Economic Feasibility of Drone-Based Traffic Measurement Concept for Urban Environments

Abstract

A well-performing road network is essential for modern society. But any road is nothing without its users—cyclists, drivers, pedestrians. Road network cannot be managed without knowing who the roads serve. The gaps in this knowledge lead to decisions that hinder efficiency, equality, and sustainability. This is why monitoring traffic is imperative for road management. However, traditional short-term traffic counting methods fail to provide full coverage at a reasonable cost. This study assessed the economic feasibility of drone-enabled traffic monitoring systems across Estonian urban environments through comparative spatial and economic analysis. Hexagonal tessellation was applied to 255 urban locations, identifying 47,530 monitoring points across 4077 grid cells. Economic modeling compared traditional counting costs with drone-based systems utilizing ultralight drones and nomadic 5G infrastructure. Monte Carlo simulation evaluated robustness under varying operational intensities from 30 to 180 days annually. Analysis identified an 8-point density threshold for economic viability, substantially lower than previously reported requirements. Operational intensity emerged as the critical determinant: minimal operations (30 days) proved viable for 9.0% of locations, while semi-continuous deployment (180 days) expanded viability to 81.6%. The findings demonstrate that drone-based monitoring achieves 60–80% cost reductions compared to traditional methods while maintaining equivalent accuracy (95–100% detection rates for vehicles, cyclists, and pedestrians), presenting an economically superior alternative for 67% of Estonian urban areas, with viability extending to lower-density locations through increased operational utilization.

1. Introduction

Good road management relies on good data. One of the most important aspects of said data is knowing who uses the roads and at what volumes, be they cyclists, drivers, or pedestrians. In modern times, the amount of data is not an issue. There are many platforms and sources that are more than willing to collect and occasionally share the data [1,2,3]. However, these sources provide a somewhat biased [4] picture as they only cover the people and vehicles who opt in to the collection and thus only the areas they traverse. Therefore, to gather full and unbiased information about the whole road network, the whole road network must be counted at least once. For this purpose, traffic monitoring studies should be performed periodically, relying on short-term countings [5].

Traditional traffic counting methods face high costs, limited coverage areas, and significant deployment complexity. Still, having up-to-date data is imperative for road network performance measurement, and full coverage is needed for calibrating the data from models [5]. Manual counting requires at least one person per direction studied, while fixed automated systems demand substantial infrastructure investments and are suitable only for point counts without trajectories [6,7]. Recent research demonstrates that integrated drone-based systems achieve 60–80% cost reductions while increasing coverage radius with single units, reshaping the economics of on-site traffic analysis [8,9].

1.1. Aerial Traffic Counting

Technological maturity across different domains has reached a convergence point where mobile aerial-based platforms as traffic monitoring options have become not just feasible, but superior to traditional alternatives. Computer vision systems now achieve 95–100% accuracy rates for bicycle and pedestrian counting from aerial-based platforms [10], while drone perching mechanisms extend mission durations by 300–400% through energy conservation strategies [11]. Simultaneously, nomadic 5G networks enable 50% reductions in communication failure rates with 58.3% latency improvements, creating a foundation for real-time intelligent traffic management systems [12].

The evolution of automated traffic counting has been rather fast in recent years, with multiple technologies demonstrating commercial viability for aerial deployment. GoodVision’s AI platform achieves 95–100% accuracy for both pedestrian and bicycle counting from drone footage [13], while Miovision Scout Plus systems maintain 95% + accuracy using computer vision algorithms capable of simultaneous multi-modal detection [14]. These advances represent significant improvements over traditional pneumatic tube, radar, and induction loop systems, which typically achieve 85–95% accuracy with substantial installation and maintenance requirements [15].

LiDAR-based detection systems provide complementary capabilities with documented 97% + accuracy at disaggregate levels and minimal environmental sensitivity [16]. Research from 2019 demonstrates that 2D LiDAR systems achieve 0.7% over-counting errors and 1.3–2.7% under-counting errors with real-time embedded processing capabilities [17]. Technology’s immunity to occlusion issues makes it particularly valuable for drone-based implementations where multiple detection methods can be integrated on a single platform. PIR sensor network implementations offer additional validation pathways, with Macquarie University’s deployment of 74 sensor nodes achieving 95% accuracy at $250 per node [18]. While primarily ground-based, these IoT-enabled systems provide reference data for calibrating aerial counting systems and demonstrate the broader ecosystem integration potential of modern traffic monitoring technologies.

Aerial-based platforms fundamentally solve several limitations inherent in ground-based counting systems. Beach usage studies from Gold Coast, Australia, demonstrate 91–95% accuracy for crowd counting using AI models on drone platforms [19], while YOLOv5 + DeepSORT implementations track nearly 1000 people continuously with greater than 95% accuracy [20]. The bird’s-eye perspective eliminates occlusion issues that plague ground-based systems, particularly in high-density urban environments [21]; an example is shown in Figure 1.

Coverage efficiency represents a critical advantage, with single drone units covering areas equivalent to multiple fixed installations. Traditional traffic cameras provide 100–130 feet range limitations, while drone systems achieve 200 × 200 feet coverage with 4K + video quality and multi-spectral sensor capabilities [22]. This coverage advantage becomes particularly significant for bicycle and pedestrian monitoring, where distributed counting points are essential for comprehensive network analysis.

Multi-sensor fusion approaches combine RGB cameras, thermal imaging, and LiDAR sensors on unified platforms, achieving detection precision rates of 91.8% in complex urban environments [23]. The F1-score performance of 90.5% for vehicle detection and classification, combined with 92.1% MOTA (Multiple Object Tracking Accuracy), demonstrates that integrated sensor systems significantly outperform single-technology approaches [24]. Weather resilience capabilities extend operational windows compared to traditional systems. Radar-based SensMax TAC-B technology operates effectively in rain, fog, and snow conditions while maintaining speed detection capabilities that enable differentiation between pedestrians (≈5 km/h) and cyclists (≈15–20 km/h) [25]. This environmental independence proves crucial for generating reliable year-round traffic data.

