Directed energy weapons (DEWs), such as high power microwave (HPM) systems, have received substantial press attention over the last year. Videos in which a whole swarm of drones fall out of the sky under HPM attack have been very impressive, seemingly offering a solution to swarming tactics that seem to be difficult to counter in other ways.

However, countermeasures are developed against all weapons sooner or later. The question of how swarm (or individual) drones may counter HPM attack (beyond straightforward electromagnetic hardening) has not been addressed in the literature.

Will the advantages of HPM in countering drone swarms survive simple drone countermeasures? How will those countermeasures relate to the unique properties of HPM weapons?

Our research aim in this paper is to consider for the first time what drone swarms can do to protect themselves by modifying their swarming behaviour. We shall see that HPM weapons (unlike laser or kinetic weapons) have a particular weakness that is a direct consequence of their strength: their ability to impact simultaneously a large area means that they cannot avoid alerting an even wider area to their presence.

Drones or UAVs have become very prominent in the media and on the battlefield. Small uncrewed systems and loitering munitions have moved from niche to ubiquitous, with the Russia–Ukraine war accelerating adoption across regular and irregular forces and normalising first person view (FPV) drones as improvised loitering munitions (Brenner and Ferran, 2024; Hvizda et al., 2025; International Institute for Strategic Studies [IISS], 2024; Watling et al., 2023). Authoritative surveys define loitering munitions as expendable assets combining intelligence, surveillance, and reconnaissance (ISR) and terminal attack, distinct from multi-rotor UAVs that drop ordnance (Brenner and Ferran, 2024), while current policy and strategy analyses emphasise rapid innovation cycles, mass, and electromagnetic warfare (EW) susceptibility observed at scale in Ukraine (Bondar, 2025; Hvizda et al., 2025). Procurement signals echo these trends (e.g. Germany’s move to acquire loitering munitions), and global actors are openly pursuing mass production of “suicide drones.” Across these sources, a consistent picture emerges: low unit cost and operator-in-the-loop precision drive proliferation; survivability depends on the counter-uncrewed arial systems (C-UAS) environment (things such as EW, kinetic weapons, and DEWs). Despite the rapid progress of artificial intelligence (AI), autonomy beyond short-horizon waypointing remains constrained, with human oversight still central (Bondar, 2025; International Institute for Strategic Studies, 2024).

Military interest in UAV swarms stems from their potential to overwhelm traditional air-defence decision loops by exploiting simultaneity, dispersion, and redundancy. Swarming emphasises many-on-few dynamics in which numerous, relatively simple agents impose disproportionate cost on defenders by saturating sensors, command-and-control (C2), and interceptors while preserving mission outcomes despite attrition (Arquilla and Ronfeldt, 2000). From an engineering perspective, swarms leverage local rules (collision avoidance, goal seeking, and limited neighbourhood awareness) to produce robust collective behaviour without a single point of failure; decades of work on distributed flocking and collective motion provides a theoretical basis for such emergent coordination even with minimal or intermittent communications (Couzin et al., 2002; Reynolds, 1987).

Within counter-UAS, HPM DEWs are maturing from decades of NATO/US Department of Defense research on coupling, susceptibility, and hardening into fieldable systems that promise one-to-many effects against swarms (NATO RTO, 2003; NATO Science and Technology Organization (STO), 2010; Reding et al., 2023). Recent demonstrations include Air Force Research Laboratory’s (AFRL) THOR engaging multi-drone swarms and industry systems (e.g. Epirus “Leonidas”) disabling dozens of UAVs in a single activation, underscoring the potential of wide-area electromagnetic effects versus laser “one-by-one” engagements (AFRL, 2023; Epirus Inc., 2025). At the same time, government assessments note enduring transition challenges—power/aperture scaling, repeatability and characterisation of lethality/safe severities, fratricide/electromagnetic compatibility (EMC), logistics, and Concept-of-Operations (CONOPS) integration—alongside policy and acquisition hurdles that can slow operationalisation despite increasing investment. For small UAVs specifically, the literature highlights vulnerable coupling paths (power electronics, flight-control signal paths, and radio-frequency [RF] front-ends), the importance of platform hardening, and the prospect that HPM becomes a central layer in layered C-UAS architectures where cost-per-shot and swarm coverage dominate (NATO Science and Technology Organization (STO), 2010).

