Optimizing Medicine &
Vaccine Delivery Routes in
New York City
Group members: Yuan Zou / Hao Wang
Research Background &
Objectives
P A R T 0 1
3
● Medicines/vaccines are highly temperature- and time-
sensitive: Cold chain violations directly lead to drug
spoilage and public health risks.
● Current Pain Points: Under NYC’s high population
density and complex traffic, existing fixed/experience-
driven routes suffer from low efficiency, high costs,
and unstable compliance rates.
1. Reduce total costs (vehicle activation, mileage, cold chain
energy consumption).
2. Improve cold chain compliance rates (reduce violations
and drug loss).
3. Increase on-time delivery rate (target: 95%).
≥
4. Provide actionable scheduling policy recommendations for
DOHMH.
Core Challenges Why routing realism matters
P A R T 0 1
Waiting-Time Paradox (Why delays feel
worse than averages)
1. Travel time is variable in NYC; we don’t “experience” the
mean travel time.
2. We are more likely to be caught in longer congestion
periods (inspection/waiting-time paradox).
3. Motivation: use peak/off-peak travel-time matrices +
service-level metrics (on-time / cold-chain risk), not just
average time.
Benchmark Service System
P A R T 0 2
5
Core Components
● Depot: NYC Public Health Laboratory (455 1st Ave, Manhattan), operating hours: 7:00–19:00.
● Receiving Sites (29): Gotham Health Community Health Centers, covering NYC’s 5 boroughs
(standardized addresses, coordinates, and opening hours).
● Current Operation Mode: Fixed/experience-driven scheduling (non-optimal).
● Key Stakeholders: DOHMH, PHL, 29 Health Centers, Patient Groups.
Benchmark Operational Indicators
● Core Metrics: Number of vehicles, total mileage, average arrival time, cold chain compliance
rate.
P A R T 0 2
Core Parameterization
P A R T 0 3
7
P A R T 0 3
8
P A R T 0 3
9
Optimization Model &
Methods
P A R T 0 4
11
P A R T 0 4
Objective Function (Minimize Total Cost)
12
P A R T 0 4
Constraints
13
P A R T 0 4
Constraints
Data Collection
P A R T 0 5
15
Data Collection
● Real NYC case: 1 depot + 29 Gotham Health clinics
● Collected five types of data:
● Facility locations (address, coordinates, borough)
● Clinic opening hours time windows
→
● Population & demand estimates
● Distance / travel time between all nodes
● Vehicle & cold-chain parameters
P A R T 0 5
16
17
18
5.1 Data Sources
● NYC Health + Hospitals / Gotham Health websites
● clinic addresses, boroughs, opening hours
● NYC population / demographic data
to estimate population served by each clinic
● Public health guidelines (CDC / DOHMH)
vaccine usage rates, cold-chain limits
● Map / routing services
coordinates, distances, travel-time estimates
● Vehicle specification sheets
capacity, fuel cost, refrigeration properties
P A R T 0 5
19
5.2 Preprocessing &
Assumptions
P A R T 0 5
Results
P A R T 0 6
21
Scenario-Level Performance
Comparison
P A R T 0 6
Scenario Avg. Total Distance (km) Avg. Total Duration (hr) Avg. Total Load (kg)
Peak 81.79 12.95 70.84
Off-Peak 83.64 11.05 70.84
22
Peak vs. Off-Peak
● Duration: Off-peak average route duration (11.05 hrs) is 17% shorter
than peak (12.95 hrs) due to reduced congestion.
● Distance: Off-peak average distance (83.64 km) is slightly higher
(+2.3%) than peak (81.79 km), reflecting optimized route coverage.
● Load: Identical average load (70.84 kg) across scenarios—demand
consistency confirms efficiency differences stem from traffic, not
delivery volume.
23
Vehicle-
Level Total
Distance &
Load
Distribution
P A R T 0 6
Vehicle
(Scenario)
Total Distance
(km)
Total Load (kg)
Peak 1 140.85 102.36
Peak 3 15.13 9.11
Peak 4 89.40 101.04
Off-Peak 0 35.53 33.10
Off-Peak 2 84.54 91.69
Off-Peak 3 130.86 87.72
24
Clinic Demand Distribution
P A R T 0 6
25
EfficiencyHighlights
● Load balance: Off-peak routes distribute load more evenly (e.g., 91.69 kg vs. 87.72 kg for
top vehicles) vs. peak’s unbalanced allocation (2 vehicles handle 98% of peak load).
● Geographic alignment: 70% of demand is in Brooklyn (42%) and Manhattan (28%)—model
prioritizes these areas, cutting unnecessary long trips (e.g., Staten Island: 2% demand,
minimal routes).
26
VRP Map
P A R T 0 6
27
VRP Map
P A R T 0 6
28
Thanks!