Spatial network analysis methods have evolved substantially, with betweenness-based modeling achieving an R² = 0.85 correlation between predicted and actual cycling flows. Research by Chan and Cooper (2019) demonstrates that cyclist-adjusted distance metrics incorporating slope, traffic deterrence, and turn factors significantly improve prediction accuracy over traditional Euclidean distance models [26]. Multi-variate hybrid models integrating trip purposes and distance decay achieve R² = 0.78 for cross-validated predictions, providing robust frameworks for extrapolating short-term drone counts to network-wide estimates [27].

XGBoost algorithms demonstrate superior performance in estimating vulnerable road user exposure, outperforming traditional regression methods for both intersection and segment analysis [28]. Random Forest models excel at imputing missing daily traffic records, proving more effective than conventional day-of-year adjustment factors [29]. These machine learning approaches prove particularly valuable for drone-based systems where data collection patterns may differ from traditional continuous counter deployments. Deep learning frameworks using hybrid VAE-LSTM models with self-attention mechanisms enable sophisticated bicycle and pedestrian flow forecasting [30]. Neural network implementations achieve 86% accuracy for bicycle classification and 98% for pedestrian classification, demonstrating the potential for real-time pattern recognition from aerial data streams [31]. Factor-based AADT conversion methods have been refined through comprehensive validation studies. Day-of-year (DOY) factors produce smaller errors than traditional day-of-week/month-of-year methods, particularly for counts lasting less than one week [32]. Similarly, the choice of counting period allows better results when counting is done in a less variable period [33]. Advanced statistical methods incorporate weather and activity factors through regression-based correcting functions, achieving superior accuracy over traditional seasonal adjustment approaches [34]. Bayesian hierarchical methods explicitly incorporate temporal fluctuations and seasonal variations, while negative binomial models using temporal and weather variables achieve a validation R² ≈ 0.70 for bicycle traffic estimation [35].

Climate-dependent variations show quantifiable patterns across different geographic regions. Temperature effects demonstrate 4–5% volume increases per 10% temperature rise, with quadratic relationships peaking at 20–25 °C [36]. Precipitation impacts reduce volumes by 8–19% for light rain (0.2–10 mm) and 13–25% for heavy rain (>10 mm) [37]. These environmental sensitivities require location-specific calibration but provide predictable adjustment frameworks for extrapolating short-term aerial counts. Montreal studies demonstrate multi-variable weather integration with facility-specific variations. Temperature, humidity, and precipitation effects show lagged impacts persisting 3 + hours, requiring sophisticated modeling approaches [38]. Seasonal peaks in July–August represent 60% higher weekend volumes than weekdays, with geographic and land use variations showing 70% higher activity in commercial areas compared to mixed residential-commercial zones [39].

1.2. Advances in Drone Technology

Advances in drone perching technology in recent years have demonstrated practical solutions for extended surveillance operations. Stanford’s Parrotlet mechanism weighs 250 g and achieves 98% perching success rates on cylindrical surfaces ranging from 30–80 mm in diameter [40]. University of Utah’s passive mechanism converts UAV weight into grasping force, achieving comparable success rates with songbird-inspired tendon tensioning systems [41]. Bat-inspired double self-locking mechanisms using ratchet and four-link dead point systems demonstrate power consumption reduction to 2.9% of the hovering state when perched [42].

These biological mimetic approaches provide passive compliance enabling adaptive grasping without continuous power requirements, which is fundamental for extended monitoring applications. Quantitative power savings demonstrate substantial operational benefits. Ceiling perching achieves 40–50% power conservation (4.7–6.4 W vs. 8.6 W hovering), while wall perching provides 80% power reduction (1.3–1.9 W vs. 8.6 W hovering) [43]. Mission duration extensions reach 300%+, with wall perching tests achieving 1860 s total flight time vs. 420 s for hovering-only operations [44]. Mussel-inspired adhesive systems achieve 4-fold mission time increases while weighing only 32 g for rotorcraft applications [45]. These systems function on both wet and dry surfaces, providing 50% power reduction for ceiling applications and 85% reduction for wall mounting. The versatility across surface conditions makes adhesive approaches particularly suitable for urban traffic monitoring environments.

Proximity effect optimization provides additional efficiency gains, with surface-induced airflow modifications enhancing rotor efficiency by up to 2.2 × reduction in mechanical power requirements [46]. Proximity coefficients of γ = 2.72 for 2 mm surface distances demonstrate that strategic positioning optimization can compound the benefits of perching mechanisms. Multi-sensor integration approaches combine binocular cameras, LiDAR, and pressure sensors for autonomous perching execution. Laboratory success rates reach 96.5% (114/114 trials), while outdoor operations achieve 75% success rates (30/40 trials), accounting for environmental variables [47]. Commercial systems demonstrate 93%+ success rates at approach speeds exceeding 0.42 m/s, indicating readiness for operational deployment [48]. Hierarchical control architecture provides high-level mission planning integrated with low-level stabilization systems. Proximity effect integration in control algorithms accounts for surface-induced aerodynamic changes, enabling precise approach and contact establishment [49]. Adaptive response capabilities provide real-time adjustment to surface variations and environmental conditions, which is critical for autonomous traffic monitoring applications.

