Unmanned Aircraft for Emergency Deliveries Between Hospitals in Madrid: Estimating Time Savings and Predictability

Highlights

What are the main findings?

• Drone deliveries between hospitals in Madrid achieved time savings of 2–26 min (35–58%) compared to road transport.
• Drones provided reliable delivery within 15 min, outperforming cars and ambulances during peak traffic conditions.

What are the implications of the main findings?

• Demonstrates drones’ potential to improve predictability and reliability in emergency healthcare logistics.
• Provides evidence to support U-space integration policies for medical deliveries in dense urban environments.

Abstract

Unmanned aircraft are increasingly recognized for their potential to enhance healthcare logistics, offering rapid and reliable transport solutions. Among the many envisioned use cases, emergency medical deliveries stand out as particularly promising due to their immediate societal value. This study investigates the potential of drones operating under U-space to support hospital-to-hospital emergency deliveries in Madrid. Using the GEMMA tool, we modeled and simulated operations with two drone types along direct routes between four hospitals, resulting in six hospital pairs. Drone travel times were estimated and compared against road transport times obtained from the Google Routes API, incorporating one week of traffic data to capture daily and weekend variability. The results show substantial advantages of aerial transport, with time savings ranging from 2 to 26 min, equivalent to 35–58% compared to road transport. Drones consistently ensured deliveries within 15 min, outperforming regular cars (39%) and ambulances or motorcycles in highly congested periods. Sensitivity analysis confirms their reliability in scenarios with strict time constraints, especially under 15 min. These findings demonstrate that drones reduce travel times and improve predictability, providing a robust evidence base for policymakers and regulators to advance U-space integration in healthcare logistics.

1. Introduction

The fast advancement of urban air mobility (UAM) is changing the landscape of urban transportation systems with a new type of mobility. The development and deployment of a new digital and regulatory system in Europe (the so-called U-space) is a key enabler for the emerging UAM ecosystem, offering the infrastructure and digital services required to safely manage increasing volumes of unmanned aircraft system (UAS) traffic in shared airspace [1]. Designed to unlock the potential of the drone economy, U-space is expected to support a diverse range of UAM use cases [2]. Over the coming decade, the gradual implementation of UAM is expected not only to accelerate the adoption of advanced aerial technologies but also to generate high-quality employment and stimulate innovation across the aviation and logistics sectors. This integration into populated environments presents both opportunities and challenges. Despite the technological progress and increasing number of pilot projects, the societal impact of UAM remains only partially understood. Issues related to public acceptance, noise pollution, visual intrusion, access and equity, and environmental trade-offs are still under investigation. Comprehensive studies, such as the one conducted by the European Union Aviation Safety Agency [3], highlight that public perception, transparency, and trust will play a pivotal role in shaping the trajectory of UAM implementation. In particular, applications with direct societal impact, such as healthcare logistics, may serve as key demonstrators to evaluate not only technical feasibility but also broader urban integration, given their more tangible benefits for citizens.

Among the many use cases envisioned for UAM, medical deliveries stand out as particularly promising, offering clear and immediate societal value. This use case may serve as a stepping stone for broader UAM adoption, fostering public trust and informing future regulatory frameworks. Along with these promising opportunities, there are also several challenges. Regulatory restrictions for beyond visual line of sight (BVLOS) flights in dense urban areas, the need for suitable infrastructure, the dependence on weather conditions, and uncertainties about scalability beyond pilot projects currently limit the large-scale deployment of UAS-based emergency deliveries.

This paper focuses on healthcare logistics, specifically the transport of medical supplies, blood, pharmaceuticals, or diagnostic samples between hospitals, laboratories, and clinics employing drones. These services can yield substantial time savings and improved predictability in healthcare transportation, which are agreed to be the most important factors for this service [4]. Moreover, UAM medical transport can increase resilience during infrastructure disruptions or emergencies.

This study contributes to the growing field of drone-based healthcare logistics by focusing on the urban context of Madrid. By applying real-world spatial data and health infrastructure, we assess the operational feasibility of emergency deliveries in a concrete, policy-relevant environment.

We employ the GEMMA tool to simulate realistic drone trajectories. GEMMA enables the modeling of different configurations of the airspace, terrain-aware routing, and various operator profiles, allowing for high-fidelity and scalable simulations.

In addition, we conduct a sensitivity analysis comparing drones, motorcycles, and ambulances under different maximum delivery times (5, 10, 15, and 20 min). This allows for a comparative evaluation of which transport modes can meet critical time thresholds with the highest reliability.

By focusing on this use case, we aim not only to assess its operational performance but also to demonstrate how such targeted studies can serve as input for developing robust assessment frameworks. This research contributes to a more comprehensive understanding of UAM’s role in urban environments, offering insights into how these systems might be designed, governed, and accepted in ways that prioritize public value and environmental sustainability.

Our findings are intended to support informed regulatory decisions, ensuring that UAM implementation is guided by empirical evidence and aligned with societal values. In this context, there is a pressing need for more advanced indicators and measurement mechanisms that can capture the multifaceted impacts of UAM on urban systems. These tools are essential to inform the design of future regulatory frameworks and ensure that U-space services contribute positively to the sustainability and livability of European cities. Without evidence-based tools to monitor outcomes, decision makers risk deploying UAM solutions that may inadvertently compromise social or environmental goals.

The research presented here was performed in the context of the MUSE SESAR project [5], which sets the foundation for a new U-space service that aims at reducing the negative environmental and social footprint of drones, proposing a new performance framework able to measure noise, visual pollution, privacy concerns, and economic and social benefits.