The United Kingdom is developing a sovereign RF DEW (RF-DEW) capability for counter-UAS missions. The current demonstrator, led by a Thales, UK-headed industry consortium under the Ministry of Defence’s “Team HERSA” initiative, has undergone live trials with the British army. Public reporting indicates effective engagement of small UAVs at ranges of approximately 1 km, with near-instant effects and a cost-per-shot on the order of pence, highlighting its potential as a low-cost complement to the existing short-range air defence systems. The demonstrator will not enter service in its current form; instead, it is intended to inform future RF-DEW requirements, doctrine, and procurement pathways as part of a layered UK air-defence architecture integrating kinetic, electronic-warfare, laser, and RF-based effectors.

China has demonstrated rapid progress in the development and fielding of HPM systems tailored for counter-UAS roles. These systems emphasise mobility, rapid engagement of small UAS, and integration within multi-layered air-defence architectures. The FK-4000, developed by the China Aerospace Science and Technology Corporation (CASC), exemplifies a long-range mobile HPM system. Public reports indicate an effective range on the order of 3 km with the ability to disable light drones in less than one second. Defence major Norinco’s Hurricane-3000 is optimised for drone-swarm engagement and sustained operations. It reportedly engages effectively within 3km, with radar/sensor detection to ~6 km and optical tracking to ~4 km.

Beyond flagship systems, at least three additional vehicle-mounted HPM variants have been shown publicly in China, including integrations on 8x8 light armoured vehicles and heavy trucks (e.g. Shacman SX2400/2500-series) with mast-mounted surveillance radars and articulating microwave emitters. These illustrate a trend towards a family of HPM systems covering different mobility, power, and sensor-integration trade-offs—supporting roles from fixed-site defence and convoy protection to rapidly re-deployable tactical missions.

This section summarises the physical and engineering reasons why modern solid-state HPM architectures tend to operate with longer pulse envelopes than classic 1950s tube-based devices, and why those longer pulses are advantageous when the mission is to disable large numbers of small UASs (swarms). There is little information in the public domain about pulse lengths used by the HPM weapons described above, but their counter-swarm rationale, the use of solid-state electronics and phased-arrays, suggests a trend towards microsecond and longer pulse widths.

Tube-based pulsed-power HPM sources (vircators, relativistic magnetrons, Marx-driven magnetrons, etc.) produce extremely high instantaneous power over very short durations (typical pulse widths in the nanosecond to microsecond range) with very low duty cycles and therefore low average power. By contrast, solid-state power amplifiers such as gallium nitride–laterally diffused metal oxide semiconductor (GaN-LDMOS) phased arrays, trade peak instantaneous power for the ability to sustain high average power and high repetition rate. Practically, semiconductor devices are constrained by device breakdown, thermal limits, and drive/capacitor technology: they are far more capable of delivering microsecond to millisecond envelopes at substantial average power than of producing single-shot gigawatt (GW)-class nanosecond spikes.

Phased-array architectures rely on coherent control of many transmitter/receiver modules. Effective electronic steering, nulling, and adaptive spatial shaping require a waveform with sufficient temporal extent to apply and maintain the desired phase relationships across the aperture, to track moving targets and to sweep or dwell over a volume. Ultra-short isolated spikes provide little time for adaptive beam manipulation; longer pulses (or trains of pulses) enable continuous steering, adaptive nulling, and quasi-continuous-wave (CW) illumination that better support area coverage. Hence, there is a general trend towards longer (1–10 µs), steerable pulses in HPM hardware.

The vulnerability mechanisms for small UAVs include: upset of digital electronics, collapse or disturbance of power rails and regulators, failure or desynchronisation of radio/telemetry links, and thermal or cumulative stress on power-electronics. Many of these mechanisms are cumulative or time-dependent: sustained or repeated energy deposition is more likely to produce widespread malfunction across many heterogeneous platforms than isolated sub-nanosecond strikes, unless those strikes reach exceptionally high peak fields. Consequently, longer pulses increase the probability of producing functional degradation across an ensemble of devices. From an engineering-logistics perspective, long pulses yielding high average power are the natural operating mode for fieldable mobile systems.

The open literature already documents practical, fieldable concepts for detecting HPM events with modest hardware. Adami et al. (2014a) describe both the limits of early broadband diode detectors and a demonstrator system based on modern logarithmic detector integrated circuits (ICs) and broadband spiral antennas with >60dB dynamic range, capable of extracting pulse amplitude, width, repetition frequency, and count across roughly 0.5GHz to 8GHz, with robustness tests reported to ~kV/m field levels (Adami et al., 2011). Their survey also records multiple off-the-shelf or prototype detectors (e.g. QinetiQ Canary, Market Central “microwave microphone,” and a Netherlands Organisation for Applied Scientific Research developed unit) and publishes indicative threshold bands: on the order of ~100V/m for continuous wave (CW) upsets and ~10³V/m for pulsed/multi-cycle events, with steeper single-pulse transients in the kV/m range (Adami et al., 2011).