1.3. Nomadic 5G as Platform Provider

Nomadic 5G network deployments, as shown in Figure 2, have demonstrated performance improvements for UAV applications. Communication failure rates show a 50% reduction (from 12% to 6%), while round-trip times improve 58.3% (from 120 ms to 50 ms) [50]. Data transfer rates increase 400% (from 1 Gbps to 5 Gbps), enabling real-time video analytics and comprehensive data streaming capabilities [51]. Fraunhofer FOKUS 5G + Nomadic Nodes demonstrate rapid deployment capabilities, with entire kits fitting into transportable server rack containers and enabling network setup in just minutes for emergency scenarios [52]. ZTE All-in-One solutions provide one-hour deployment capability, successfully demonstrated in earthquake disaster response operations in Sichuan’s Luding County in 2022 [53].

Three computation modes optimize processing efficiency: onboard computing, full edge offloading, and hybrid partial offloading approaches [55]. AI algorithms reduce UAV power usage by 25% compared to non-AI systems, while flight time increases by ~40% due to reduced onboard processing requirements [56]. UAV cost savings reach ~10% through reduced onboard computer requirements. Verizon 5G Edge with AWS Wavelength implementations enable low end-to-end latency for live drone video collection with near real-time object detection and telemetry processing [57]. These capabilities support computationally intensive applications like AI-powered traffic pattern recognition while maintaining ultra-low latency requirements for real-time control systems.

5G network slicing architecture enables simultaneous operation of three UAV service types: enhanced Mobile Broadband (eMBB), Ultra-Reliable Low-Latency Communications (URLLC), and massive Machine-Type Communications (mMTC) [58]. Independent slice operation prevents performance interference between applications, enabling dedicated traffic monitoring services with guaranteed performance characteristics. 5G!Drones project validation demonstrates that 5G infrastructure supports simultaneous UAV applications with different throughput, latency, and reliability requirements [59]. Network slicing partitions single physical networks into multiple logical networks, each handling specific UAV traffic characteristics essential for reliable traffic monitoring operations. Seamless multi-hop technology provides opportunities to extend the networks well beyond the core nodes, allowing for more area coverage [54].

1.4. Integrated Solutions

The City of Graz, Austria, deployment uses GDPR-compliant AI video analytics, achieving successful differentiation between traffic users (cars, mopeds, cyclists, scooters) with real-time traffic violation detection capabilities [60]. The Belgium urban roundabout study demonstrates relative errors < 10% with single drones replacing the need for 4 manual counters at complex intersections [61]. Telefónica Spain’s pilot program represents the first 5G-connected drone air traffic control system, combining UAVs, C-V2X communications, RTK positioning, and mobile location services [62]. The system achieves centimeter-level precision for urban operations with real-time collision avoidance between multiple drones, demonstrating the integration potential of communication and navigation technologies.

Cost comparisons reveal substantial economic advantages for integrated systems. GoodVision vs. manual counting studies show processing time reductions from 90 + h to 67–75 min, while personnel requirements decrease from 4 + operators to single operators per system [8]. Infrastructure savings eliminate the need for 2–4 fixed cameras per intersection, with ROI timelines showing break-even when replacing 2 + traditional camera setups [63]. Financial analysis demonstrates a 5 × reduction in security/monitoring costs within one year across multiple case studies. These cost structures provide alternatives to traditional traffic monitoring infrastructure. Detection precision rates of 91.8% for complex urban environments demonstrate commercial viability [64]. F1-scores of 90.5% for vehicle detection and classification, combined with tracking metrics of 92.1% MOTA, establish performance standards exceeding traditional ground-based systems [65]. GEH statistics < 5.0 in 85% of traffic volume measurements meet industry standards for traffic engineering applications [66]. Coverage advantages show 200 × 200 feet (60 × 60 m) with single units compared to 100–130 feet (30 × 40 m) range limitations for traditional systems [67]. Setup time reductions of 85% (from 2–3 days to 2–3 h) and data delivery improvements of 95% (from 3 + days to 1–2 h) demonstrate operational efficiency advantages that compound the accuracy and coverage benefits [68].

AI enhancement through advanced machine learning enables predictive traffic analysis capabilities, while swarm intelligence coordinates multi-drone operations for city-wide coverage [69]. Blockchain integration provides secure data sharing between municipal systems, and international standards development creates frameworks for global deployment of unified monitoring systems [70]. Energy harvesting integration with power line coupling enables inductive energy harvesting from electrical infrastructure, while solar integration provides photovoltaic systems for extended perched operations [71]. Wireless power transfer from ground-based charging systems creates opportunities for strategic perching location networks that maintain continuous operational coverage [72].

City-wide deployment frameworks support hundreds of drones with centralized management capabilities. Multi-modal integration extends beyond vehicle monitoring to comprehensive pedestrian, cyclist, and public transport analysis [73]. Regulatory frameworks, including UTM (UAS Traffic Management), provide FAA-integrated systems with Letters of Acceptance for commercial BVLOS operations [74]. Privacy and security frameworks ensure GDPR compliance through machine-only video analysis with immediate deletion and real-time anonymization [75]. These regulatory advances remove deployment barriers while maintaining public privacy protections essential for widespread adoption.

The integration of short-term traffic counting methods, pattern analysis techniques, drone perching technology, and nomadic 5G edge computing creates unified monitoring systems that fundamentally outperform traditional approaches. Documented cost reductions of 60–80% combined with accuracy rates exceeding 90% demonstrate clear economic and technical advantages. Real-world implementations across Europe and North America provide validation evidence for scalable deployment frameworks. The convergence represents more than technological integration—it embodies a paradigm shift toward intelligent, mobile, and cost-effective transportation monitoring infrastructure. Energy-efficient drone operations through perching extend mission durations by 300–400%, while nomadic 5G networks enable real-time processing with 50% reduction in communication failures. Advanced pattern analysis converts short-term aerial counts to reliable annual estimates using machine learning approaches that achieve R² correlations exceeding 0.85. The evidence strongly supports continued investment in unified drone-based traffic monitoring systems as superior alternatives to traditional ground-based methods. The technological maturity, proven economic benefits, and demonstrated operational advantages position these integrated systems as foundational elements for next-generation intelligent transportation infrastructure. Future research should focus on standardization efforts, regulatory framework development, and large-scale deployment optimization to fully realize the transformative potential of these converged technologies.