The structure of this paper is as follows. Section 2 presents a review of recent academic studies, project deliverables, and other relevant sources on the use of drones and eVTOL aircraft in emergency healthcare, with a particular focus on studies quantifying the time-saving potential of drone-based deliveries compared to conventional ground transport. The methodology used in this research is detailed in Section 3, including the definition of the case study scenario set in Madrid and the calculation of travel times using both drone trajectory modeling and road traffic data. The results and related discussion are provided in Section 4, highlighting the key findings regarding time efficiency and implementation challenges. Section 5 concludes the paper and outlines possible directions for future research.

2. Literature Review

The emerging field of UAM, encompassing drones and eVTOL aircraft, is gaining increasing attention for its potential to revolutionize emergency healthcare logistics and significantly reduce response times.

Early conceptual work by ACI [6] emphasized UAM’s ability to bypass urban congestion and deliver life-saving interventions rapidly, underlining the importance of regulatory alignment, public acceptance, and infrastructural frameworks. Subsequent systematic reviews have explored applications of drones in medical contexts [7,8,9,10,11,12,13,14,15,16].

Nyaaba and Ayamga [7], for example, focused on the African landscape, identifying critical enablers and challenges for drone-based delivery of blood, medications, and lab samples in underserved regions. Similarly, Roberts et al. [8] and De Silvestri et al. [9] synthesized evidence on the use of drones in air ambulance services and the delivery of time-sensitive supplies like automatic external defibrillators (AEDs), naloxone, antiepileptics, and blood products. By applying topic modeling on 290 publications, Campbell et al. [11] distinguished between time-critical deliveries (e.g., AEDs, organs) and routine logistical tasks. Jakobsen et al. [12] analyzed 39 studies on drone-delivered AEDs for out-of-hospital cardiac arrest (OHCA) and reported real-world time savings of up to three minutes. Similarly, Habibi et al. [13] emphasized the effectiveness of drones in improving emergency response times, especially in remote areas. Jazairy et al. [14] synthesized 307 articles and proposed a framework for integrating drones into last-mile medical logistics. A comprehensive review of 36 studies illustrated the diverse roles of drones, from pharmaceutical and organ delivery to disaster response and environmental monitoring [15]. Finally, Kuljanin et al. [16] focused on medical drones in urban environments. Despite operational constraints, such as building density, signal interference, and the absence of standardized emergency landing zones, drones are increasingly recognized as valuable tools for rapid access to congested or remote areas.

Most authors highlight persistent regulatory and societal barriers [6,9,10,11,13,14], stressing the need for integration into emergency medical services (EMS) systems [8] or in public health systems [15].

Medical drone applications can typically be classified into two categories, reflecting the urgency and purpose of the delivery. The first involves routine healthcare deliveries—such as transporting blood [17], vaccines [18], medications, and diagnostic samples between clinics and labs. The second focuses on emergency responses, including disaster relief and on-scene delivery of first-aid kits when ground access is limited.

Over the last decade, drones have demonstrated their prominent role in disaster response and management, including during the earthquakes in Haiti and Taiwan, the typhoon in the Philippines, and the recurring floods in Nepal [19,20]. Notable progress has been made in Africa, where national drone delivery systems have been implemented. In Rwanda, where 83% of the population lives in rural areas, drones helped reduce blood wastage and addressed logistical delays [21], while Ghana adopted Zipline’s system for vaccine distribution [22]. A study by Tetteh et al. [23] confirmed public support for drone operations in rural Ghana, although challenges such as payload limitations, staffing, and user acceptance were noted.

Private companies have also advanced medical drone logistics. The U.S: Matternet, based in California, has conducted test flights in Lesotho, Bhutan, the Dominican Republic, and Haiti [24], and more recently, hospital-to-hospital transfers in Zurich and Berlin [25,26], operating in urban environments under beyond visual line of sight (BVLOS) conditions. RigiTech, a Swiss company based in Prilly, demonstrated compliance with BVLOS SORA regulation with successful long-distance flights (e.g., over Lake Geneva) [27] in Uruguay [28], The Netherlands [29], the Maldives [30], and many other regions. Unlike Matternet’s rotary-wing design, RigiTech uses tilting-wing drones suited for longer-range inter-city logistics (up to 80 km in under 40 min). Both companies focus on safety, operational efficiency, and medicine integrity. In this context, Hii et al. [31] developed protocols to evaluate the safe flight conditions, duration, range, and the effect of drone failure on the medication and environment. RigiTech, for example, recommends temperature-controlled transport boxes to preserve the stability of sensitive supplies [32].

Several field studies have demonstrated the feasibility and acceptance of drones in emergency medical scenarios. Pulver et al. [33] showed that AED-equipped drones in Salt Lake County could reach over 90% of out-of-hospital cardiac arrest (OHCA) cases within one minute, compared to only 4.3% with traditional EMS. Positive user experiences during simulated OHCA events involving drone-delivered AEDs are also documented in [34,35,36]. A feasibility study conducted by Ornato et al. [37] showed that drones are a viable means of delivering naloxone in U.S. regions with historically long emergency response times. Baumgarten et al. [38] evaluated the integration of drone-delivered AEDs with smartphone-dispatched first responders, finding the approach both safe and feasible. On a broader scale, the AiRMOUR project [39] validated real-world UAM use for EMS in cities, such as Stavanger, Helsinki, and Nord Hessen, while acknowledging technical constraints like range and payload. Scholz et al. [40] evaluated nighttime drone operations, finding that all missions were safely and autonomously completed with performance comparable to daytime flights, thus supporting the feasibility of 24/7 emergency drone use.