Optimizing Medicine & Vaccine Delivery Routes in New York City.pptx

  • 1.
    Optimizing Medicine & VaccineDelivery Routes in New York City Group members: Yuan Zou / Hao Wang
  • 2.
  • 3.
    3 ● Medicines/vaccines arehighly temperature- and time- sensitive: Cold chain violations directly lead to drug spoilage and public health risks. ● Current Pain Points: Under NYC’s high population density and complex traffic, existing fixed/experience- driven routes suffer from low efficiency, high costs, and unstable compliance rates. 1. Reduce total costs (vehicle activation, mileage, cold chain energy consumption). 2. Improve cold chain compliance rates (reduce violations and drug loss). 3. Increase on-time delivery rate (target: 95%). ≥ 4. Provide actionable scheduling policy recommendations for DOHMH. Core Challenges Why routing realism matters P A R T 0 1 Waiting-Time Paradox (Why delays feel worse than averages) 1. Travel time is variable in NYC; we don’t “experience” the mean travel time. 2. We are more likely to be caught in longer congestion periods (inspection/waiting-time paradox). 3. Motivation: use peak/off-peak travel-time matrices + service-level metrics (on-time / cold-chain risk), not just average time.
  • 4.
  • 5.
    5 Core Components ● Depot:NYC Public Health Laboratory (455 1st Ave, Manhattan), operating hours: 7:00–19:00. ● Receiving Sites (29): Gotham Health Community Health Centers, covering NYC’s 5 boroughs (standardized addresses, coordinates, and opening hours). ● Current Operation Mode: Fixed/experience-driven scheduling (non-optimal). ● Key Stakeholders: DOHMH, PHL, 29 Health Centers, Patient Groups. Benchmark Operational Indicators ● Core Metrics: Number of vehicles, total mileage, average arrival time, cold chain compliance rate. P A R T 0 2
  • 6.
  • 7.
    7 P A RT 0 3
  • 8.
    8 P A RT 0 3
  • 9.
  • 10.
  • 11.
    11 P A RT 0 4 Objective Function (Minimize Total Cost)
  • 12.
    12 P A RT 0 4 Constraints
  • 13.
    13 P A RT 0 4 Constraints
  • 14.
  • 15.
    15 Data Collection ● RealNYC case: 1 depot + 29 Gotham Health clinics ● Collected five types of data: ● Facility locations (address, coordinates, borough) ● Clinic opening hours time windows → ● Population & demand estimates ● Distance / travel time between all nodes ● Vehicle & cold-chain parameters P A R T 0 5
  • 16.
  • 17.
  • 18.
    18 5.1 Data Sources ●NYC Health + Hospitals / Gotham Health websites ● clinic addresses, boroughs, opening hours ● NYC population / demographic data to estimate population served by each clinic ● Public health guidelines (CDC / DOHMH) vaccine usage rates, cold-chain limits ● Map / routing services coordinates, distances, travel-time estimates ● Vehicle specification sheets capacity, fuel cost, refrigeration properties P A R T 0 5
  • 19.
  • 20.
  • 21.
    21 Scenario-Level Performance Comparison P AR T 0 6 Scenario Avg. Total Distance (km) Avg. Total Duration (hr) Avg. Total Load (kg) Peak 81.79 12.95 70.84 Off-Peak 83.64 11.05 70.84
  • 22.
    22 Peak vs. Off-Peak ●Duration: Off-peak average route duration (11.05 hrs) is 17% shorter than peak (12.95 hrs) due to reduced congestion. ● Distance: Off-peak average distance (83.64 km) is slightly higher (+2.3%) than peak (81.79 km), reflecting optimized route coverage. ● Load: Identical average load (70.84 kg) across scenarios—demand consistency confirms efficiency differences stem from traffic, not delivery volume.
  • 23.
    23 Vehicle- Level Total Distance & Load Distribution PA R T 0 6 Vehicle (Scenario) Total Distance (km) Total Load (kg) Peak 1 140.85 102.36 Peak 3 15.13 9.11 Peak 4 89.40 101.04 Off-Peak 0 35.53 33.10 Off-Peak 2 84.54 91.69 Off-Peak 3 130.86 87.72
  • 24.
  • 25.
    25 EfficiencyHighlights ● Load balance:Off-peak routes distribute load more evenly (e.g., 91.69 kg vs. 87.72 kg for top vehicles) vs. peak’s unbalanced allocation (2 vehicles handle 98% of peak load). ● Geographic alignment: 70% of demand is in Brooklyn (42%) and Manhattan (28%)—model prioritizes these areas, cutting unnecessary long trips (e.g., Staten Island: 2% demand, minimal routes).
  • 26.
  • 27.
  • 28.