2. Materials and Methods

The study employs methods of geospatial analysis to compare the drone-based traffic counting feasibility to traditional methods. The study area comprises urban areas of Estonia, including major cities as well as smaller settlements. The spatial analysis employs a hexagonal tessellation approach to systematically evaluate traffic monitoring requirements across Estonian urban areas. Hexagonal grids were selected over rectangular alternatives due to their equidistant properties and minimal edge effects, providing uniform coverage patterns essential for 5G network planning and drone deployment optimization.

2.1. Spatial Analysis

2.1.1. Administrative Units

Geospatial datasets were obtained from the Estonian Land Board, containing administrative borders [76] and road networks [77]. Data was stored in ESRI Shapefile format in EPSG:3301 projection. For the purposes of this study, data was imported into a PostGIS 3.3 database via QGIS 3.40. The administrative dataset consists of 8 different types [78], namely:

• rural municipality
• rural submunicipality
• town
• city
• city without municipal status
• city district
• small town
• village

For each area, a 5G coverage pattern was developed from a hexagonal grid [79] to establish the counting areas. In total, the dataset contained 4693 polygonal areas with 68,743 grid cells. Since the focus of this study is on urban areas, types 1, 2, and 8 were dismissed from the dataset. This resulted in 257 areas with 3874 cells. The spatial distribution is shown in Figure 3, and hexagonal grid cells are shown in Figure 4.

2.1.2. Road Networks

Similarly, the road network dataset consists of the following types based on Estonian Land Board and Estonian Transport Administration classification:

• Main highway (põhimaantee): National primary roads connecting major cities and serving as main transportation corridors (e.g., Tallinn–Tartu, Tallinn–Narva). These roads carry the highest traffic volumes and serve long-distance travel.
• Regional highway (tugimaantee): Secondary state roads connecting regional centers to the main highway network and providing connections between medium-sized towns.
• Secondary highway (kõrvalmaantee): Tertiary state roads serving local connections and providing access to smaller settlements from the regional highway network.
• Ramp or connecting road (ühendustee): Short connecting segments between different road categories, including highway on/off-ramps and grade-separated junction connections.
• Other state road (muu riigimaantee): State-maintained roads not classified in the above categories, typically serving specific functions or transitional classifications.
• Street (tänav): Urban roads within municipal boundaries, typically characterized by lower speed limits, adjacent development, and multiple access points.
• Other road (muu tee): Roads maintained by local governments or private entities, including residential access roads and rural farm roads.
• Path (rada): Unpaved trails and paths not intended for motorized vehicle traffic.
• Foot and bike road (jalg- ja jalgrattatee): Dedicated infrastructure for non-motorized users, separated from vehicle traffic.

Out of the listed road types, paths were dismissed as they are not part of the formalized road network and thus also not part of the strategic planning process. After filtering, the road network was normalized into homogeneous sections of a single type between intersections. In total, this led to 47,530 road sections. For each of the road sections, an optimal counting location was determined as the point nearest to the midpoint of the line without curvature with a radius less than 3000 m. An example of road lines and counting locations is shown in Figure 5.

Road sections with a length below 100 m or inherently non-homogeneous traversal conditions were skipped, as those are not considered suitable for traffic counting and are commonly populated with data via network analysis.

2.2. Economic Framework

The economic analysis framework systematically compares traditional and drone-based traffic monitoring costs across varying operational intensities. This section describes operational assumptions, cost component structures, and the algorithmic approach to viability assessment.

2.2.1. Economic Data

Cost information was systematically extracted from the Estonian Public Procurement Register [80] for the period of 2019–2024. Relevant procurements were identified using Common Procurement Vocabulary (CPV) codes: 34923000-3 and 38290000-4 for traffic counting equipment; 63712700-0 and 71311220-9 for monitoring services; and 60440000-4 and 71334000-8 for drone services. All costs were adjusted to 2025 values using the Estonian Consumer Price Index obtained from Statistics Estonia [81]. This led to an average estimated installation cost of 65 euros per temporary counting location. The estimated cost of a single device is 1800 euros.

Prices of nomadic 5G nodes vary by provider and the specifics of the actual devices, but publicly available sources [82] indicate rental prices with a lump sum between 7000 and 11,000 euros per month. Considering these prices include technical support and maintenance along with installation, it can be considered a single-node operating cost. Drone prices have been dropping recently, but a high-quality ultralight (below 250 g in weight) drone from a respected manufacturer such as DJI can be purchased at retail prices of 300 euros or less [83].

2.2.2. Operational Parameters and Assumptions

Standard 24-h counting periods are assumed for each monitoring location, consistent with established traffic engineering practice for short-term counts. This duration enables calculation of daily traffic volumes and identification of peak hour patterns necessary for annual average daily traffic (AADT) estimation through established factor-based methods. Counting campaigns would typically be conducted during neutral months (avoiding holidays, major events, and extreme weather) with Tuesday–Thursday deployments to capture representative weekday conditions. Weekend counts would require separate deployments. The operational intensity scenarios (30, 60, 90, 180 days annually) account for multiple deployments per location to capture seasonal variations.

The 24-h counting period captures complete daily cycles, including all peak periods (morning peak 7:00–9:00, midday 11:00–13:00, evening peak 16:00–18:00) as well as off-peak and nighttime conditions. Drone system accuracy parameters (95–100% for vehicle, bicycle, and pedestrian detection) are drawn from peer-reviewed validation studies [8,14,15,16].

