Two persons are running on a circular track either in the same direction or in the opposite direction, indefinitely. The speed of both of them is given, say $u$ and $v$. Now, how do I find out the the number of distinct points at which they will meet on the circle.

  • 1
    $\begingroup$ Have you tried to solve this problem yourself? $\endgroup$ – Ben Millwood Aug 30 '12 at 18:41
  • $\begingroup$ Yeah, well the first meeting point is C/(v-u), considering the circumference is C and v>u. Now the meetings points are going to be multiples of this. So nC/(v-u) is going to be equal to some kC/(v-u). Since, the path is circular, I'm expecting n can be decomposed to something less than C using modulo, but I'm getting nowhere with this approach :| $\endgroup$ – n0nChun Aug 30 '12 at 18:45
  • $\begingroup$ Yeah, I've gotta say, the problem is slightly less straightforward than it looks... it would have saved me some time if you'd mentioned that you'd already tried the obvious approach :) $\endgroup$ – Ben Millwood Aug 30 '12 at 18:50

$\omega_A=\frac{v}{C/2}$ and $\omega_B = \frac{u}{C/2}$ will be the angular speeds. Say you have angular speeds $\omega_A$ and $\omega_B$, and initial angles $\theta_{A0}$ and $\theta_{B0}$, then A's angle on time $t$ is $\theta_A(t)=\theta_{A0}+\omega_At$ and B's angle on time $t$ is $\theta_B(t)=\theta_{B0}+\omega_B t$.

The meeting angles are the solutions of the equation

$\theta_A(t)=\theta_B(t) + n2\pi$ or $\theta_{B0}+\omega_B t=\theta_{A0}+\omega_At + n2\pi$ where $n$ is $a$ integer.

This gives solutions:

$t = (n2\pi +\theta_{A0}-\theta_{B0})/(\omega_B-\omega_A)$

(assuming $\omega_B \neq \omega_A$, which would mean they'd always or never meet)

substituting this in A's angle equation gives meeting angles:

$\theta_{A0} + \omega_A(n2\pi +\theta_{A0}-\theta_{B0})/(\omega_B-\omega_A) = \text{constant} + \frac{\omega_A}{(\omega_B-\omega_A)}n2\pi$

as you can see, if $\omega_A/(\omega_B-\omega_A)$ is not rational ($p/q$ where $p$ and $q$ are integers), then you will get infinite meeting points because you will never get multiples of $2\pi$ added to the constant. And note that $\omega_A/(\omega_B-\omega_A)$ is rational iff $\omega_A/\omega_B$ is rational, which happens iff $u/v$ is rational.

So basically, if the ratio between the speeds is not $a$ rational number, there will be infinite meeting points.


This problem becomes much easier if one uses the right units and the right coordinate system.

Let the runners be $A$ and $B$, and let $A$ be the slower runner. It’s convenient to measure distance in laps and time in $A$-laps, where one $A$-lap is the time it takes $A$ to run one lap. Consider first the case in which $A$ and $B$ circle the track in the same direction.

$B$’s speed is $v+1$ laps per $A$-lap for some $v>0$, so he gains $v$ laps on $A$ in each $A$-lap and passes $A$ at those times $t$ such that $vt$ is an integer. At time $t$ runner $A$ is at position $t-\lfloor t\rfloor$ on the track, having completed $\lfloor t\rfloor$ full laps. Thus, the runners meet at those points $t-\lfloor t\rfloor=t\bmod 1$ at which $vt$ is an integer.

Suppose first that $v$ is rational, say $v=m/n$ in lowest terms. Then $vt=k\in\Bbb Z$ iff $t=kn/m$, so the runners meet precisely at those times $t$ that are integral multiples of $n/m$. Since $m$ and $n$ are relatively prime, $\{kn\bmod m:k\in\Bbb Z^+\}=\{0,\dots,m-1\}$, so the set of meeting points is

$$\left\{\frac{kn}m-\left\lfloor\frac{kn}m\right\rfloor:k\in\Bbb Z^+\right\}=\left\{\frac{k}m:k=0,\dots,m-1\right\}\;,$$

and there are $m$ of them.

If $v$ is irrational, the set of meeting times is $\{kv^{-1}:k\in\Bbb Z^+\}$, and the set of meeting points is $\{kv^{-1}\bmod 1:k\in\Bbb Z^+\}$, which is well-known to be dense in $[0,1)$ and therefore infinite. Of course one can observe directly that the set is infinite: if it weren’t, we’d have $mv^{-1}=nv^{-1}+k$ for some positive integers $m,n$, and $k$, and then $v^{-1}=\frac{k}{m-n}$ would be rational.

Now suppose that $B$ runs in the opposite direction around the track at a speed $u$, not necessarily greater than $1$. Let $v=u+1$, his speed relative to $A$’s position, so they meet whenever $vt$ is an integer. As before $A$ is at position $t-\lfloor r\rfloor=t\bmod 1$ at time $t$, so the analysis proceeds exactly as before.

The algebra to convert to other units is routine. In particular, if $r$ is the ratio of $B$’s speed to $A$’s, $v=r-1$ when they’re running in the same direction and $r+1$ when they’re running in opposite directions.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.