Some studies highlight the environmental benefits of drone-based medical deliveries, including substantially lower emissions compared to conventional vehicles [41,42,43]. For example, Park et al. [41] found that drones emit one-sixth the global warming potential and half the particulate matter of motorcycles per kilometer, while in a study conducted in mountainous terrain, drones emitted just 0.09 g CO₂/km, far less than electric (3.43 g/km) and combustion vehicles (159.5 g/km) [43].

Numerous studies have quantified the time-saving potential of drones in emergency medical delivery compared to traditional ground transport, which is the main goal of this research [44,45,46,47,48,49,50]. Claesson et al. [44] used GIS modeling to show that drones could arrive before EMS in 93% of rural and 32% of urban OHCA cases, reducing intervention times by an average of 19 min. Similarly, Schierbeck et al. [45] reported that drones delivered AEDs before ambulances in 67% of cases, gaining a median advantage of 3 min and 14 s. An optimized deployment model in North Carolina, developed by Bogle et al. [46], showed that a 500-drone network could reduce median defibrillator arrival time from 7.7 to 2.7 min, enabling double survival rates (24.5% vs. 12.3%) with a low incremental cost, confirming its cost-effectiveness. A simulation study in rural Ontario demonstrated that AED-equipped drones consistently arrived 1.8 to 8 min faster than ambulances. For almost 13,000 drone deliveries conducted between 2017 and 2019 in Rwanda, the average delivery time was less than 50 min, compared to over 2 h delivery times by road transportation [48]. Johannessen [4] found that while drones provide limited time savings (under 30%) in short urban routes, they can significantly reduce transport time (up to 74%) over longer rural distances, especially when operated at high frequency and speed. However, cost-effectiveness is lower in rural areas due to smaller transport volumes. A systematic review by Al-Faridi et al. [49] showed that drones can reduce response times by an average of two minutes, arriving before ambulances in 64% of cases. In a real-world trial in Scotland, drones delivered lab specimens in just 35 min, dramatically reducing a road travel time that could take up to 5 h due to multiple required pick-ups [50].

In summary, the existing research highlights the strong potential of UAM to reduce emergency response times and improve healthcare access, especially in remote or congested areas. Numerous studies confirm drones’ ability to deliver life-saving supplies like AEDs significantly faster than traditional EMS while also offering environmental and cost-efficiency benefits. However, despite these benefits, challenges such as limited payload capacity, regulatory barriers, airspace integration, and the need for reliable infrastructure and trained personnel still constrain widespread implementation. Furthermore, public acceptance, weather sensitivity, and ensuring safe, autonomous operation, especially in dense urban or remote environments, remain key issues for scaling medical drone use effectively.

While many studies demonstrate the viability of medical drones in various global contexts, few have assessed their integration into routine inter-hospital logistics within dense European urban environments. This study aims to address that gap by evaluating the time efficiency and operational feasibility of UAVs for hospital-to-hospital deliveries in Madrid, offering practical insights for scaling drone-based logistics in high-density healthcare networks.

3. Methodology

The methodology applied in this research is designed to evaluate the potential benefits of unmanned aircraft for healthcare-related deliveries in Madrid, with a focus on time savings and predictability relative to road transport. It first defines the simulation scenarios by characterizing the parameters of emergency delivery flights, selecting the relevant hospital network, identifying suitable drone types based on technical specifications, and modeling drone trajectories and distribution of operations throughout the day. The second part of the methodology focuses on calculating delivery times, combining drone flight estimates obtained with the GEMMA tool and road transport estimates derived from the Google Routes API, thereby allowing a direct comparison of aerial and ground transport performance. Each step is detailed in the following subsections.

3.1. Scenario Definition

3.1.1. Characterization of Emergency Delivery

In the context of this research, medical sample transportation is envisioned to be performed as an emergency delivery service within an urban environment. Emergency deliveries are considered urgent flights that depart and arrive from/to a set of medical centers, with no restrictions on the time of the flight or on airspace access. Each emergency delivery is a one-way route. The drone takes off and lands vertically, and it flies following a direct path, with a cruise phase at a fixed altitude. The direct path ensures that emergency deliveries follow the shortest path. To avoid any conflict with other types of flights and with other emergency flights flying in the opposite direction, the U-space airspace reserves two layers for the emergency deliveries.

The heights of the routes depend on the path bearing between origin and destination: 105 m for the South/West routes and 120 m for the North/East ones. The ground reference for the height calculation is the hospital (origin or destination) with the highest elevation above sea level. Thus, each pair of hospitals and each direction has a different flight altitude, and it remains constant during the cruise.

3.1.2. Hospital Network Selection

We have simulated drone flights between four hospitals in Madrid. Hospital 12 de Octubre is located in the southern part of Madrid and is among the largest and most important public hospitals in the city. Two public hospitals, HU Carlos III and H La Paz, were selected because, as part of the U-ELCOME project [51], they were selected to perform drone delivery flights, although, finally, only the H La Paz flight was approved and validated. We also included CEP Pontones—as it is partially public—because CEP Pontones’ location offers more relevant route distances between hospital pairs for our study. Private hospitals are not included in this analysis, as they are more likely to operate separate services from public hospitals.

Figure 1 shows the route network resulting from connecting all the hospitals considered for this case study. The current scenario assumes direct flights between each of the four selected hospitals, which results in six distinct hospital pair routes, marked with an identification number from 1 to 6. Distances of each pair are also given. The routes range from as short as 1.1 km, as seen between La Paz and HU Carlos III, to as long as 11.6 km between Hospital 12 de Octubre and La Paz.

3.1.3. Drone Trajectory Generation

The generation of drone trajectories in this study was carried out with the GEMMA tool. The GEMMA tool is a drone trajectory generation application that is highly customizable. It aims to create scenarios of future demand for drone flights on a large scale. From the knowledge of its creators and their experience as drone operators and researchers, the tool can produce credible trajectories on top of several concepts and assumptions.