A follow-on chapter by the same group, “HPM detector with extended detection features” (Adami et al., 2014a), details a multi-channel approach (four polarisation-independent antennas and logarithmic detector chains) that augments simple thresholding with coarse direction finding and pulse parameterisation, further underscoring that HPM engagements are electronically conspicuous even to relatively austere sensors. Complementing these engineering demonstrations, a recent review in IEEE Transactions on Instrumentation and Measurement synthesises HPM classification, sources, and detection techniques (including operation in strong electromagnetic [EM] environments), reinforcing that robust, low-SWaP HPM detection is technically routine, rather than speculative (Jia et al., 2024).

Taken together, these sources support the central premise of this paper: HPM use against one UAV typically produces signatures detectable by others nearby, enabling threshold-based warning and autonomous dispersion with minimal sensor sophistication.

In summary, the detection problem considered here is not precision HPM metrology but threshold warning. The relevant question is therefore whether a low-mass detector can trigger before destructive field levels are reached, and whether the resulting warning interval is tactically exploitable.

We begin to model the effect of HPM weapons on UAV systems. This requires some routine consideration of beam geometry and potential power and field delivered by such systems. This is necessarily an approximate, given the variety of systems discussed above and the fact that many of their parameters and performance characteristics are not in the public domain. The model is therefore best interpreted as a sensitivity framework, rather than as a prediction for any specific classified HPM system: the qualitative conclusion is robust when the detector trip field is below the destructive field and when the interval between vanguard engagement and exposure of main body permits non-negligible lateral dispersion.

The analytical model used in this paper is intended as a deliberately simple, transparent framework for exploring the tactical consequences of HPM detectability, rather than as a high-fidelity prediction for any specific classified weapon system. The following assumptions define the scope of the model and clarify how the illustrative parameter values should be interpreted. They also identify the main limitations and the variables most suitable for later sensitivity analysis or hardware-in-the-loop validation.

This framing should be read as a sensitivity framework: the central qualitative result is strongest where the detector trip threshold is materially below the destructive field threshold and where the warning interval permits non-negligible lateral dispersion before the main body enters the lethal engagement volume.

For a circular aperture, the half-power beam width (HPBW) in degrees may be approximated by

vwhere D is the effective aperture diameter and λ is the wavelength. For a rotationally symmetric beam of full-angle HPBW θ, the solid angle is

and the corresponding antenna gain (assuming high efficiency) is approximately

Common reference values (circular beam) are set out in Table 2.

In typical use cases: narrow beams (<10°) maximise field intensity at range (hard-kill, single-UAV engagement), while wide beams (30°–60°) enable “blanket” effects suited to swarm defeat at reduced peak field strength.

Assuming far-field conditions and a fixed beam divergence (solid angle Ω), the on-axis (power) intensity and electric field scale with range r are as follows. The radiated power spreads over a spherical sector of area Ωr², giving an approximate power flux density,

where P_tx is the transmitted RF power. Thus, the intensity falls as 1/r².

The electric field strength is related to the intensity by

where η₀ ≈ 377 Ω is the impedance of free space. Substituting this into the expression above yields

so the peak electric field strength falls approximately as 1/r.

In practical terms, this means that while total energy delivered per unit area drops rapidly with distance, the electric field (which governs induced voltages and the likelihood of electronic upset or breakdown) decays more slowly. A wider beam (larger Ω) reduces on-axis field strength at a given range, whereas narrowing the beam (smaller Ω) increases it. These relations assume a constant beam divergence; near-field effects, side-lobes, and electronically steered changes in Ω can introduce deviations.

For the purposes of this analysis, we consider the simplest possible HPM detector. A very simple and lightweight method of detecting an HPM strike is to fit UAV with a miniature gas-discharge (e.g. neon) indicator lamp connected, so that its leads act as a small dipole antenna. When exposed to a sufficiently strong microwave electric field, the gas within the lamp undergoes electrical breakdown, and atoms ionise, recombine with electrons and emit a brief flash, which can be detected optically onboard. Indeed, it is easily detected through electromagnetic screening, such as a copper mesh that may be used to “harden” the main body of UAV to HPM attack. This simple gas-discharge device provides a low-cost threshold sensor for HPM exposure.

