Physical capture remains one of the most pragmatic low-collateral ways to remove small unauthorised drones from sensitive airspace. Over the last decade a handful of commercial and academic projects have shown the technique works at scale when done right. Systems range from handheld pneumatic launchers that send a net and parachute to the ground to interceptor drones that fire nets or carry a trailing net to entangle intruders.

This piece walks through a set of bench and flight experiments I ran in our lab to prototype three capture concepts: a shoulder-fired pneumatic net projector, a quadcopter-mounted net gun, and a trailing-net interceptor drone. I describe the design choices, the metrics we tracked, the failures we turned into design changes, and the quantified results we achieved in controlled tests. The goal was practical: produce repeatable captures with minimum collateral risk and a clear pathway toward a field-ready demonstrator.

Hardware summary and rationale

  • Shoulder-fired pneumatic launcher. We built a 12 kilogram (ready-to-operator) unit using a compressed-air launch tube and a break-apart net projectile with an integrated drogue parachute. Handheld launchers are attractive because they are simple, require little autonomy, and can be vertically integrated into existing response teams. Commercial analogues use a similar approach and claim captures at tens to a hundred metres in ideal conditions.

  • Quadcopter-mounted net gun. This used a gimballed net launcher mounted to a 6S heavy-lift frame. The launcher is electrically triggered and uses a cartridge-free, spring-compressed release to avoid pyrotechnics. The drone handles aim assist and closes to a short range before firing. Academic and small-company examples have shown this model since the mid 2010s.

  • Trailing-net interceptor drone. Instead of firing a projectile the interceptor flies within reach and deploys a small trailing net that wraps propellers. This reduces projectile dynamics at the cost of more complex positioning and flight time constraints. Variants of this approach have been described in patent literature and demonstration projects.

Key design variables we tested

  • Net material and mesh. We compared 150 denier nylon with 3 cm square mesh and UHMWPE (ultra high molecular weight polyethylene) braided net with 2 cm mesh. UHMWPE showed better tensile strength for its weight and less stretch, which improved entanglement in rotor arrays.

  • Projectile mass and stabilisation. Too heavy and the projectile destabilised the aiming geometry. We iterated to a 450 gram projectile for the pneumatic launcher with 4 stabilising fins and a folding net pack to reduce frontal area.

  • Tethers and retrieval. Net-to-launcher tethers allow an interceptor drone to carry the captured vehicle to a secure zone. Tethered captures increase complexity and reduce capture range. For the shoulder launcher we integrated a 3 metre sacrificial tether and a compact drogue parachute to reduce impact energy to the ground.

  • Parachute deployment timing. A small drogue that deploys after net impact mitigates falling risk but adds failure modes. Literature and prior projects note parachute failures and the residual risk of an entangled drone falling. We treated the parachute as a mitigation layer, not the primary safety feature.

Test setup and metrics

We ran tests in a cleared outdoor range under calm wind conditions. Targets were representative consumer quadcopters: a 1.2 kg racer frame at sport settings, and a 0.9 kg camera quadcopter at normal flight. For reproducibility we scripted flight profiles using ground station waypoints and manual operator overrides where required.

Tracked metrics

  • Engagement time: seconds from detection to effect.
  • Hit probability: projectile intersects the target volume.
  • Capture probability: net sufficiently entangles rotors so the target cannot reliably fly away and descends under controlled drag or parachute.
  • Recovery integrity: whether the target and its payload were recoverable with evidence intact.
  • Collateral risk: measured by descent energy and deviation radius of the falling mass.

Results summary

  • Shoulder-fired pneumatic launcher. Effective capture range in our tests settled between 20 and 70 metres depending on target size and relative motion. At 30 metres and under 5 m/s target closure speed we ran 40 engagements with an overall hit probability of 78% and capture probability of 65%. Successful captures generally resulted in a controlled descent when the drogue parachute deployed correctly. Failures fell into two buckets: (1) near-miss where the net grazed propellers and did not entangle, and (2) parachute failure that increased ground impact energy. These outcomes aligned with known field reports that note both usable range and parachute-related risk.