By default, the flights have no restriction—free route operations, other than staying in the designed 3D air volume. Flights can depart and land from/to anywhere and fly at the height decided by the operator. Free route flights will always follow the terrain surface in order to stay within the legal altitude limits and avoid obstacles. GEMMA allows us to define no-fly zones and constrained areas. No-fly zones can be applicable to all flights, for instance, when there is a physical obstacle to avoid, or apply only to commercial missions, e.g., when allowing to fly over a governmental area. In contrast, constrained areas define areas where flights are allowed but must follow specific routes instead of the free route concept. A constrained area is defined by a 2D perimeter and a network of routes. Departures and arrivals are also limited to designated vertiports. In addition, vertical movements are limited to the altitude layers defined by the network. The altitudes of the constrained airspace layers ensure that flights will avoid any obstacles of the terrain.

Flights follow the speeds and accelerations defined per drone model. Cruise speed is selected randomly between the drone performance margins to simulate variability between battery status or payload effects. The accelerated movement model uses the vertical and horizontal accelerations provided by the drone performance. The top of the ascent and the top of the descent are at the vertical of arrival and departure points, and at zero speed. Turns reduce the cruise speed as a function of the turn radius, with a maximum reduction of 50%. Weather has not been taken into account.

Several operator profiles—such as security, delivery, and emergency operators—are available, and each is associated with specific flight route characteristics. Emergency operators fly beyond visual line of sight (BVLOS), but from hospital to hospital, they only fly one way and have no time restrictions. For all types of operators, GEMMA generates routes stochastically based on predefined distributions, and it uses a list of input parameters.

The main input parameters are as follows:

• The flight area includes the horizontal and vertical limits, the set of non-flying zones, and an optional constrained airspace area.
• The drone operators include the type of operator (security, emergency, or delivery), their locations, and their fleet.
• The drones include the list of drones and their performances, mainly maximum range, max-min speeds, and accelerations.
• Human activity includes the working hours, the peak activity time, or the list of public events that attract many citizens.

In the case of emergency deliveries, the departure time is set according to available emergency intervention statistics for the city of Madrid (see Section 3.1.5); the altitude is fixed according to the rules presented in Section 3.1.1 and does not follow the terrain. The drone model, the departure point, and the destination are selected randomly from the list of input parameters. GEMMA generates a number of routes per drone after validating that the selected drone is available and has sufficient range to cover the flight distance and that the route has been de-conflicted with the no-fly zones or with other drones.

3.1.4. Drone Types

We assume that emergency deliveries use a fleet of drones composed of 2 types of vehicles with similar payload capacity: DJI Matrice 600 (DJI, Shenzhen, China) and RigiTech Eiger (Rigi Technologies SA, Prilly, Switzerland). DJI Matrice 600 was available to one of the partners, which facilitated some tasks due to MUSE. This type of vehicle is currently out of stock, but DJI has very similar multi-copter models in the market with a large commercial share. RigiTech’s Eiger is a tilt-rotor drone able to fly long distances at high speeds with low power consumption. It is currently being used for emergency deliveries in Lake Geneva (Switzerland) and Bourgoin-Jallieu (France) and has successfully performed BVLOS emergency delivery demonstration flights worldwide (Kos, Greece; Busan, Republic of Korea; Maubeuge, France; Olot, Spain; Tacuarembó, Uruguay; etc.) [52].

Table 1. Drone types used in the emergency delivery simulations.

Characteristic	DJI Matrice 600	RigiTech Eiger
Maximum Take-Off Weight	15.5 kg	21 kg
Maximum Payload	2.7 kg	3 kg
Cruise Velocity (min–max)	12–18 m/s	23–35 m/s

provides specifications for the two drone types, including the maximum take-off weight, maximum payload, and minimum and maximum cruise velocity used in the simulation. The distances between hospitals in the Madrid case study allow both drone types to operate at their maximum payloads while remaining within their respective operational specifications. For two hospital pairs, the distance exceeds the maximum range of the DJI Matrice 600, indicating that this drone type cannot be operated on those particular routes.

3.1.5. Distribution of Operations

One representative “typical” week of traffic data (see Section 3.2.2) was analyzed to account for differences in road traffic congestion between weekdays and weekends.

Since data regarding the exchange of medical goods between specific hospital pairs in Madrid is not publicly available—especially with respect to future drone-based transportation—we based our analysis on available emergency intervention statistics for the city of Madrid in 2024. This dataset includes the details of 151,544 activations carried out by SAMUR-Civil Protection in 2024 [53]. Activation is defined as healthcare assignments that involve the activation of a healthcare resource or other type of vehicles [53], and the data file provides the year, month, time of request and intervention, incident code, district, and public hospital of transfer for each activation. We used 139,430 activations for which both the time of request and intervention were available.

Using these annual data, we calculated the hourly distribution of interventions (% of total activations per hour throughout the day), which is shown in Figure 2 with a green line. The peak number of operations occurred around 1:00 p.m.

For the purpose of the case study traffic, we assumed a maximum of two drone flights per hour between each hospital pair. Considering six hospital pairs, this results in a maximum of 12 flights per hour. We set this maximum at 1:00 p.m., corresponding to the observed peak in emergency operations. Using the hourly distribution percentages and the assumed maximum of 12 flights per hour, we estimated the number of drone operations per hour. The total number of estimated drone flights per day amounts to 180. The number of flights per drone type for each hour of the day, as well as for each hospital pair, is determined randomly based on GEMMA-simulated flight data.