For a small neon bulb with a strike voltage V_th of around 100V, the effective electric field threshold depends on dipole length (formed by lamp leads). Typical order-of-magnitude values are as follows:

In the radiating far field, assuming a fixed beam divergence (solid angle Ω), the on-axis electric field of the HPM pulse scales as E(r) ∝r. The approximate warning range, r_trip at which the detector triggers is as follows:

This shows that halving the electric-field threshold doubles the warning distance, and that narrower beams (smaller Ω) reduce the time available for evasive action.

As an illustration, for a plausible peak effective isotropic radiated power, P_EIRP, in the range;

P_EIRP ~ 10¹⁰ – 10¹² W

for modern HPM counter-UAS systems (US Government Accountability Office (GAO), 2023), a 6-cm neon-dipole detector may trigger at ranges of order of hundreds of metres to multiple kilometres. Shorter dipoles (e.g. 1.5 cm), while compact, raise the trip threshold and therefore reduce warning distance. Pulsed HPM also increases the likelihood of ionization, because repeated peaks provide multiple opportunities for the strike voltage to be exceeded.

This form of detector is therefore most useful for early warning and evasive manoeuvre initiation, rather than for precise field measurement. Multiple lamps of different dipole lengths and orientations are employed to improve frequency and polarisation coverage.

We now compare this detection range with the UAV “kill” range.

Open sources do not provide a single universal “kill field” for small UAVs; instead, thresholds vary with frequency, pulse width, beam width, aspect, and coupling path (“front-door” via antennas vs. “back-door” via wiring and seams). Nonetheless, the literature and test practice indicate a consistent banding: (i) upset/loss-of-control can begin from a few hundred volts per metre into a low thousands of volts per metre regime for vulnerable configurations; and (ii) permanent damage/burnout to susceptible electronics is often discussed around 10 to 25 kV/m on-target fields for short pulses, depending on coupling (Chaari, 2021; Karcz et al., 2023).

For comparison, baseline radiated-immunity requirements for military and automotive electronics (not HPM, but indicative of design targets) are typically specified at about 200V/m CW over wide bands (with higher levels in certain radar sub-bands), meaning civilian commercial off-the-shelf (COTS) avionics and consumer-grade drones—often designed to much lower electromagnetic compatibility (EMC) levels—will generally show lower upset/damage thresholds than hardened platforms (International Organization for Standardization (ISO), 2021; US Department of Defense, 2015). Hardened small UAVs employing bonding, shielding, filtering, and partitioning can shift upset upward (and extend destruction distance), but are not invulnerable to multi-kV/m HPM environments (GAO, 2023).

Assuming far-field conditions and fixed beam divergence, the on-axis electric field scales as E(r) ∝ 1/r.

Using effective isotropic radiated power P_EIRP (or P_tx and solid angle Ω), the approximate trip/destruction range for a threshold E_th, is

Approximate, illustrative ranges: Using the equation above, representative thresholds, and peak P_EIRP give:

These bands are consistent with lower thresholds (shorter destruction distances) for unprotected “civilian” quadcopters; and higher thresholds (longer distances) for hardened military-class small UAVs. This is also consistent with programmatic reporting that HPM CUAS systems aim to deliver disruptive or destructive effects within sub-kilometre to kilometre-class ranges, depending on beam width and target set (GAO, 2023; Karcz et al., 2023).

A key question for survivability is whether a simple neon lamp-based detector provides sufficient early warning for a small UAV to take evasive action before experiencing disruptive or destructive HPM field strengths. Using peak effective isotropic radiated powers in the range P_EIRP ~ 10¹⁰ – 10¹² W, consistent with publicly discussed HPM counter-UAS concepts (GAO, 2023), the 6-cm neon detector would typically provide warning at distances from several hundred metres to multiple kilometres, depending on system parameters. The corresponding time margin for evasive action (e.g. rapid alteration of trajectory, attitude, or speed) would be on the order of several seconds to over one minute for small UAVs travelling at 30–50 m/s, subject to propagation effects, beam pointing, polarisation alignment, and whether the UAV is illuminated by main lobe or side lobe.

This analysis suggests that, in many engagements, a lightweight neon-based detector can provide meaningful early warning against HPM attack, particularly for wide-beam systems or initial detection of a scanning main lobe. Performance can be enhanced by using multiple orthogonally mounted lamps of differing lengths to reduce polarisation and frequency sensitivity, and by linking detection to a predefined automatic evasive-manoeuvre protocol.