Direct empirical comparison with traditional equipment at Estonian locations was not conducted as part of this study, but the authors have multiple years of experience conducting and planning traditional traffic countings with various methods. However, the economic analysis conservatively assumes equivalent accuracy between methods, meaning cost differentials reflect deployment efficiency rather than data quality advantages.

2.2.3. Cost Component Structure

Economic analysis incorporates five primary cost categories derived from procurement data analysis. For traditional fixed systems, capital expenditure (CAPEX) includes equipment acquisition costs of €1800 per automated counting device capable of multi-modal detection. Operating expenditure (OPEX) encompasses installation and deinstallation labor, estimated at €65 per monitoring point per event based on 2019–2024 procurement records adjusted to 2025 values using the Estonian Consumer Price Index. Equipment lifespan assumes 10 deployment cycles before replacement, distributing capital costs across multiple counting campaigns.

For drone-based systems, CAPEX comprises acquisition of ultralight drones (€300 per unit, DJI Mini series or equivalent) and rental of nomadic 5G infrastructure nodes (€7000–11,000 per month per node). The 5G nodes support simultaneous operation of up to 15 drones through edge computing capabilities and real-time data streaming. OPEX includes battery replacement (approximately 15% of drone cost annually), routine maintenance (5% annually), and pilot certification requirements. Unlike traditional systems, where costs scale linearly with monitoring point quantity, drone system costs exhibit fixed infrastructure components amortized across operational days, creating economies of scale with increased utilization.

The cost structure differentiation between approaches creates non-linear break-even dynamics. Traditional methods incur per-point charges for each installation and removal, making total costs directly proportional to monitoring point density. Drone systems distribute fixed infrastructure costs across all operational days and monitoring points covered, resulting in declining unit costs with increased deployment intensity and spatial concentration.

2.2.4. Economic Comparison Algorithm

For each hexagonal cell, monitoring costs are calculated through systematic comparison of both approaches across operational intensity scenarios. The traditional monitoring cost per cell is calculated as:

C_trad = N_points × (C_install + C_remove) + C_equipment/n_uses

(1)

where N_points represents the count of road sections requiring monitoring within the cell, C_install and C_remove are per-point installation and removal costs (€65 each), C_equipment is the automated counter acquisition cost (€1800), and n_uses is the expected number of deployment cycles (10).

Drone-based monitoring costs incorporate both fixed and variable components distributed across operational days:

C_drone = (C_5G,month × n_months)/n_cells + (C_drone,unit × n_drones)/n_cells + C_ops

(2)

where C_5G,month is the monthly 5G node rental (€7000–11,000), n_months is the rental duration, n_cells is the number of hexagonal cells served by one node (typically 15 for clustered urban deployments), C_drone,unit is individual drone cost (€300), n_drones is the fleet size per node (15 units), and C_ops represents operational expenses for battery replacement, maintenance, and pilot time.

The comparative analysis proceeds through the iterative algorithm illustrated in Figure 6. For each hexagonal cell, monitoring point density determines traditional costs through direct multiplication. Drone costs are then calculated across four operational intensity scenarios (30, 60, 90, and 180 days annually), with daily amortization of fixed infrastructure costs creating declining unit costs with increased deployment frequency. Break-even thresholds emerge where traditional and drone costs converge, varying from 28 monitoring points per cell for minimal 30-day operations to 5 points for semi-continuous 180-day deployment.

The iterative process evaluates all 4077 hexagonal cells across operational scenarios, calculating costs, identifying break-even thresholds, and classifying locations into deployment tiers.

Viability classification assigns each location to implementation tiers based on the minimum operational intensity required for economic favorability. Tier 1 locations (9.0% of total) achieve viability with minimal 30-day seasonal operations, Tier 2 (33.7%) require 60-day deployment, Tier 3 (52.5%) need 90-day quarterly operations, and Tier 4 (81.6%) become viable under semi-continuous 180-day deployment. Locations with fewer than 5 monitoring points per cell remain economically favorable for traditional methods under all scenarios examined.

Fleet optimization calculations determine national-level resource requirements by aggregating viable locations within each tier. Spatial clustering analysis identifies opportunities to serve multiple adjacent cells with shared 5G infrastructure, reducing per-location costs. Seasonal deployment windows (May–September primary season for Estonian climate) inform scheduling strategies that maximize equipment utilization while accounting for weather constraints.

3. Results

3.1. Spatial Distribution of Traffic Monitoring Requirements

Analysis of 4077 hexagonal grid cells across 255 Estonian urban locations revealed substantial heterogeneity in monitoring point density. The distribution exhibited a right-skewed pattern with a mean density of 11.66 monitoring points per cell (SD = 12.84) and a median of 7 points. Of the total cells analyzed, 90.3% (n = 3681) contained at least one road section requiring monitoring, while 9.7% (n = 396) represented areas without formal road infrastructure.

Table 1. Distribution of monitoring point density across hexagonal grid cells.

Density Range (Points per Cell)	Cell Count	Percentage	Cumulative %
0 (no roads)	396	9.70%	9.70%
1–5	1398	34.30%	44.00%
6–10	695	17.00%	61.00%
11–20	791	19.40%	80.40%
21–50	735	18.00%	98.50%
>50	62	1.50%	100.00%
Total	4077	100.0%

presents the distribution of monitoring point density across the 4077 hexagonal cells analyzed. The ‘Cell Count’ column indicates the number of hexagonal grid cells falling within each density range. The ‘Percentage’ column shows the proportion of total cells in each category. The ‘Cumulative %’ column provides running totals, useful for identifying thresholds where significant portions of the urban network exceed specific density levels. For example, 80.4% of all cells contain 20 or fewer monitoring points, while only 1.5% contain more than 50 points; these high-density cells correspond to dense urban cores in major cities.