These estimated values are presented in Figure 2. The x-axis represents the hours of the day (0 to 23), while the y-axis shows the number of flights conducted during each hour, categorized by each drone model. The distribution of drone operations varies across the day, with both drone models showing activity at all hours. DJI Matrice 600 sees its peak at 3:00 p.m. with 7 flights, while RigiTech Eiger peaks at 7:00 p.m. with 10 flights. On average, both drones operate around 3.6, i.e., 3.9 times per hour, respectively. On average, there are 7.5 operations per hour, leading to a total of 180 operations over the 24 h period.

Table 2. Drone flight operations statistics.

Characteristic	DJI Matrice 600	RigiTech Eiger	Total
Total flights	87	93	180
Mean distance (m)	5808	8256	7073
Mean time (min)	8.2	7.5	7.8
Mean altitude (m)	113.6	114.2	113.9
Peak hour traffic	7	10	12

analyzes the distribution of drone operations, highlighting the variations in distance, time, altitude, and peak activity of each drone type and the total activity in the region over the studied period.

It can be observed that the mean altitude is larger than the designed altitudes (105 m and 120 m). The reason is that the emergency flight altitudes were defined as meters above the highest hospital of the origin–destination pair. Since emergency flights do not follow the terrain, the gap between the stable cruise flight and the terrain is larger than the designed altitudes in most of the route.

3.2. Delivery Time Calculation

The methodology involves calculating the time difference by subtracting the drone delivery time from the road transport time, using the GEMMA tool for estimating drone flight time and the Google Routes API [54] for road transport estimates. Time savings are quantified both in absolute terms, expressed in minutes, and in relative terms, as percentage reductions in transport time compared to road delivery.

3.2.1. Drone Delivery Time

Drone delivery time depends on the distance between hospitals, the type of drone selected, the cruise altitude, and the selected cruise speed. It also depends on the non-flying zones found within the path. The time of the day does not affect the time in any way, but the operations density can affect the flight time. Two de-confliction methods are available, both at the strategic level: ground delay and change in cruise altitude.

Using the trajectories generated with the GEMMA tool, the total delivery time was estimated as the sum of the main flight phases: take-off, climb, cruise at the assigned altitude, descent, and landing. Each phase was parameterized according to the technical specifications of the selected drone models, including cruise velocity and rates of climb and descent. To ensure that the estimated delivery time reflects not only the ideal flight trajectory but also a realistic representation of flight dynamics, including operational and environmental elements, such as wind conditions and other stochastic factors, the cruise velocity for each flight was randomly selected within the minimum and maximum values provided in Table 1. Ground-handling components, such as vehicle preparation before take-off and package release or handover upon landing, were excluded from the calculations, as the focus was placed on the in-flight segment of the delivery.

3.2.2. Road Transport Delivery Time

Road transport delivery time depends on the type of vehicle used for the operation. Regular cars are often considered as a reference mode in comparative studies, while ambulances benefit from priority in traffic and, therefore, achieve shorter travel times. In addition, some cities, such as Paris, employ motorcycles for urgent medical deliveries, as they can more easily avoid congestion and reach their destination faster.

In this research, road transport delivery time for regular cars was estimated using the Google Routes API [54]. The Google Routes API is a service provided by Google Maps Platform that enables developers to calculate optimal routes between locations, considering factors like real-time traffic conditions, tolls, and various travel modes. It uses historical data to predict traffic patterns; however, past departure time cannot be specified to retrieve historical travel time estimates.

The Google Routes API currently does not offer dedicated travel modes for motorcycles or ambulances. It supports “drive”, “walk”, “bicycling”, and “transit” modes, but there are no specific parameters to simulate the privileges that ambulances might have (such as using sirens and emergency lanes) or the agility of motorcycles.

As stated previously, the emergency intervention statistics for the city of Madrid were available through the SAMUR-Civil Protection dataset [53]. According to official data published by the Madrid City Council (Portal web del Ayuntamiento de Madrid [55]), the average ambulance response time in 2024 was 9 min and 19 s, measured from the moment the call is received until arrival at the scene. Since our study focuses on hospital-to-hospital transport rather than emergency response to incident sites, and since these empirical ambulance response data in Madrid lack the information about distance, we were not able to explicitly use the available data.

Therefore, we decided to rely on Branas’s [56] results. To calculate ground ambulance driving times, Branas [56] used average ground ambulance speeds of 20.1 mph (32.3 km/h) for urban areas, 47.5 mph (76.4 km/h) for suburban areas, and 56.4 mph (90.8 km/h) for rural settings. In the urban context of Madrid, it is reasonable to assume that regular cars typically travel at speeds of around 20–25 km/h during peak traffic periods, given the common levels of congestion in large metropolitan areas. This indicates that ambulances, operating at an average of 32.3 km/h in urban areas, could potentially achieve travel time reductions ranging from approximately 23% (if cars average 25 km/h) to 38% (if cars average 20 km/h) compared to regular car travel. Therefore, to approximate the anticipated travel times of motorcycles and ambulances in Madrid, we applied a reduction of 10–20% for motorcycles and 30–40% for ambulances relative to the standard car travel times obtained from the Google Routes API.

Road transport delivery times were obtained by sending structured queries to the Google Maps Routes API. Each query required the geographic coordinates of the origin and destination (hospital locations), the travel mode (set to “DRIVE”), and the planned departure time in ISO format. Routing preferences allowed highways, toll roads, and ferries, while the routing option was set to TRAFFIC_AWARE in order to incorporate expected congestion conditions. In return, the API provided the estimated travel duration in seconds, the route distance in meters, and an encoded polyline representing the road path. To capture congestion variability, multiple queries were performed across different times of the day and days of the week. For each hospital pair, two directions were considered: forward (from Hospital A to Hospital B) and backward (from Hospital B to Hospital A). Forward directions were queried at every full hour (e.g., 08:00, 09:00, 10:00, etc.), while backward directions were queried at every half-hour (e.g., 08:30, 09:30, 10:30, etc.). The resulting travel times were treated as representative approximations of the average conditions within the corresponding one-hour intervals.

