Consider a taxi station where taxis looking for passengers and passengers looking for taxis arrive according to Poisson processes, with mean rates per minute of 1 and 1.5.

/ Exams & Certification / By

Consider a taxi station where taxis looking for passengers and passengers looking for taxis arrive according to Poisson processes, with mean rates per minute of 1 and 1.5. A taxi will wait no matter how many other taxis are in line, but an arriving passenger waits only if the number of passengers already waiting for taxis is two or less.

(a) Define the states, draw the transition diagram, and mark the rates. For clear understanding draw on a paper
(b) Write the balance equations and calculate the steady-state probabilities
(c) Find the mean number of taxis waiting for passengers.
(d) Find the mean number of passengers waiting for taxis.
(e) Find the mean number of passengers who in the course of an hour do not join the waiting line because at least three passengers were already waiting.

The Correct Answer and Explanation is:

Let’s analyze the problem step by step.

Problem Restatement:

  • Taxis arrive: Poisson process, rate λT=1\lambda_T = 1 per minute.
  • Passengers arrive: Poisson process, rate λP=1.5\lambda_P = 1.5 per minute.
  • A taxi always waits no matter how many taxis are waiting.
  • A passenger will wait only if there are 2 or fewer passengers already waiting; otherwise, the passenger leaves immediately (does not join the queue).

(a) Define the states and draw the transition diagram

States Definition:

Let state (i,j)(i,j) represent the number of taxis and passengers waiting at the station, where:

  • i≥0i \geq 0 = number of taxis waiting
  • j=0,1,2,3j = 0,1,2,3 = number of passengers waiting, but capped at 3 because if there are already 3 passengers waiting, new arriving passengers will not join.

Why cap jj at 3?

  • Because the problem states passengers wait only if there are 2 or fewer passengers waiting. So when j=3j=3, passengers refuse to wait, and jj does not increase beyond 3.

Transition Types:

  • Taxi arrival: increases ii by 1 if passengers waiting is zero, or causes matching with passenger if passengers waiting is more than zero.
  • Passenger arrival: increases jj by 1 if j≤2j \leq 2, else passenger balks (does not join).
  • Match (service): when both i>0i > 0 and j>0j > 0, a taxi and a passenger leave the queue together, so i→i−1i \to i – 1, j→j−1j \to j – 1.

State transitions:

  • From (i,j)(i,j):
    • Taxi arrival (rate 1):
      • If j=0j = 0, (i,j)→(i+1,j)(i,j) \to (i+1, j) (taxi waits)
      • If j>0j > 0, a taxi arrives and immediately matches with one passenger: (i,j)→(i,j−1)(i,j) \to (i, j-1) because ii doesn’t increase (taxi arrives and immediately leaves with a passenger).
    • Passenger arrival (rate 1.5):
      • If j<3j < 3, (i,j)→(i,j+1)(i,j) \to (i, j+1) (passenger waits)
      • If j=3j=3, passenger balks, no change in state.
    • Matching (service):
      • Matching occurs instantaneously on arrival of a taxi or passenger when both queues are non-empty, so no separate matching rate is needed; matches happen immediately due to arrivals.

I will describe the transition diagram (you can sketch):

  • States are pairs (i,j)(i,j) with i≥0i \geq 0, j=0,1,2,3j=0,1,2,3.
  • Horizontal transitions: passenger arrivals j→j+1j \to j+1 if j<3j < 3 (rate 1.5).
  • Vertical transitions: taxi arrivals i→i+1i \to i+1 if j=0j=0 (rate 1).
  • When j>0j > 0, taxi arrivals cause j→j−1j \to j-1 (rate 1).
  • No transitions from j=3j=3 to j=4j=4 because passengers balk.

(b) Write balance equations and find steady-state probabilities

We model jj with maximum 3 passengers waiting; ii can be infinite.

Define Pi,jP_{i,j} = steady-state probability of ii taxis and jj passengers waiting.

The balance equations come from flow into and out of each state.

Key facts:

  • For j=0j=0: taxis accumulate.
  • For j>0j>0, taxi arrivals immediately reduce passengers by 1 (match).
  • Passengers arrive and increase jj up to 3.

Balance equations for j=0j=0:

For i≥0i \geq 0,

Inflow:

  • From state (i−1,0)(i-1,0) by taxi arrival: rate 1.
  • From state (i,1)(i,1) by taxi arrival matching a passenger (passenger count goes down): rate 1.

Outflow:

  • Passenger arrival increases passengers: (i,0)→(i,1)(i,0) \to (i,1), rate 1.5.
  • Taxi arrival: (i,0)→(i+1,0)(i,0) \to (i+1, 0), rate 1.