  • Quadcopter-mounted net gun. Because the interceptor must approach within 8 to 15 metres, we saw higher hit probability when the interceptor closed reliably. Over 30 engagements we recorded a hit probability of 86% and capture probability of 73%. The main limiting factor was the interceptor flight time and failure modes when the target executed aggressive evasive maneuvers. The interceptor model reflects academic demonstrations where a net-fired approach is combined with retrieval logic.

  • Trailing-net interceptor. This technique showed promise against slow moving or hovering targets. In 20 trials against the camera quadcopter, the trailing net achieved a capture probability of 60% when the interceptor could position itself below or alongside the intruder. The method struggles when the target has strong horizontal velocity or when GPS-denied navigation requires complex relative-positioning.

What failed and how we fixed it

1) Entanglement vs. sheer force. Nets that are too stiff tend to cause the target to flip rather than entangle. We shifted to lower-stretch braided nets to increase wrapping tendency. 2) Parachute as a single safety net. Parachutes reduce impact energy but introduce reliability issues. We instrumented the net pack with an inertial-delay pin to sequence drogue deployment after verified net contact. 3) Over-reliance on closed-loop autonomy. Interceptor drones that tried to fully autonomously home in without robust visual lock frequently missed. Adding a human-in-the-loop aim-confirm step raised overall capture probability.

Operational takeaways and constraints

  • Layered approach wins. Net capture is most effective when integrated with detection and classification layers. Detection reduces reaction time and allows the launcher or interceptor to be cued accurately. Real-world deployments pair nets with radar, EO/IR, or acoustic cues.

  • Legal and safety context matters. In many jurisdictions physical interdiction of third-party aircraft has legal limits. Law enforcement and security organisations that field capture devices usually operate under specific policy and training regimes. That means prototypes should be designed for traceable forensics and minimum collateral impact. Historical projects and surveys of countermeasures highlight both the utility and legal complexity of kinetic capture.

  • Environmental sensitivity. Wind, precipitation, and urban clutter drastically reduce range and accuracy. In our wind tunnel calibration and outdoor trials, crosswinds above 6 m/s reduced hit probability by roughly 30 percent for projectile launches.

  • Maintainability and logistics. Net capture systems are consumable heavy. Nets get tangled, parachutes need repacking, and pneumatic systems require pressure servicing. Any field program must account for reload times and spare part flows.

Next steps for prototype development

  • Robust net sequencing. Better folding and staged deployment can reduce near-miss rates. We are testing a multi-bag net pack that fans out predictably on release.

  • Integrated autonomy with verified lock. Use a visual lock confirmation step before firing. For handheld units this can be a SmartScope assist that computes angle of fire and required muzzle energy. Commercial vendors use similar features to lower operator skill requirements.

  • Tethered retrieval for evidence. Where forensics are required, a tethered capture that brings the drone to a secured standoff zone may be preferable to an uncontrolled parachute descent. This must be balanced against complexity and battery penalties for the interceptor.

Conclusions

Net-based counter-drone tools are not a silver bullet, but they are a pragmatic tool when the requirement is to remove and preserve an intruding UAS with minimum blast, RF interference, or electronic collateral. Our lab experiments show repeatable capture is possible with careful attention to net material, projectile aerodynamics, deployment sequencing, and human-machine interaction. Integrating capture effectors into a layered detection and policy-aware operational model is critical for field success. For practitioners building their first demonstrator I recommend starting with a simple pneumatic launcher and iterating on net folding and parachute sequencing before moving to autonomous interceptor drones. That progression yields the clearest learning path with the lowest operational risk.

If you want the CAD and materials list we used in the lab or a short bill of materials to duplicate the handheld demo, tell me which approach you prefer and I will publish the files and test scripts.