Detection of an imminent HPM activation by a simple threshold sensor (e.g. a neon lamp dipole) does not itself harden a UAV platform, but it can materially change the attacker–defender engagement geometry by inducing an immediate, local, and autonomous avoidance response. Even in the absence of inter-vehicle communication or coordinated command, such individualised, randomised evasive manoeuvres rapidly reduce target areal density and thereby decrease the probability that a subsequent narrowly directed HPM pulse will illuminate multiple surviving vehicles. This phenomenon can be understood formally, but is intuitively captured by the “wasp swarm” analogy: Dispersing a compact cluster into many independent trajectories makes it significantly harder to neutralise the threat in subsequent attempts, even though each individual element remains vulnerable.

Consider a swarm of small UAVs approaching a defended sector. At range r, the HPM emitter subtends an irradiated footprint on the target plane of area

A(r) = Ωr²,

where Ω is the beam solid angle, as introduced in the section above. Let the initial mean areal density of UAVs in the swarm be λ₀ (vehicles per m²), typically a good deal less than unity. The number of vehicles illuminated at that instant is (approximately) Poisson-distributed with mean,

E[N₀] = λ₀A(r)

Suppose the HPM activation destroys each illuminated vehicle independently with probability p_kill. The expected number of immediate survivors within the illuminated footprint is therefore

E[S₀] = E[N₀](1 – p_kill)

Assume that each surviving vehicle executes an autonomous, randomised evasive manoeuvre immediately upon detection. We model this as isotropic lateral dispersion at effective sideways speed v_disp sustained for a delay Δt before any subsequent HPM pulse can be delivered. After this interval, the survivors occupy, on average, an area of the order

A_after ≈ π(R₀ +v_dispΔt)²,

where R₀ is the initial effective radius of the illuminated group (so that A(r) ≈ πR₀²). The mean areal density after dispersion, λ_after, is approximately

A subsequent narrow beam pulse at the same range again subtends area A(r), so the expected number of vehicles re-illuminated is λ_afterA(r). The ratio of the expected re-engagement count to the initial count is therefore

This expression highlights two multiplicative effects that both reduce follow-on engagement effectiveness: (i) incomplete initial kills (1 – p_kill), and (ii) geometric dilution of areal density because of dispersion. Two immediate implications follow:

First-shot effectiveness is critical. Unless p_kill is close to unity, a single narrowly focused HPM engagement is unlikely to remove the swarm as a threat. Any surviving vehicles disperse and become significantly harder to re-acquire within a narrow beam on subsequent pulses.

This directly parallels the “wasp swarm” analogy: dispersing a tight cluster of individually fragile agents transforms a single-direction defensive problem into a multi-vector one. Even without communication between UAVs, random autonomous dispersion significantly increases the defender’s resource and time requirements for neutralisation.

To counter a detect-and-disperse swarm effectively, a defender must consider one or more of the following:

Achieving sufficiently high p_kill on the first pulse through increased P_EIRP, multiple simultaneous beams, or optimised beam shaping.

The above model omits additional factors, such as near-field effects, ground reflections, polarisation mismatch (relevant for dipole-type detectors), and potential partial degradation of flight control following initial exposure. These effects should be incorporated in higher-fidelity simulation or hardware-in-the-loop trials.

To illustrate orders of magnitude, consider an initial compact group with effective radius R₀ = 20 m (so A(r) ≈ πR₀² ≈ 1,257 m²) and mean areal density λ₀ = 0.032 m^–2, yielding an expected initial count of forty UAVs

E[N₀] = λ₀πR₀² ≈ 40.

Assume that the first HPM pulse achieves per-vehicle kill probability p_kill = 0.6, leaving

E[S₀] = E[N₀](1 – p_kill) ≈ 16

immediate survivors within the footprint. Suppose each survivor executes an autonomous lateral evasion with effective sideways speed v_disp = 20 m/s and there is a Δt = 2 s interval before a possible follow-up pulse, giving a lateral offset v_disp Δt = 40 m. The post-dispersion effective area is then

A_after ≈ π(R₀ + v_dispΔt)² = π60² ≈ 11,310 m²,

and the resulting mean areal density becomes

If a subsequent narrow beam again subtends area A(r) ≈ πR₀² at the same range, the expected number re-illuminated on the second pulse is

E[N₁] = λ_afterA(r) ≈ (1.4 × 10^–3) × 1,257 ≈ 1.8,

viz. a reduction from ≈ 40 initially illuminated vehicles to fewer than two on the next pulse, even before considering any further time for additional dispersion. The governing factor is the dilution ratio

which here implies roughly a 95% reduction in expected re-engagement count between the first and second pulses. Modest changes in p_kill, v_disp, or Δt map directly to this ratio and can be used as design or doctrinal tuning parameters.

