2. MOTIVATION
■ According to the Guardian, an estimated 150000 people die unnecessarily
because emergency first-aid is not available on time.
■ Blocked airways cause 2500 deaths a year, while Sudden Cardiac Arrest
(SCA) kills more than 29000 people.
■ These deaths are avoidable if there are epinephrine syringes( Epi-Pen) and
AED (Automated External Defibrillator) available to the patient on time.
■ The time between a blocked airway and death is around 15 mins, while for
SCA it’s about 10 mins.
■ Average ambulance response times in NYC are around 8.5 mins, while in
other cities around the world it’s a lot more.
■ Can we get these emergency first-aid kits to the incident sooner?
4. PROPOSED SYSTEM
■ We decided to explore the idea of drones delivering these emergency life-
saving kits when not available nearby on time.
■ Drones will be stationed at 24-hour pharmacies across the city.
■ When an incident happens and a call to 911 is placed, check the kind of
emergency.
■ If it fits the system’s definition, grab the incident coordinates.
■ Package is processed and deployed within 30 seconds ( very much possible).
■ The drone takes off, flying at a safe height for the region, controlled by a
registered employee.
■ The drone reaches the point coordinates, descends, drops package.
■ Contains a smartphone with ER nurse on video call to guide through first-aid.
■ If serious, ambulance arrives to take to hospital. Otherwise, no ambulance
deployed.
5. DRONE SPECIFICATIONS
■ DJI Matrice 600 PRO
■ Chosen for high pay-load capacity (6kgs).
■ High speeds for ascent (5m/s), descent (3m/s) and flight speed (40 miles/hr)
■ Battery Life: 16-18 minutes on full load.
■ Dimensions: 1668 mm × 1518 mm × 727 mm with propellers, frame arms and GPS
mount unfolded (including landing gear)
■ Controller range: 5 miles.
6. TEST PILOT: MANHATTAN
■ Why Manhattan?
– Response times here are very fast compared to other major cities.
– Manhattan Skyline is one of the tallest, most complex skylines in the
world, that is hard for drones routing.
■ Manhattan might be the biggest challenge for proposed system, against
benchmark.
Let’s take a flyover through Manhattan to get an understanding of the skyline.
The next few slides take us over an aerial trip through Manhattan, very much
like a drone.
12. ZONING
■ From this, it’s visually obvious that there are zones in Manhattan with
different max heights.
■ The tallest zone is the Downtown Manhattan zone, with the One World
Trade Center at 541 m.
■ The next tallest building is a residential building at 421 meters.
■ At what height should the drone fly to not worry about the height of the
skyline?
■ For this system, we assumed the worst case scenario in terms of height, and
then used that as a benchmark against the existing fast response times for
ambulances.
13. The System
For the worst case, we’re considering that the drone flies at a height of 500m. This is done to not only avoid the tall
skyscrapers of NYC, but also to then have the option to choose Euclidean distance as our shortest path to the
incident location. Then the time of flight for the drone becomes:
Ttotal = Tascent + Tdescent + Tflight
The time to beat in the worst case scenario is 7 mins, which is faster than the median response times for FDNY in
emergency medical scenarios.
We can predetermine the Tascent and the Tdescent for the worst case scenario through the drone constraints. Given
the Ascent speed, Sascent =5m/s, Sdescent =3m/s, we get :
Tascent = 100 seconds = ;
Tdescent = 167 seconds ;
This leaves us with Tflight = 420 - 267= 253 seconds. In that time, the drone, with 40 miles/hour, can cover:
Distance = 2.81 miles = 4.52 km
And with that distance, one 24 hr pharmacy can service a region with an area of 24.83 sq miles.
Manhattan’s area is 22.82 sq miles.
15. Example of routing
● Here’s a zone in Manhattan, with max height
=300m .
● We add nodes to the border of the zone to
measure max service time in the zone.
● With that height, the ascent + descent times
reduce to: 60s(ascent) + 100s (descent) =
160s.
● Max distance in this zone between a facility
and a node is from point A to facility 1, which is
1.52 miles.
● Response time becomes: service time (30
seconds) + 160 s (ascent+ descent) + 136.8
seconds (flight time) = 326.8 seconds = 5.45
mins.
● Shortest distance of point A to a facility in the
zone is to facility 2, with distance = 1.2 miles.
● Response time for 1.2 miles: 4 mins, 28
seconds.
16. Facility Selection Problem
● Now that we have zones and heights for each zone, the problem then becomes of selecting which facilities to
use as drone deployment facilities.
● The cost of one drone is $4000, which can be considered as facility cost. Not included in objective function to
select facility.
● Each zone has the number of facilities such that the response time in that zone doesn’t exceed 7 mins,
regardless of max height in zone.
● The edge cases for each zone are considered to check for response time.
● Since there are more 24 hour facilities than we need to efficiently solve this problem, for each zone, we select
the facility with fastest response time for that zone. And then the second best one to account for accidental
failures or redundancies.
● Therefore, to cover Manhattan under a response time faster than 7 mins in case of medical emergencies, we use
6 facilities to cover the area. That is double the number of facilities needed, but accounts for backup.
● Objective function: Minimize time to serve each edge node.
● Constraints: Response Time <7 mins, Height of Zone <500m.
17. The Process, summarized
The process, on a high level, works as follows:
1. Incident occurs, call placed to 911.
2. Incident location determined, transmitted to system. Zone determined.
3. System calculates shortest distance to each drone depot pharmacy in zone, the one
being closest selected.
4. Required emergency first-aid kit loaded into drone
5. Drone dispatched
6. Drone arrives at incident location, along with a paramedic on video call on the
onboard smartphone to guide through first-aid, if required.
18. FUTURE IMPROVEMENTS
■ For this problem, we’re taking manually controlled drones. Autonomous drones
might achieve better results with automated height response.
■ We are also assuming that most of the time these drones will be unused, since these
are to handle cases of emergency.
■ We do not consider the revenue at this time, comparing the cost of benchmark and
proposed system might improve the system.
■ We do not consider real-time climate conditions at the time of incident. In that case,
we would calculate real-time flight speed of drone, calculate real-time coverage
radius of incident, and choose facility closest, regardless of zone.