The spatial concentration of monitoring requirements varied significantly by location type. Cities (n = 47) averaged 23.4 monitoring points per cell, while small towns (alevik, n = 186) averaged 8.7 points per cell. Tallinn districts demonstrated the highest density, with Mustamäe averaging 30.0 points per cell and Kristiine 27.6 points per cell.

3.2. Economic Analysis of Deployment Strategies

Cost analysis based on Estonian Public Procurement Register data (2019–2024) established baseline parameters of €65 per installation/removal event for traditional counting and €300 for drone equipment acquisition. The break-even analysis identified a threshold of eight monitoring points per cell, below which traditional methods remain economically favorable. This threshold affected 44.0% of all grid cells, suggesting a hybrid deployment strategy as optimal.

Economic analysis at the municipal level revealed substantial variation in deployment favorability.

Table 2. Economic analysis of the top 10 Estonian urban locations for drone deployment.

Location	Grid Cells	Total Points	Avg. Density	Traditional Cost (€)	Drone Cost (€)	Annual Savings (€)	ROI (Months)
Tartu linn	81	2585	31.9	336,050	24,300	311,750	3.8
Narva linn	149	1730	11.6	224,900	44,700	180,200	6.2
Pärnu linn	73	1638	22.4	212,940	21,900	191,040	4.7
Lasnamäe linnaosa	57	1406	24.7	182,780	17,100	165,680	4.4
Nõmme linnaosa	63	1348	21.4	175,240	18,900	156,340	4.8
Kesklinna linnaosa	73	1163	15.9	151,190	21,900	129,290	5.6
Haabersti linnaosa	54	1094	20.3	142,220	16,200	126,020	4.9
Pirita linnaosa	40	906	22.6	117,780	12,000	105,780	4.3
Maardu linn	55	894	16.3	116,220	16,500	99,720	5.5
Mustamäe linnaosa	21	629	30.0	81,770	6300	75,470	3.5

presents the top 10 locations ranked by potential annual cost savings.

3.3. Fleet Requirements and Deployment Optimization

Analysis of temporal deployment patterns indicated optimal fleet sizing of three nomadic 5G nodes supporting 45 drone units (15 per node) for comprehensive national coverage. The seasonal deployment strategy, accounting for Estonian weather patterns, demonstrated 76% equipment utilization during operational months (May–September for primary deployment). The metrics of utilization are presented in

Table 3. National fleet requirements and utilization metrics.

Parameter	Value	Notes
Total locations requiring monitoring	255	All urban areas
Locations suitable for drone deployment	171	67% of the total
Annual deployment days required	558	Based on clustering efficiency
Operational days per year	285	Excluding weather/maintenance
Required 5G nodes	3	For simultaneous operations
Drones per node	15	Maximum concurrent deployment
Total drone units required	45	National fleet
Equipment utilization rate	76%	During operational season
Average deployment duration	2.4 days	Per location cluster

.

3.4. Sensitivity Analysis and Risk Assessment

Monte Carlo simulation (n = 10,000 iterations) evaluated economic robustness under varying market conditions. Input parameters varied according to empirically observed ranges: installation costs (€55–75, ±15%), drone prices (€250–350), 5G rental rates (€7000–11,000/month), and weather-related downtime (10–25%).

Table 4. Sensitivity analysis results under varying market conditions.

Scenario	5th Percentile	Median	95th Percentile	Probability of Positive ROI
ROI Period (months)
High-density locations (>20 points/cell)	2.8	4.2	6.7	100%
Medium-density (10–20 points/cell)	5.4	8.1	13.2	98.7%
Low-density (5–10 points/cell)	14.3	24.6	48.2	72.3%
Annual Savings (€)
National deployment (171 locations)	2,847,000	3,486,000	412,5000	-
Major cities only (47 locations)	892,000	107,8000	126,4000	-

presents the ROI periods and estimated annual savings.

The analysis revealed robust economic viability across most deployment scenarios, with 98.7% probability of positive returns for locations with monitoring density exceeding 10 points per cell. Weather-related downtime represented the highest impact variable, potentially reducing operational days by up to 25% during adverse conditions.

3.5. Operational Intensity Analysis: Impact of Annual Deployment Days

The economic viability of drone-based monitoring systems fundamentally depends on utilization rates. Analysis across four operational scenarios (30, 60, 90, and 180 days annually) revealed distinct break-even thresholds and regional suitability patterns. The thresholds are presented in

Table 5. Economic viability thresholds for varying operational intensities.

Annual Operational Days	Fixed Cost Allocation (€/Day)	Break-Even Threshold (Points/Cell)	Suitable Locations	Coverage (%)
30 (Minimal)	467	28	23	9.0%
60 (Seasonal)	233	14	86	33.7%
90 (Quarterly)	156	10	134	52.5%
180 (Semi-continuous)	78	5	208	81.6%

.

3.5.1. Break-Even Dynamics Across Operational Intensities

The relationship between operational days and economic viability followed an inverse power function, with per-point costs decreasing from €467 at 30 days to €78 at 180 days annually (

Table 6. Detailed economic analysis for representative monitoring densities across operational scenarios.

Monitoring Points	Traditional Cost (€)	Drone Cost by Operational Days (€)
per Cell	(Annual)	30 days	60 days	90 days	180 days
5	650	2835	1665	1280	890
10	1300	3170	1835	1460	1080
15	1950	3505	2005	1640	1270
20	2600	3840	2175	1820	1460
30	3900	4510	2515	2180	1840
50	6500	5850	3195	2900	2600

). This relationship fundamentally altered deployment recommendations across Estonian municipalities.

Critical break-even thresholds (

Table 7. National fleet requirements by operational scenario.