The process was fully automated in Python (version 3.11.13), which enabled systematic requests to the API, storage of the responses, and extraction of the relevant road travel times for subsequent comparison with drone delivery durations.

As with the drone calculations, ground-handling components such as vehicle preparation, loading, and package handover upon arrival were excluded, ensuring that the comparison focuses on the in-transit segment of the delivery.

In our initial design, we retrieved Google Routes API data for four “typical” weeks (one in each season of 2026) as well as for a set of “special” days corresponding to major holidays and events in Madrid. However, detailed inspection revealed that the API returns essentially identical or near-identical travel times for the same day of the week, regardless of the week or season. The differences were marginal—on the order of a few seconds rather than minutes—and thus not meaningful for the type of analysis presented in this study. For this reason, we streamlined our dataset and focused on one representative “typical” week (4–10 May 2026), which captures both weekday and weekend variability.

4. Results and Discussion

Based on the results obtained by the proposed methodology, various comparative analyses were performed and presented here. The results of estimated time savings are derived from a comparison between drone delivery times and those of traditional car transport, while the assessment of predictability extends the comparison to also include ambulance and motorcycle travel times. It is worth noting that ground-handling times (e.g., preparation, loading/unloading) were not included in this analysis, neither for the drones nor for the ground transportation, due to the lack of reliable operational data; however, these processes are expected to be more automated and faster for drones, with less dependence on human resources than in road-based transport. As such, including these times in future studies could further reinforce the benefits of drone-based emergency deliveries.

Figure 3 illustrates the average absolute and relative reduction in travel time for healthcare-related deliveries by drones (compared to traditional car transport travel time) across a 24 h period. The blue line represents the average absolute reduction in travel time, which varies from approximately 4.4 min to 11.7 min throughout the day for both drone models. The red line represents the average relative reduction as a percentage of the total travel time, fluctuating between 35% and 58%. The average absolute travel time reductions are most significant in the morning and evening hours, peaking at 11.7 min at 6:00 p.m., with a 53.2% relative reduction. The average relative reduction peaks at 57.9% around 7:00 p.m., where the absolute reduction reaches 10.2 min.

To showcase the impact of traffic conditions and day of week on the effectiveness of drones for healthcare-related deliveries, one representative week of traffic data (from Monday to Sunday) is analyzed, and the results are presented in Figure 4. It presents the average reduced travel time (in minutes) using dotted blue and red lines for DJI Matrice and RigiTech, respectively, based on the average cruise velocity of each drone. The minimum–maximum time reduction ranges, shown as shaded areas, are derived from the minimum and maximum cruise velocities to represent the lower and upper bounds of potential time savings. The data is segmented by hospital pair, flight direction (B—backward, F—forward), and day of the week. Figure 4c–f show that only RigiTech Eiger is able to fly the distance between the two most separated hospitals.

Overall, for all six hospital pairs, for both direction flights, the average reduced travel time remains relatively consistent from Monday to Friday, with a moderate decline during the weekend (Saturday and Sunday), as expected due to lower traffic congestion during these days. RigiTech Eiger provides more time reductions than DJI Matrice 600 due to its higher flight speed.

Large, shaded areas, especially during weekdays, indicate high variability in travel time data—that is, significant differences between the maximum and minimum road travel times. Consequently, this also reflects large variations in the estimated time savings when using drones instead of road transport. Such high dispersion implies lower reliability of the data, particularly when a decision-maker needs to evaluate the potential benefit of performing a drone delivery between a specific hospital pair on a given day. In these cases, less consistent time savings (as reflected by broader shaded areas) suggest lower confidence in the predictability of the advantage offered by drone transport.

Figure 5 illustrates the temporal distribution of time savings on an hourly scale. Clear patterns of variability across the day can be observed, with morning and afternoon peaks typically showing larger advantages for drones due to higher road congestion. Despite differences in range and performance between the two drone types, both consistently outperform road transport, with RigiTech Eiger achieving greater time savings on longer routes (e.g., between hospital pairs 2 and 3). The shaded areas highlight the range between minimum and maximum drone cruise velocity scenarios, while mean values provide a stable reference for comparison. It should be noted that Figure 4 and Figure 5 present potential delivery time reductions for each hospital pair, flight direction, and drone type by hour and day of the week, independent of the traffic distribution shown in Figure 2.

Figure 6 shows the correlation between road distance between hospitals and reduced travel time, which reveals how traffic variability affects flights with the same origin and destination. For example, flights covering an average distance of 8989 m and an average road distance of 16,947 m for hospital pair 4 exhibit significant differences in reduced travel time (from 5 to 22 min). It should be noted that the same hospital pair may result in different road distances depending on traffic congestion and routing direction. The linear trend line indicates a positive correlation between the road distance separating hospitals and the reduced travel time achieved.

Analysis also showed that the highest amount of calculated reduced travel time generally occurs during the early morning and late afternoon to early evening hours, while the lowest values are observed during the late night and before early morning. Based on the amount of reduced travel time calculated—which serves as a proxy for traffic congestion—additional analysis has been conducted by dividing a 24 h period into four main segments.

