The U.S. Military Wants Drones To Fly Through Enemy Caves and Map Them

If you’ve ever tried to find your way through a dark hallway during a power outagephone flashlight in one hand, mild panic in the otheryou already understand
the vibe of subterranean warfare. Now swap “hallway” for “enemy cave complex,” remove GPS, add dust, dripping water, weird echoes, and the ever-present chance
of someone on the other side of a bend having very unfriendly opinions about your life choices.

That’s why the U.S. military has been pushing hard for tools that can go underground before humans doespecially small drones and robots that can slip
into caves and tunnels, scout ahead, and generate usable 3D maps fast. The goal isn’t sci-fi sightseeing. It’s reducing ambush risk, finding hidden branches,
spotting hazards, and answering the question every soldier would like answered in advance: “What’s around that corner, and how bad is it?”

Why caves and tunnels still matter (unfortunately)

Modern conflict isn’t only “open desert with clear lines of sight.” Adversaries have repeatedly used underground spaces to survive airpower, hide command posts,
store weapons, move personnel, and pop up where you least want surprises. Underground networks also show up in dense urban areasthink basements, subway
infrastructure, utility tunnels, and purpose-built fighting positions.

The U.S. Army has treated subterranean operations as a real readiness requirement, investing in training across brigade combat teams and developing handbooks and
tactics for fighting below the surface. Because once you go underground, a lot of the usual advantages of a modern forceoverwatch, sensors at range,
precision navigationget muted or outright canceled.

What the U.S. military is actually asking for

This isn’t just a movie-inspired daydream. There are concrete efforts aimed at cave and tunnel mapping:

1) “We need a portable way to map tunnelsfast.”

The Army’s Rapid Equipping Force has sought industry input on portable systems that soldiers can carry to a tunnel entrance and use to map the interior with
ground robots or drones. The key themes are speed, practicality, and reducing reliance on “flashlights, maps, and pencils” in dark passages where the margin
for error is tiny.

2) DARPA has been stress-testing underground autonomy at full difficulty

If you want a preview of where cave-mapping drones and robots are headed, DARPA’s Subterranean (SubT) Challenge is basically the robotics equivalent of
telling contestants, “Congratsyour new office is a cave. Also, your Wi-Fi is bad on purpose.”

SubT pushed teams to deploy autonomous systems (including aerial drones) to map, navigate, and search in three underground “flavors”: human-made tunnel
networks, urban underground spaces, and natural caves. Teams scored points by correctly identifying and locating “artifacts” like backpacks, cell phones,
radios, and even invisible hazards such as gasunder time pressure, with no GPS, and communications that get worse the deeper you go.

Why underground mapping is so hard for drones

Flying a drone outdoors is already a balancing act of batteries, wind, and “please don’t crash into that tree.” Underground flight adds a special collection of
problems that sound like they were designed by someone who dislikes both drones and joy:

• No GPS: Satellite signals don’t reach caves and tunnels. Your drone can’t “just navigate.” It has to figure out where it is.
• No light (or bad light): Cameras struggle without illumination, and dust can turn headlights into a snowstorm of particles.
• Degraded communications: Radio signals get blocked by rock, reinforced concrete, turns, and distance. Video feeds drop. Control links fade.
• Complex geometry: Branches, vertical shafts, tight squeezes, slopes, rubble, and unexpected obstacles are common.
• Harsh conditions: Moisture, mud, dust, and collisions are not rare eventsthey’re expected events.

GPS-denied navigation: the “Where am I?” problem

Underground drones rely on a toolbox of methods often grouped under SLAM (Simultaneous Localization and Mapping). In plain English: the drone
builds a map while using that map to estimate its own position. It can combine sensors like LiDAR, cameras, inertial measurement units (IMUs), and sometimes
radar or thermal imaging depending on the mission.

The trick is doing SLAM in conditions that bully sensors: low light, repeating textures, dust, and uneven surfaces. That’s why many advanced systems use
redundancyif vision gets noisy, LiDAR can still provide geometry; if LiDAR is degraded by particulate, the drone can lean more on inertial and visual cues.