The “wasp swarm” analogy is a useful, intuitively accessible framing for dispersion dynamics. In collective animal motion, brief local rules (repulsion/avoidance, heading change, and speed adjustment) generate rapid spatial de-correlation and group reconfiguration without any global communication (Couzin et al., 2002; Reynolds, 1987). In military studies, swarming has long been analysed as a many-on-few paradigm in which distributed, loosely coupled agents exploit simultaneity, manoeuvre, and geometry rather than persistent communications (Arquilla and Ronfeldt, 2000). By analogy, threshold-triggered autonomous dispersion in UAV swarms—without inter-vehicle messaging—achieves the operational effect of transforming a dense, beam-vulnerable cluster into a sparse, multi-vector set of trajectories within a few seconds. This increases the defender’s resource burden for follow-on engagements (wider beams, multi-beam simultaneity, or cross-domain layering), consistent with the general doctrine that defeating swarms often requires early, broad-pattern effects or near-simultaneous multi-axis fires rather than sequential, narrow-beam re-attacks (Arquilla and Ronfeldt, 2000).

Much of the literature on UAV swarming focuses on inter-vehicle communications and distributed algorithms for formation, task allocation, and resilience, but these same RF links create obvious vulnerability vectors: exposed antennas, radios and power electronics provide coupling paths for jamming, spoofing, and directed-energy effects and significantly raise the platform’s EM signature compared with a strictly stand-alone loitering munition. Yet detector demonstrations from Fraunhofer, TNO, and others show that small, low-mass HPM warning/detection packages are technically plausible. We have shown that this would materially change the attacker–defender calculus, even without inter-UAV communication. This then raises the question of what is the best strategy for UAV deployment under such conditions. We can draw an analogy with the doctrine of “movement to contact (MTC).”

Movement to contact is an offensive task conducted to develop the situation and establish or regain contact with the enemy when the tactical picture is incomplete (Headquarters, Department of the Army, 2012, 2023). Its logic is to preserve the freedom of action of the main body by ensuring that first contact occurs with a smaller, purpose-organised element that can report, fix, and set conditions for decisive action by follow-on forces (Headquarters, Department of the US Army, 2023).

Doctrinal fundamentals of MTC include: (i) focusing all efforts on finding the enemy; (ii) making initial contact with the smallest force possible, consistent with protecting the force; (iii) using small, mobile, self-contained forward elements to avoid decisive engagement of the main body on terrain of the enemy’s choosing; (iv) task-organising and employing movement formations to deploy and attack rapidly in any direction; (v) keeping subordinate forces within supporting distance; and (vi) maintaining contact once gained (Headquarters, Department of the US Army, 2023). At small-unit level, movement techniques are explicitly designed to enable initial contact by the smallest feasible element so that leaders can establish a base of fire and maneuver the remainder without first disengaging (Headquarters, Department of the Army, 2024).

Movement to contact commonly uses a security framework of screen/guard/covering elements to provide early warning, develop the situation, and buy time/space for the main body to deploy. Variations such as search-and-attack or cordon-and-search adapt MTC principles to stability or dispersed enemy situations while retaining the core aim of controlled first contact and rapid development of the situation (Headquarters, Department of the Army, 2023, 2024). We observe that MTC is a useful principle to apply to UAV swarming in the presence of HPM weapons.

As we have seen, if an alert mechanism is in place in each UAV, the effectiveness of the HPM in countering a swarm becomes critically dependent on an effective “first shot” to disable most, if not all, of them. Figure 1 shows the schematic approach of a UAV swarm.

In this configuration, as shown in Figure 2, it’s possible for the HPM weapon to wait until most, or all, of the UAVs are close enough to be destroyed before initiating HPM pulses.

Thus, an appropriate strategy in the spirit of MTC for the attacker may be to deploy a vanguard UAV in advance of their main swarm, as shown in Figure 3.

The power density at the vanguard UAV is

where r₁ is the closest tolerable distance of approach.

This can warn the main UAV swarm if each vehicle has a microwave detector triggered at a threshold power density p₂, where

A defender cannot ignore vanguard threat without risk, but need not necessarily answer it with HPM. The vanguard concept is therefore strongest where the attacker can make non-HPM interception costly, saturated, delayed, or uncertain. Importantly, unlike kinetic weapons, the use of HPM cannot prevent some of the directed electromagnetic energy from also alerting the main UAV force to the microwave attack, provided that the spacing behind the vanguard UAV is small enough and the detectors on the following UAVs are sufficiently sensitive.