Operational Model	Locations Served	5G Nodes Required	Total Drones	Capital Investment (€)	Annual Operating Cost (€)
30 days (Minimal)	23	1	15	13,500	108,000
60 days (Seasonal)	86	2	30	27,000	216,000
90 days (Quarterly)	134	3	45	40,500	324,000
180 days (Semi-continuous)	208	5	75	67,500	648,000

) emerged at distinct monitoring densities: 28 points for minimal operations (30 days), 14 points for seasonal deployment (60 days), 10 points for quarterly operations (90 days), and 5 points for semi-continuous monitoring (180 days).

3.5.2. Regional Suitability Classification

Analysis of location-specific requirements (

Table 8. Location classification by minimum required operational intensity for economic viability.

Location Category	Count	30 Days	60 Days	90 Days	180 Days	Not Viable
Major Cities (>1000 points)	15	15	0	0	0	0
Regional Centers (500–1000)	32	8	19	5	0	0
Medium Towns (200–500)	68	0	14	37	17	0
Small Towns (100–200)	72	0	3	24	42	3
Minor Locations (<100)	68	0	0	2	31	35
Total	255	23	36	68	90	38

) revealed four distinct operational categories based on economic optimization:

• Category 1: Minimal Operations (30 days annually) Suitable for 23 locations (9.0%) with very high monitoring density (>28 points/cell). Limited to major urban cores: Tartu city center (31.9 points/cell), Mustamäe (30.0 points/cell), and Kristiine (27.6 points/cell). These locations achieve positive ROI within 3–4 months despite minimal utilization.
• Category 2: Seasonal Operations (60 days annually) Encompasses 86 locations (33.7%) with moderate–high density (14–28 points/cell). Includes most Tallinn districts, regional cities (Pärnu, Narva, Viljandi), and larger towns. ROI period extends to 6–8 months with seasonal deployment aligned to Estonian summer conditions (May–September).
• Category 3: Quarterly Operations (90 days annually) Covers 134 locations (52.5%) with moderate density (10–14 points/cell). Represents the optimal balance for most Estonian municipalities, achieving a 12-month ROI while maintaining operational flexibility for weather variations.
• Category 4: Semi-Continuous Operations (180 days annually) Required for 208 locations (81.6%), including all areas exceeding 5 points/cell. This intensity enables drone deployment even in lower-density small towns and suburban areas, though ROI extends to 18–24 months.

3.5.3. Cost-Effectiveness Surface Analysis

Three-dimensional analysis of the cost–benefit relationship revealed non-linear interactions between operational intensity and monitoring density. Maximum savings occurred at the intersection of high density (>40 points/cell) and moderate operations (90 days), achieving annual savings exceeding €5000 per location.

The cost-effectiveness surface identified three distinct regions:

• High-Efficiency Zone: Monitoring density > 20 points with 60 + operational days, achieving > 75% cost reduction
• Moderate-Efficiency Zone: Density 10–20 points with 90+ days, achieving 40–75% reduction
• Low-Efficiency Zone: Density < 10 points requiring 180+ days, achieving < 40% reduction

3.5.4. Fleet Optimization Under Variable Intensity

Fleet requirements varied significantly with operational intensity as presented in Table 7, depending on operational model.

3.5.5. Sensitivity to Utilization Rates

Monte Carlo analysis (n = 10,000) evaluated economic robustness across utilization scenarios, as presented in

Table 9. Probability of positive ROI within 24 months by operational intensity.

Monitoring Density	30 Days	60 Days	90 Days	180 Days
Very Low (5–10 points)	2.3%	18.7%	41.2%	78.4%
Low (11–15 points)	14.6%	52.3%	76.8%	94.7%
Medium (16–20 points)	38.9%	74.5%	89.3%	98.9%
High (21–30 points)	71.2%	91.8%	97.6%	99.8%
Very High (>30 points)	94.5%	99.2%	99.9%	100.0%

.

The analysis revealed critical utilization thresholds: locations with <15 points/cell require a minimum of 90 operational days for reliable positive returns, while high-density locations (>20 points/cell) achieve viability even with a minimal 30-day deployment. The distribution of locations is shown in Figure 7.

3.5.6. Regional Implementation Recommendations

Based on operational intensity analysis, implementation should follow a tiered approach:

• Tier 1 (Immediate Deployment): 23 major urban centers with minimal operational requirements. Single 5G node rental for May–June achieves full cost recovery within the first year.
• Tier 2 (Year 2 Expansion): Additional 63 locations requiring 60–90 day operations. Two additional nodes for April–September deployment, targeting regional centers and medium towns.
• Tier 3 (Full Coverage): Remaining viable locations through semi-continuous operations. Requires year-round equipment availability but serves 81.6% of all Estonian urban areas.

The operational intensity analysis demonstrates that drone deployment viability extends well beyond initial estimates when utilization rates increase. Even modest 90-day annual operations enable cost-effective monitoring for 52.5% of Estonian urban locations, while semi-continuous operations achieve 81.6% coverage with positive economic returns.

4. Discussion

4.1. Economic Viability and Deployment Thresholds

The identification of an eight-point break-even threshold for drone deployment represents an improvement over the previously reported requirement. While earlier studies suggested minimum thresholds of 50–100 monitoring points for economic viability [8,13,63], the Estonian context demonstrates feasibility at significantly lower densities. This difference primarily stems from three factors: the reduction in drone costs, the rental-based model for 5G infrastructure eliminating capital expenditure, and the aggregation benefits achieved through hexagonal clustering approaches.

The break-even dynamics align with theoretical predictions from network economics, where fixed cost distribution across operational days creates inverse relationships between utilization and unit costs. Our finding that per-point costs decrease from €467 at 30 operational days to €78 at 180 days validates the importance of utilization optimization in deployment strategies. This relationship suggests that even sparse road networks become economically viable for drone monitoring when equipment utilization approaches semi-continuous operations.