Mapping that humans can actually use (not just a cool point cloud)

A raw 3D scan is impressive, but soldiers need actionable outputs: passable routes, likely choke points, branches, dead ends, vertical drops, and
potential hiding spots. That means turning sensor data into:

• 3D tunnel models (mesh or voxel maps) that can be rotated and zoomed
• Floor plans / cross-sections for quick decision-making
• Annotated hazards (gas, heat sources, blocked passages, unstable rubble)
• Confidence cues (what the drone mapped clearly vs. what it only partially observed)

In the SubT-style approach, “mapping” often includes identifying objects of interest and pinning their locations in the map. In combat terms, that could
translate to “possible firing position here,” “branching tunnel here,” “potential storage chamber here,” or “this shaft connects levels.”

Communications in caves: when your signal rage-quits

Underground radio behaves badly. One solution is to accept that the drone won’t be continuously piloted and instead make it more autonomous: it explores,
maps, and returns when it can. Another approach is to build a breadcrumb trail for communications, using:

• Relay nodes placed or dropped along the route
• Mesh networking so each device extends the network
• Heterogeneous teams (ground robots acting as comms anchors while drones move ahead)

DARPA’s SubT concept highlighted the value of relay-style approaches because underground comms are not a “maybe” problemthey are a guaranteed constraint.

Survivability: cave flight is basically controlled crashing

In tight subterranean spaces, minor collisions are normal. That’s why some cave-capable drones emphasize protective cages, ducted fans, or designs intended to
bounce and keep flying (within reason). The mission isn’t “perfect cinematic footage.” It’s “get the map, don’t get stuck, and preferably come back.”

What cave-mapping drones could change on the battlefield

A reliable tunnel reconnaissance drone doesn’t replace soldiers; it reshapes the tempo and safety of underground operations:

Faster decisions at the entrance

Instead of committing a team into an unknown maze, a unit could launch a drone, collect an initial map, and make smarter choices: which branch to clear first,
where to place security, where to expect a chokepoint, and whether the network is larger than anticipated.

Reduced risk from booby traps and ambushes

Underground spaces are ideal for trapstripwires, IEDs, blocked passages meant to funnel movement. A drone that can scout ahead, detect anomalies, and flag
hazards can reduce the odds of walking into a carefully prepared disaster.

Better integration with combined teams (air + ground robots)

The most promising direction is not “one heroic drone.” It’s teams of robots: small aerial drones for rapid scouting and larger ground robots
for endurance, payload, comms relays, or specialized sensing. SubT-style systems showed how heterogeneous teams can cover each other’s weaknesses.

It’s not just combat: disaster response is the friendly twin of this problem

The same technology that maps a tunnel network in a hostile environment can also map:

• collapsed buildings and underground parking structures
• mines after accidents
• subway tunnels after explosions or flooding
• industrial underground spaces with gas hazards

DARPA has consistently framed subterranean autonomy as beneficial for both warfighters and first respondersbecause “dangerous, dark, or deep” doesn’t care if
you’re wearing camouflage or a rescue helmet.

What still needs to improve (before this becomes routine)

Cave-mapping drones are real, but making them dependable enough for routine operational use requires progress in a few areas:

1) More reliable autonomy under messy conditions

Autonomy has to handle uncertainty: dust clouds, confusing geometry, degraded sensors, and sudden dead ends. The drone must decide when to push forward, when
to return, and how to avoid becoming expensive cave décor.

2) Better human interfaces (because nobody wants to read a 3D map while stressed)

The best map in the world is useless if it takes 12 minutes to interpret and your mission timeline is measured in seconds. Interfaces need to summarize:
“Here are the branches, here are the hazards, here are the likely routes, here’s what we don’t know.”

3) Smaller, cheaper, more disposable systems

In some missions, a drone should be treated like a flashlight battery: used aggressively, possibly lost, and replaced. That favors lower-cost platforms,
modular sensors, and the ability to launch quickly with minimal setup.

