R] ,,:nd with W. R. Heller [106.2R]. This thesis includes four
graphs which show how the aerodynamic pursuit curve deviates from
the pure pursuit curve. An elegant extension to three dimensionswas
made by L W Coh . "E . it
C ,,': en In quations for Aerodynamic Lead PurSU!
ur-ves . (Apphed Math. Panel Report 153.1R).
Pursuit curves were . d wi . ] d
d ou btl ess much of th concerve
t
WIth piracy on the high seas [B1 an
'. .. t
e recen research IS classified as a military secre .
 ARTHUR BERNHART 205

But man does not live by aerial dogfights alone. While one member
of Napoleon's expedition to Egypt discovered the Rosetta stone,
another observed the first courbe du chien in the sand. But, if life
began in the sea, long before this trace launched a thousand other
spoors,there must have been countless examples of dynamic pursuit in
the ocean depths. The following query was published in L'Tntermedi-
aire des Mathematiciens, v. 1 (1894), p. 183. "Supposing that the light
of a star takes twelve hours to reach the center of the circle which the
star traverses in twenty-four, what is the path of a fish which, starting
from the center, moves with a given speed constantly directly to the
point where it sees the star" [quoted by R. C. Archibald and H. P.
Manning in Amer. Math. Mo., v. 28 (1921), p. 92]. If the fish remains
near the center of its watery abode and if the light rays are considered
parallel, then the fish swims in a circle away from the star toward its
diametrically opposite image. But if an exact solution is demanded,
we need to know the index of refraction and the radius of the hydro-
sphere. The fish swims toward the image of the star, but the position
of this image depends on the eccentric location of the fish, so this is not
pursuit of a given point.
[E3] Geodesic Pursuit. Benjamin F. Finkel, who founded the
American Mathematical Monthly, proposed [v. 9 (1902), p. 271] the
followingproblem. "A dog at the vertex of a right conical hill pursues
a foxat the foot of the hill. How far will the dog run to catch the fox,
if the dog runs directly toward the fox at all times, and the fox is con-
tinually running around the hill at its foot, the velocity of the dog being
6 feet per second, the velocity of the fox being 5 feet per second, the
hill being 100 feet high and 200 feet in diameter at the base?" The
next year [v. 10 (1903), p. 104-106] a solution by G. B. M. Zerr was
published. A differential equation was derived and the numerical
solution s = 314 feet was obtained by introducing the approximation
that vertex-dog-fox are collinear at all times. But J. E. Sanders ob-
jects (p. 205) that "the dog cannot run toward the fox at all times and
keep between the fox and the vertex of the hill. If he keeps between,
Sanders computes s = 167 feet from the formula s = md arcsin. 11m
whered is the original distance between dog and fox. R. C. Archibald
and H. P. Manning comment: "Professor Finkel made clear the
equivalence of his problem" with circular pursuit, conceiving "the sur-
faceof the COneto be spread out on a plane." The reader will note that
this "equivalence" holds only if we interpret "directly toward" as the
shortest path between dog P and fox Q when the curved sur~ace.of t~e
COneis developed into a plane. We call this geodesic pursuit SIDceIt
suggests a natural extension to pursuit on a non-developable surface,
 206 CURVES OF GENERAL PURSUIT

bringing to mind a spider chasing a fly across a doughnut. But the

geodesic path is not the only reasonable interpretation, especiallysince
the dog cannot see the fox. An alternate point of view, suggestedby
motion of a particle constrained to move on a smooth surface, wouldbe
to project the vector PQ onto the tangent plane at P.
An earlier version with the hill 200 feet in radius appeared in 1888in
Mathematical Questions from Educational Times. The proposer obtains
a differential equation which is a quarter page long but does not solve
it. A numerical solution by B. C. Wallis [v. 5 (1904), p. 30] givess =
266 feet while the fox runs 222 feet.
This sort of problem is reminiscent of the spider which must follow
a path along the walls and ceiling of a room. To obtain a solution tbe
walls must be folded onto the same plane. Since there is more tban
one way of efiecting this projection, the alternative paths must be com-
puted for the room dimensions given, and then the least of these rela-
tive minima is the acceptable answer.
And God said, "Let there be light"; and there was light. The
Hebrew text uses the same word for the command and its fulfilhnent.
But we can imagine the angelic architect asking for more details:
"What path shall light follow in going from P to Q?" And the answer
might have been, "Don't bother me with such details. See that it
makes the trip in a minimum time." From this minimal principle one
finds that for reflection the angle of incidence should equal the angle of
reflection, while for refraction at an interface the ratio of the sine of the
angle of incidence to the sine of the angle of refraction must equal the
ratio of speeds in the two media. And God saw that the light was
good.
UNIVERSITY OF OKLAHOMA