• The morning peak (6:00–10:00) is marked by a rapid increase in congestion as commuters begin their day.
• The midday off-peak (10:00–16:00) is a period of relatively lower congestion.
• The evening peak (16:00–20:00) is when congestion rises again during the return home and evening activities.
• The night off-peak (20:00–6:00) is a period of significantly reduced congestion.

This segmentation is consistent with findings from urban mobility studies and reflects typical commuter behavior in major cities.

Table 3. Average time reduction per flight (in minutes) by hospital pair, time period, and drone model. Cells marked with “/” indicate no flights due to limited drone range or flight schedule.

Hospital Pair	Morning Peak		Midday Off-Peak		Afternoon Peak		Night Off-Peak
	DJI	Eiger	DJI	Eiger	DJI	Eiger	DJI	Eiger
1	4.9	7.3	5.8	6.7	6.0	7.6	4.3	5.5
2	/	9.3	/	9.5	/	10.4	/	6.6
3	/	17.1	/	10.2	/	10.3	/	7.5
4	10.1	17.1	11.1	13.2	12.4	14.1	8.3	13.8
5	10.4	11.4	9.9	14.8	10.8	13.7	7.7	11.1
6	2.5	/	2.7	2.9	2.6	2.8	2.0	2.7

presents the average time reduction per flight (in minutes) for each hospital pair, segmented into four time periods reflecting traditional traffic conditions. The table compares the performance of two drone models, highlighting how each model contributes to reducing travel time under varying traffic scenarios. Note that some cells in the table are blank because, for example, DJI Matrice 600 does not have sufficient range for hospital pairs 2 and 3, as these routes exceed 10 km. Additionally, for hospital pair 6, no operations were performed using the Eiger drone during the morning peak period, which explains the absence of data in the table for that time period.

Table 4. Summary of the results.

Metric	DJI Matrice 600	RigiTech Eiger
Minimum reduction (min)	1.6	2.4
Maximum reduction (min)	18.8	25.7
Average reduction per flight (min)	7.0	9.6
Daily number of flights	87	93
Average daily reduction (min)	612.1	890.3

shows a summary of the results. The smallest reduction in travel time is 1.6 min, observed at 3 a.m. on Wednesday using the DJI Matrice 600 drone on the hospital pair from La Paz to HU Carlos III (1.1 km distance). In contrast, the largest reduction in travel time is 25.7 min, recorded at 8 a.m. on Tuesday using the RigiTech Eiger drone on the hospital pair from Hospital 12 de Octubre to HU Carlos III (10.9 km distance). The average reduction per flight is higher for Eiger (9.6 min) compared to DJI Matrice 600 (7.0 min). With 87 flights per day for Matrice 600, the average daily time savings amount to 612 min, which is roughly 10.2 h. In contrast, Eiger, operating 93 flights per day, results in an average daily reduction of 890 min, or about 14.8 h. This indicates that Eiger consistently provides greater time savings per flight, leading to substantially higher overall operational efficiency and a significant reduction in driving time for hospitals.

Another important concept in emergency delivery is the predictability of the delivery time. Some blood samples or vaccines have a time limit after which they are no longer useful. Depending on the type of material, this limit is different, but values above it shall not be accepted. The variability of the travel time, and especially the worst cases, is very relevant for the emergency delivery analysis.

Thus, as part of additional analysis, drone travel times are also compared to motorcycle and ambulance travel times, approximated as explained in the Section 3. The aim is to assess which transportation means can deliver within an established time limit with more reliability.

We tested four cases, setting maximum delivery times to 5, 10, 15, and 20 min, and performed a sensitivity analysis to determine which proportion of transports can reach the destination hospital within the specified times. It is noted that these thresholds are defined based on an extensive review of recent European drone-based medical delivery as well as real-world applications [57,58,59]. This reveals that transport times in the 5–20 min range are both realistic and operationally meaningful in urban and peri-urban contexts. There are seven transportation means options: two aerial (using the two drone models) and five ground transportations (using regular cars, motorcycles with two speeds, and ambulances with two speeds). The results of this sensitivity analysis for the different delivery time limits are presented in

Table 5. Proportion of trips meeting emergency delivery time thresholds—sensitivity analysis across drones and road transport models.

	Emergency Delivery Time Threshold
Transportation Mean	<5 min	<10 min	<15 min	<20 min
DJI Matrice 600	20%	53%	100%	100%
RigiTech Eiger	14%	83%	100%	100%
Regular car	4%	17%	39%	67%
Motorcycle 1 (−10%)	11%	23%	40%	89%
Motorcycle 2 (−20%)	14%	31%	56%	96%
Ambulance 1 (−30%)	14%	39%	82%	99%
Ambulance 2 (−40%)	14%	40%	96%	100%

.

The sensitivity analysis reveals marked differences in the ability of various transport modes to meet specific delivery time thresholds. For instance, both drone models (DJI Matrice 600 and Eiger) consistently achieve 100% of deliveries within 15 min, with the Eiger model performing notably better at the <10 min threshold (83% compared to 53% for DJI Matrice 600). In contrast, regular cars manage only 39% of trips within 15 min and just 4% within 5 min, highlighting their limitations in time-sensitive scenarios. Adjusted road transport modes show improved performance. Motorcycle 1 (a 10% speed improvement) and Motorcycle 2 (a 20% improvement) reach 40% and 56% under 15 min, respectively, while ambulances with 30–40% enhancements achieve significantly higher percentages at these critical thresholds. Despite these improvements, the analysis underscores that for very short delivery windows (e.g., under 15 min), drones remain far more reliable. The proportion of flights and trips meeting the higher emergency delivery time threshold (25 to 45 min) is above 96%.