Monitoring point density varies substantially across Estonian urban areas (0–71 points per hexagonal cell), with the highest concentrations in Tallinn districts (Mustamäe: 30.0 points/cell, Kristiine: 27.6 points/cell). While 90.3% of urban cells contain monitoring requirements, only 38.9% exceed the threshold for immediate drone deployment, suggesting a hybrid implementation strategy. This aligns with international experiences where selective deployment at complex intersections, while maintaining traditional counts for simple segments, achieves optimal cost reductions [61,64].

4.2. Operational Intensity as Deployment Determinant

The operational intensity analysis reveals a previously underexplored dimension of drone deployment economics. The progressive expansion of viable locations from 9.0% at minimal operations (30 days) to 81.6% at semi-continuous deployment (180 days) demonstrates that equipment utilization, rather than absolute monitoring complexity, often determines economic feasibility. This finding challenges the conventional assumption that drone deployment requires high-complexity environments, suggesting instead that consistent utilization enables viable deployment even in moderate-density settings.

The seasonal deployment model, particularly relevant for Estonian climate conditions, provides a practical framework for phased implementation. The concentration of counting activities during favorable weather months (May–September) aligns with international best practices while acknowledging operational constraints.

4.3. Limitations and Implementation Challenges

Several limitations constrain the generalizability of our findings. First, the economic analysis assumes Estonian labor costs and regulatory requirements, which may not translate directly to other contexts. Countries with higher pilot certification costs or stricter aviation regulations may experience less favorable economics. Second, the weather dependency of drone operations, while quantified in our sensitivity analysis, may prove more restrictive in practice, particularly for locations requiring year-round monitoring capability.

Integration challenges, particularly coordination of multiple drones within limited airspace and data stream management, represent operational rather than technological constraints, with regulatory frameworks being the primary barrier to scaling beyond current 15-drone operations [62]. Privacy concerns, though addressed through technical measures, remain potential barriers to public acceptance. GDPR compliance templates exist [60], but social acceptance varies across cultural contexts and requires location-specific engagement strategies.

4.4. Implications for Transportation Planning

The economic viability of drone-based monitoring at relatively low-density thresholds fundamentally alters the cost–benefit calculations for traffic data collection. Municipalities previously unable to justify comprehensive counting programs due to budget constraints can now achieve network-wide coverage through strategic drone deployment. This democratization of traffic data access particularly benefits smaller cities and rural municipalities, potentially reducing the urban–rural divide in transportation planning capabilities.

The shift from point-based to area-based monitoring enabled by drone platforms provides unprecedented insights into traffic flow patterns, pedestrian–vehicle interactions, and network dynamics. Traditional tube counts capture volume at specific locations but miss turning movements, lane changes, and conflict points—all readily observable from aerial perspectives. This richer data enables more sophisticated modeling and evidence-based infrastructure investments.

4.5. Future Research Directions

Several research priorities emerge from our analysis. The development of autonomous perching capabilities, while technically demonstrated in laboratory settings [36,37,38,39,40,41,42,43,44,45], requires field validation in operational environments.

Machine learning applications for automated pattern recognition from aerial footage remain underdeveloped relative to ground-based computer vision systems. The unique perspective and lighting conditions of aerial monitoring require specialized training datasets and algorithm optimization. Transfer learning from existing ground-based models provides a starting point, but dedicated research into aerial-specific detection algorithms could improve accuracy and reduce processing requirements.

The potential for real-time traffic management based on drone observations, while beyond our current scope, represents a natural evolution of technology. Integration with adaptive signal control systems, dynamic routing applications, and incident management platforms could transform static monitoring into active management tools. This transition requires not only technological development but also regulatory frameworks for real-time intervention based on automated observations.

5. Conclusions

This study establishes the economic feasibility of drone-based traffic monitoring systems for urban environments through a comprehensive spatial and economic analysis of 255 Estonian urban locations. The analysis demonstrates that drone-based systems achieve economic viability at substantially lower monitoring densities than previously reported. The identified break-even threshold of eight monitoring points per hexagonal cell represents a five-fold reduction compared to the requirements reported in earlier studies. This finding expands the applicability of aerial monitoring systems to 67% of Estonian urban areas, including smaller municipalities previously excluded from cost-effective comprehensive traffic monitoring.

Operational intensity emerges as the critical determinant of economic feasibility. Increasing the annual deployment from minimal operations (30 days) to semi-continuous monitoring (180 days) expands viable coverage from 9.0% to 81.6% of urban locations. This relationship alters deployment strategy recommendations, suggesting that consistent utilization enables viable implementation even in moderate-density environments where traditional cost structures would prove prohibitive.

The integration of nomadic 5G infrastructure with ultralight drone platforms creates multiplicative rather than additive benefits. The rental-based infrastructure model eliminates capital expenditure barriers while enabling simultaneous multi-point monitoring, achieving documented cost reductions of 60–80% compared to traditional manual counting methods while maintaining superior accuracy rates (95–100% vs. 85–95%).

The findings provide actionable frameworks for transportation agencies. The tiered implementation strategy—progressing from 23 high-density locations requiring minimal operations to comprehensive coverage of 208 locations through semi-continuous deployment—offers a pragmatic pathway for phased adoption aligned with budgetary and operational constraints.

Limitations include the study’s focus on Estonian urban contexts, weather-dependent operational windows, and reliance on procurement data for cost parameters. Future research should validate findings across diverse geographic and regulatory contexts, develop autonomous perching capabilities to extend mission durations, and integrate real-time traffic management applications beyond descriptive monitoring.

The convergence of low-cost aerial platforms, mobile network infrastructure, and advanced computer vision represents a paradigm shift in traffic data collection. This research demonstrates that the technological maturity, proven economic advantages, and scalable deployment frameworks position drone-based systems as viable alternatives to traditional ground-based monitoring for most urban environments.