4) Countermeasures and electronic warfare reality

Any advantage becomes contested. Underground drones must assume jamming, spoofing attempts, and adversaries who learn quickly. Autonomy and mission design
have to account for “what if the link dies?” and “what if the environment is intentionally booby-trapped for robots?”

The big picture: why the U.S. military is chasing cave-mapping drones now

The U.S. military’s interest in drones that can fly through caves and map them is part of a broader shift: preparing for environments where traditional
strengths don’t automatically apply. Underground spaces punish GPS reliance, punish communications, punish visibility, and punish anyone who assumes the map is
accurate.

Programs like DARPA’s SubT Challenge demonstrated that autonomous aerial and ground robots can explore, map, and report meaningful information in caves and
tunnel systemsunder constraints that resemble real-world missions. Meanwhile, Army efforts have emphasized portable mapping tools and realistic training,
because subterranean operations are not an edge case. They’re a repeat guest star in modern conflict… the kind that never leaves when the party ends.

Field Diary: Experiences Related to Cave-Mapping Drones (Composite Training Scenarios)

Note: The following is a composite of commonly described training dynamics and operational-style constraintswritten as a scenario to illustrate what the
technology is meant to solve.

You’re standing at a cave mouth that looks like a shadow spilled into a hillside. Outside, everything is normal military reality: radios chatter, someone
adjusts kit straps, and a senior leader asks the question nobody loves“How far does it go?” Inside, reality changes. Light disappears after a few meters.
Sound gets weird. Distance becomes a guess. And if there’s an enemy down there, they’ve had time to decide where they want you to be uncomfortable.

In a training lane, the first drone goes in low and slow, props whining like it’s complaining about the assignment. The video feed looks fine for about ten
secondsthen the cave swallows the signal just enough to make everyone tense. That’s the moment you realize the real requirement isn’t “cool drone footage.”
It’s “keep flying and come back even when we can’t see you.”

The drone’s autonomy takes over. It starts building a map in the background, turning darkness into geometry: a bend here, a widening chamber there, a side
branch that drops slightly. Back at the entrance, the tablet display updates like a living sketch. The map isn’t perfectthere are fuzzy edges where dust
floated through the sensors or where the drone didn’t get a clean anglebut it’s enough to change the plan. The team can see that the first left branch dead
ends quickly (great, we can ignore that), while the right branch continues and splits again (less great, but at least it’s known).

Then the drone bumps somethinglight collision, the kind that would end a consumer quadcopter’s career. In the subterranean setup, collisions are expected,
not exceptional. The drone wobbles, corrects, and keeps moving. The group collectively exhales. That’s a weird thing to celebrate, but underground work is
basically a series of small victories over physics and anxiety.

Halfway in, the comms degrade further. In some scenarios, a ground robot follows as a relay, or the team drops small nodes like breadcrumbs to keep a thin
connection alive. In others, they accept silence and rely on the drone’s return. Either way, the experience teaches a brutal truth: underground is the one
place where “just call it in” is not guaranteed.

When the drone returns, it brings more than video. It brings decision speed. The map shows a constriction point that would force single-file movement, a
chamber that could conceal multiple people, and a vertical feature that suggests a second level. Instructors love that part, because it forces the team to
think like this is real: How do you secure a branching network? Where do you set security? How do you avoid funneling into a trap? The drone doesn’t solve the
mission. It changes the odds.

The most valuable “experience” isn’t the tech itselfit’s what it does to the team’s mindset. Instead of stepping into the unknown with only courage and a
flashlight, you step in with a rough map, a sense of scale, and fewer surprises. And underground, “fewer surprises” is basically the dream.

Conclusion

The push for drones that can fly into enemy caves and map them is about making one of the worst environments in warfare slightly less punishing. The U.S.
military’s interest spans practical Army needs (portable tunnel mapping) and high-end research (DARPA’s SubT-style autonomy) because the underground domain is
too importantand too dangerousto leave to guesswork.

The technology isn’t magic. It’s a hard-won mix of GPS-denied navigation, SLAM, resilient communications concepts, and rugged drone design. But every meter of
mapped tunnel is a meter that humans didn’t have to enter blind. And that’s the kind of advantage nobody argues with.