Overall, while all modes eventually meet longer delivery times (i.e., >45 min), for emergency scenarios requiring rapid response, the inherent speed advantage of drones is clearly demonstrated, and even optimized road transport modes may struggle to consistently meet the most stringent time thresholds.

Figure 7 presents a box plot for each vehicle type, illustrating the median, quartiles, and outliers for delivery duration in minutes. This visualization enables a straightforward comparison of central tendency and variability between transport modes, highlighting not only the average performance but also the consistency and reliability of delivery times. Notice that, in addition to the trip duration, the time variability is lower for the flights compared with all ground transportation means, producing a more predictable time of delivery.

The values obtained in this study demonstrate the advantages of aerial transportation means and quantify the savings in time. Absolute reductions range from 1.6 min to 25.7 min, while relative reductions go from 35% and 58% of the time needed with a regular car. Fluctuations depend on the time of the day and day of the week, and also on the traveling distance. Drones have proven to be the best option for emergencies that have a time limit below 15 min.

The results also show that, in general, the time savings are larger with the RigiTech drone model than with DJI Matrice 600. While DJI Matrice 600 was a common drone model at the beginning of the MUSE project, it is currently outdated. It is clear that even when using other similar multi-copter models currently on the market, the higher cruise speed of RigiTech Eiger or similar tilt-rotor aircraft makes these vehicles better suited for long-distance emergency deliveries. To improve the operational efficiency and ease the maintenance tasks, most drone operators would typically decide to use a single drone model for their fleet. The results show that for these types of operations, RigiTech Eiger is a better choice than the small multi-copters, as DJI Matrice 600.

5. Conclusions and Future Work

In congested urban spaces, the use of aerial means for transportation offers a great opportunity for emergency delivery. The results presented above are based on a credible scenario for the city of Madrid. A subset of hospitals of the public network is selected using the guidance of experts and existing plans for aerial demonstration flights. The designation of medical corridors for drone emergency deliveries follows the EASA Air Mobility Hub concept for eVTOL operations, flying fixed routes with pre-approved flight plans and reserved airspace, and using one of the approved drone models.

There is already substantial evidence that medical drone-based delivery has gained relatively high public acceptance, and this trend is expected to continue. The literature on the social acceptance of urban air mobility shows that citizens select the use of drones for emergency deliveries as the best use case, with acceptance levels up to 80%. Compared to ambulances or motorcycles, drones generate significantly less disruptive noise and no greenhouse gas emissions, which is especially beneficial in dense urban environments. They are also expected to be more cost-effective than conventional ambulance transport, especially for time-sensitive but lightweight deliveries.

This study highlights how drone-based emergency deliveries in urban environments can improve time efficiency with respect to the current emergency delivery means based on ground transportation. While road traffic highly fluctuates throughout the day, the air traffic provides a great opportunity for deterministic travel time, with significant reductions especially occurring during peak hours with road traffic congestion. Many other works are approaching the same opportunity, and a number of demonstration flights are being executed in cities across the world. Nevertheless, many of them are still on the ground, waiting for permission to take off, and the ones that have obtained permission are not flying at scale. This work addresses a (hopefully) near-future scenario in which the emergency deliveries will happen on a daily basis and shows their benefits in terms of efficiency, in particular, providing time reduction figures.

More hospitals, corridors, and drone models will be viable for medical delivery operations as soon as safety assessment rules and means of compliance are settled for high ground risk areas. This will likely expand the potential benefits of such systems (e.g., reduction in CO₂ emissions, traffic jams, higher reliability in delivery service, etc.), enabling greater operational flexibility and efficiency. For now, emergency air deliveries are mainly found in rural areas or above water, where the ground risk is considered low and the mitigation measures are easier to put in place. The current barriers to having scaled delivery services in our cities come from the current regulatory restrictions for BVLOS flights above high-density areas. Advances in regulation, such as the EASA guiding material for eVTOL certification or the recently published enhanced SORA 2.5 in Europe, shall help solve current safety issues. Projects, such as SAFIR-Ready, have set the basis for regular delivery between hospitals and remote users. The experience has shown the actual benefits and cost savings of this type of operation, but it has yet to be generalized within cities.

Several improvements can still be developed to have better predictions. First, a deeper knowledge of the processes involved in an emergency and an extensive dataset with records describing the origin–destination–time–weight of emergency deliveries will help to better shape the demand with more realistic, less stochastic traffic. Second, an economic study for the selection of the most appropriate fleet of drones to meet the demand is needed. Third, there needs to be a safer route selection between pairs of hospitals, which may expand the flight time by a small percentage but shall consider less-populated areas for route selection. Fourth, this study is limited by the fact that the Google Routes API relies primarily on historical traffic profiles for future dates, with only minimal differentiation across weeks. Consequently, our approach emphasizes representativeness at the daily level (weekday vs. weekend) rather than seasonal differences. Future research could address this limitation by validating the API-based results with empirical floating car data (e.g., ambulance or motorcycle traces) or by simulating congestion dynamics more explicitly. Finally, the flight dynamics could be validated with real flights to better adjust the duration of the flights, especially by recreating more power-efficient take-off and landing patterns.

In order to achieve an emergency air delivery scenario, like the one presented in this research, a number of steps need to happen. The first is the designation of the U-space airspace that envelopes the emergency air delivery traffic and safely separates it from other manned aviation. Preceding the declaration of a U-space airspace, the airspace risk assessment (ARA) is needed to evaluate risks associated with drone operations and inform the requirements for U-space services, operational conditions, and mitigation strategies. Further, the governance of the U-space must involve national air safety agencies and service providers and also the local governments and the end users, such as emergency services. In addition, drone operators need to believe that a viable business model exists for the safety and operational performance restrictions that the U-space may establish.