From the two relations above, we obtain

so that if p₂ is 1% of p₁, and the minimum distance the HPM (with its own safety in mind) engage the vanguard UAV is 10 m (say), then the optimum placement of the main UAV force is r₂ – r₁, or just <90 m behind the vanguard UAV. This would provide enough time and space for individuals within the main force to disperse before entering any zone where they will be destroyed. Random dispersal, at high speed, might be a sensible choice as soon as HPM is detected, as shown in Figure 4.

The concept of vanguard UAV places a forward, sacrificial element ahead of the main swarm to force an adversary’s HPM emitter to demonstrate its intent and geometry (cf. the movement-to-contact framing in the section of that title above). If the vanguard UAV compels the defender to fire in order to protect defended assets, the resulting emission will likely destroy the vanguard UAV but risks triggering threshold detectors carried by the follower UAVs; those detectors in turn can initiate autonomous evasive responses that rapidly dilute the areal density of the swarm (dispersion).

In summary, the vanguard concept converts HPM engagement from a hidden terminal event into an observable tactical signal. Its value is greatest where the defender must either reveal the HPM layer early or expend alternative defensive resources against cheap forward probes.

For concreteness, we adopt representative (publicly discussed and internally modelled) values: a neon lamp trip threshold for a ~6 cm dipole of

E₂ ≈ 3 kV/m

and an illustrative destructive threshold for small UAV electronics of

E₁ ≈ 15 kV/m

so that E_dest/E_M ≈ 5 (cf. GAO, 2023). Using the far-field effective isotropic radiated power (EIRP) formulation as defined above, the trip and destructive distances scale with inverse field and thus satisfy

independent of EIRP; numerical values, however, do depend on effective radiated power. Table 4 presents worked ranges and time margins for three example peak EIRP envelopes commonly used in HPM order-of-magnitude reasoning (10¹⁰, 10¹¹, and 10¹² W). The time margins shown assume radial closure speeds of v_uav = 30 m/s and 50 m/s, respectively.

These examples show that, for plausible EIRP envelopes, the neon (or other) detector can provide an order-of-magnitude spatial margin between detection and destructive field levels. Converting that margin to time yields conservative windows ranging from a few seconds to many tens of seconds depending chiefly on emitter power and platform speed. Such time margins are practically useful: even a short automatic lateral manoeuvre or a short-duration reduction in radar/coupling exposure can move a high proportion of the swarm out of the main lobe (see dispersion model below).

If the vanguard elicits a defensive HPM shot and the main force detectors trigger, survivors execute autonomous, randomised dispersion. The dilution of areal density and its impact on subsequent narrow-beam engagements is readily quantified. Using the notation introduced above, for an initial illuminated effective radius R₀, the area is A₀ = πR₀². The expected initial count is E[N₀] = λ₀A₀. If the first pulse achieves per-vehicle kill probability p_kill, the survivors E[S₀] = E[N₀](1 – p_kill) disperse to an effective radius R_after = R₀+ v_dispΔt after delay Δt, producing areal density,

Hence, the expected number re-illuminated by a subsequent narrow beam of the original footprint area is

A worked numerical example (consistent with the above parameters) further highlights the impact. Let R₀ = 20 m, initial expected count E[N₀] = 40, effective sideways dispersion speed v_disp = 20 m/s, and take the conservative time margin Δt from the P_EIRP = 10¹¹ W row in Table 4 for v_uav = 30 m/s (Δt ≈ 21.7 s), then for p_kill = 0.6, we obtain

R_after = 20 + (20)(21.7) ≈ 454 m,

and hence,

that is a lower than 50–50 chance of any UAV being illuminated on a subsequent narrow engagement. Even for shorter time margins and more modest lateral speeds, the dilution factor is large: dispersion converts tens of targets into a spatially sparse cloud that is expensive for the defender to re-engage with a narrow solid-angle beam.

Two strategic conclusions follow:

First-pulse lethality must commensurate with dispersion penalty. If the defender relies on a single, narrow beam, the required first-shot per-vehicle kill probability grows rapidly with initial swarm size and expected dispersion. For practical swarm sizes, a defender must either deliver near-unity p_kill across the footprint or accept that survivors will persist and disperse.

To operationalise the vanguard UAV approach and validate the numerical trade space, we recommend a combined modelling and test programme:

Monte Carlo simulation sweeps over P_EIRP, beam HPBW (Ω), E_M, E_dest, p_kill, v_uav, and v_disp to map survivor distributions and expected re-engagement counts (probabilistic mission-level risk).