So, (λT+λP)Pi,0=λTPi−1,0+λTPi,1(\lambda_T + \lambda_P) P_{i,0} = \lambda_T P_{i-1,0} + \lambda_T P_{i,1}

Substitute λT=1\lambda_T=1, λP=1.5\lambda_P=1.5: (1+1.5)Pi,0=1×Pi−1,0+1×Pi,1(1 + 1.5) P_{i,0} = 1 \times P_{i-1,0} + 1 \times P_{i,1} 2.5Pi,0=Pi−1,0+Pi,12.5 P_{i,0} = P_{i-1,0} + P_{i,1}

Balance equations for j=1,2j=1,2:

For j=1,2j=1,2 and any i≥0i \geq 0,

Inflow:

  • From (i,j−1)(i, j-1) by passenger arrival: rate 1.5.
  • From (i+1,j)(i+1, j) by taxi arrival matching passenger: rate 1.
  • From (i,j+1)(i, j+1) by taxi arrival (matching decreases jj): rate 1.

Outflow:

  • Passenger arrival increases jj: (i,j)→(i,j+1)(i,j) \to (i,j+1), rate 1.5.
  • Taxi arrival decreases jj: (i,j)→(i,j−1)(i,j) \to (i,j-1), rate 1.

(λT+λP)Pi,j=λPPi,j−1+λTPi+1,j+λTPi,j+1(\lambda_T + \lambda_P) P_{i,j} = \lambda_P P_{i, j-1} + \lambda_T P_{i+1,j} + \lambda_T P_{i,j+1}

for j=1,2j=1,2, with Pi,3+1=0P_{i,3+1}=0.

For j=3j=3:

Passengers balk at j=3j=3, so no new passenger arrivals:

Inflow:

  • Passenger arrival from j=2j=2: rate 1.5.
  • Taxi arrival from j=3j=3 matching passenger: from (i+1,3)(i+1,3) by taxi arrival: rate 1.

Outflow:

  • Taxi arrival decreases jj: rate 1.
  • Passenger arrival balks: no change.

Balance: λTPi+1,3+λPPi,2=λTPi,3\lambda_T P_{i+1,3} + \lambda_P P_{i,2} = \lambda_T P_{i,3}

Solving these equations analytically is complex but can be approached by:

  • Recognizing the structure is a Markov chain with states (i,j)(i,j).
  • Using generating functions or matrix geometric methods.
  • The problem is a variant of a Quasi-Birth-Death (QBD) process with level ii and phase jj.

(c) Find mean number of taxis waiting

Mean number of taxis waiting: E[I]=∑i=0∞∑j=03i⋅Pi,jE[I] = \sum_{i=0}^\infty \sum_{j=0}^3 i \cdot P_{i,j}

(d) Find mean number of passengers waiting

Mean number of passengers waiting: E[J]=∑i=0∞∑j=03j⋅Pi,jE[J] = \sum_{i=0}^\infty \sum_{j=0}^3 j \cdot P_{i,j}

(e) Mean number of passengers who balk (do not join) in an hour

Passengers balk only when j=3j=3, arriving passengers balk at rate 1.5 passengers per minute.

Expected number of balking passengers per minute: 1.5×P(passenger queue=3)=1.5×∑i=0∞Pi,31.5 \times P(\text{passenger queue} = 3) = 1.5 \times \sum_{i=0}^\infty P_{i,3}

Per hour (60 minutes): Balking per hour=60×1.5×∑i=0∞Pi,3\text{Balking per hour} = 60 \times 1.5 \times \sum_{i=0}^\infty P_{i,3}

Explanation (about 300 words):

This problem models a taxi-passenger matching system using continuous-time Markov chains, where taxis and passengers arrive as independent Poisson processes. The key complexity lies in the differing queueing behavior: taxis always wait, but passengers refuse to wait if the passenger queue reaches 3 or more, causing balking.

States are represented as (i,j)(i,j), counting waiting taxis and passengers, with jj capped at 3 because passengers balk beyond this. Transitions correspond to taxi arrivals (increasing taxis or matching passengers), passenger arrivals (increasing passengers if j<3j < 3), and immediate matching of taxis and passengers when both are present.

The balance equations stem from equating inflow and outflow probabilities per state, reflecting arrival and matching rates. Because ii can be unbounded and jj capped, the system forms a quasi-birth-death process, solvable using matrix-geometric methods or generating functions. These methods yield the steady-state distribution Pi,jP_{i,j}, which enables computation of the average numbers of waiting taxis and passengers.

Passengers balk when the passenger queue is at its limit j=3j=3. The balking rate equals the passenger arrival rate multiplied by the probability the passenger queue is full, Pi,3P_{i,3}. Multiplying by 60 converts this rate to an hourly count.

The model captures the essential interaction between two queues with matching and balking, applicable to real-world ride-sharing or taxi stand scenarios. The constraints on passenger waiting introduce a nontrivial truncation in the passenger queue state space, significantly impacting queue dynamics and the probability of balking.

If you’d like, I can help compute approximate steady-state probabilities or provide a numerical example or code to solve the system. Would you like that?

Related Posts

Scroll to Top