Taken together, these results are likely to strengthen the strategic case for vanguard UAV employment when combined with threshold detectors: a modestly sensitive, low-mass detector provides a demonstrable temporal window that, when exploited by simple autonomous dispersion, materially increases the resource cost to the defender of neutralising the swarm. Conversely, defenders that rely on sequential, narrow-beam engagements will find that dispersion imposes prohibitive re-engagement costs unless they concentrate sufficient first-shot lethality or adopt wider/overlapping engagement geometries.

The baseline vanguard UAV model deliberately isolates the HPM-specific case in which the defender uses microwave engagement against the leading element. In practice, an adaptive defender may employ layered C-UAS fires, including kinetic weapons, laser DEWs, interceptor UAVs, electronic warfare, and HPM, and may attempt to defeat the vanguard without revealing the HPM emitter. This does not invalidate the detect-and-disperse mechanism, but it narrows the conditions under which a single vanguard is sufficient. Against such an adaptive defence, the attacker’s problem becomes one of forcing premature commitment of the HPM layer by presenting multiple low-cost probes, simultaneous axes, altitude-separated arrivals, or decoy-vanguard elements. In doctrinal terms, the vanguard should therefore be understood not only as a single sacrificial UAV but as the simplest member of a broader movement-to-contact family of probing elements. Modelling such multi-axis and layered defence interactions is beyond the scope of the present minimal analytical model, but follows naturally from the same engagement geometry and cost–exchange framework.

High-power microwave systems represent an attractive counter-swarm technology, offering the prospect of disabling multiple UAVs with a single engagement. However, the “wasp swarm” analogy is a useful way to visualise swarm dispersion dynamics and how it may frustrate the destruction of the swarm as a whole. This paper has examined the fundamental field thresholds relevant to small UAV vulnerability, analysed the feasibility of early HPM detection using a lightweight neon lamp threshold sensor, and quantified the tactical implications of autonomous evasive responses triggered by such detection. A simple but robust probabilistic dispersion model is developed, demonstrating that even brief, uncoordinated manoeuvres can rapidly dilute UAV areal density and drastically reduce the effectiveness of subsequent narrow-beam HPM engagements. Analytical and numerical studies further show that, unless near-unity per-vehicle kill probabilities are achieved on the first shot, dispersion imposes severe re-engagement penalties on the defender.

The “vanguard UAV” concept was introduced and formalised as a sacrificial forward probe, capable of eliciting the initial HPM emission and thereby granting the main swarm a tactically valuable time window for dispersion. Results indicate that, for representative detector thresholds (e.g. a 6-cm neon dipole at ~3 kV/m) and destructive field bands (~10–25kV/m), typical warning intervals range from several seconds to tens of seconds on-axis, depending on microwave power and UAV approach speed. These margins provide sufficient opportunity for dispersion to transform a compact, vulnerable swarm into a sparse, multi-vector threat, significantly increasing the defender’s resource burden for follow-on engagements. In operational terms, this shifts the defender’s requirement towards wide or overlapping beams, multi-emitter simultaneity, or layered effects, rather than repeated narrow-beam shots.

A central strategic insight emerging from this work is that, unlike laser-directed energy or kinetic C-UAS approaches that often neutralise targets without necessarily alerting others, HPM has an intrinsic self-revealing characteristic: any emission sufficient to destroy or disrupt one UAV is likely to be detectable by others in the vicinity. The neon lamp detector analysed here is intentionally elementary, and better alternatives have already been demonstrated, yet already offers meaningful early-warning capability. More sensitive, purpose-designed detectors can only strengthen this effect. That a trivial device of negligible mass, cost, and technical sophistication can materially alter the engagement economics suggests that HPM may not constitute the “silver bullet” against UAV swarms that it is sometimes portrayed to be. Rather, HPM employment against swarms is fundamentally a contest of first-shot effectiveness versus dispersion dynamics, with favourable outcomes relying on doctrine, timing, geometry, and combined effects, not HPM alone.

Future work should extend these models to include near-field effects, beam-steering dynamics, coordinated swarm behaviours, and adversarial learning. Hardware-in-the-loop testing and controlled flight trials are recommended to validate timing, survivability, and dispersion behaviours under realistic conditions. The findings herein highlight that low-cost sensing and autonomy can substantially degrade the operational value of HPM in swarm-defence scenarios, and that resilient swarming tactics remain within reach of even modestly equipped actors.