Earth GuardPlanetary Defense Academy

Choose a program

Every mode uses the same rockets, engines, clock, reliability, and physics. The difference is what you are trying to build before time runs out.

Pick a threat

You are the mission director. Something is on a collision course with Earth. Work out what it is, decide how to move it, and launch before you run out of time.

Rocket garden

No mission. No pressure. Pick any rocket and watch it fly, with a real simulated flight path.

How this works

The important numbers are calculated from physics, but this is still a teaching model. The Help page separates measured inputs, derived results, and explicit planning assumptions so none of it is a black box.

Help & Physics › Inputs › Equations › Assumptions

Nothing important is hidden

Earth Guard has three kinds of numbers: published vehicle or planetary data, values calculated from equations, and planning assumptions used where no final vehicle or mission exists. This page labels all three. Scientist mode also prints the live equation with your mission's numbers on the screen where it is used.

Expeditions and campaign clock

How destination missions are calculated

r₁ = 1 AU aₜ = (r₁ + r₂) / 2 tH = π√(aₜ³/μ☉) v∞ = |√[μ☉(2/r₁ − 1/aₜ)] − √(μ☉/r₁)| C₃ = v∞²

The destination radius and the Sun's gravitational parameter are published inputs. The transfer time and C3 are derived. The Moon uses a separate 5.9 km/s surface-delivery budget from low Earth orbit.

Assumption: circular, coplanar Hohmann baseline. Real missions leave planets on eccentric, inclined orbits and often trade launch energy against gravity assists. Europa Clipper takes about 5.5 years using Mars and Earth assists even though an ideal Hohmann calculation is shorter. Uranus concepts study roughly 13 to 15 year cruises. The game shows the minimum-energy baseline and labels it as such.
Assumption: arrival delivery fractions. After Earth departure, the game reserves 18% for an orbiter, 62% to 70% for atmospheric entry systems, and 72% to 78% for powered landing systems depending on destination. These are transparent architecture-level mass allowances, not published payload guarantees for a vehicle that does not exist yet.
Clock rule. A turn advances to the next dated event, not a fixed number of days. Development, pad cadence, transfer time, and arrival all share the same clock. Launch windows repeat at the displayed synodic period.
Launch vehicles and recovery

What the rocket model includes

ṁ = F / (Isp · g₀) q = ½ρv² ρ = 1.225e^(−h/8500) g = μEarth / r² Δv = Isp · g₀ · ln(m₀/mf)

Stage wet mass, dry mass, thrust, specific impulse, diameter, and published low-orbit payload are inputs. The ascent is a 2D point-mass integration at 0.1 second steps with drag, falling gravity, centrifugal relief, propellant use, and a closed-loop gravity turn.

Assumption: ascent guidance. There is no wind, lift, slosh, structural flex, engine-out guidance, or detailed steering loss. It makes realistic-shaped trajectories, not flight-certified trajectories.
Assumption: recovery reserve. Downrange landing holds back 15% of first-stage propellant, return to pad 27%, tower catch 25%, and parachute 4%. The resulting lost ascent delta-v is paid from payload. Compatibility also checks stage mass, restart, throttle, landing hardware, and whether a catch tower was designed with the vehicle.
Assumption: failure timing. The displayed vehicle reliability decides whether a launch completes. If it does not, a second transparent planning model selects among ignition/pad systems, first-stage propulsion, Max Q structure, guidance and flight termination, stage separation, or upper-stage propulsion. Those conditional weights teach where failures can appear; they are not manufacturer statistics. An upper-stage loss can still leave a separated first stage free to complete its recovery; a three-core vehicle can also lose the center at staging after its side cores have departed.
Mishap consequence rule. Pad damage and return-to-flight reviews run in parallel, so the longer delay controls the campaign clock. A US launch uses an FAA return-to-flight determination; other sites use their range-safety authority. The exact day ranges are game planning assumptions. The real FAA requires a mishap plan, investigation, corrective action, and a public-safety finding before return to flight. FAA mishap and return-to-flight rules.

Liquid side cores burn harder than the throttled center core. They separate first; the center then continues alone. Every recoverable core rolls and records its own outcome.

Engine laboratory

Ideal nozzle physics plus labeled engineering limits

Aₑ/A* = f(Mₑ, γ) pₑ = p₀[1 + (γ−1)Mₑ²/2]^(−γ/(γ−1)) Vₑ = Mₑ√(γRTₑ) · ηcycle F = ṁVₑ + (pₑ−pₐ)Aₑ Isp = F/(ṁg₀)

The nozzle solver is derived from ideal-gas, choked-flow equations. Propellant temperature, ratio of specific heats, gas constant, and cycle efficiency are representative approximations. Cycle pressure and single-chamber thrust caps prevent obviously impossible combinations.

Assumption: development time and cost. NASA's TRL 1 through 9 scale provides the milestones, but the exact years and dollars are game planning estimates based on cycle complexity, thrust, pressure, and nozzle novelty. They are not NASA forecasts.
Landing rule. A developed engine supports powered landing only when the cycle can throttle deeply and the selected nozzle is not an oversized vacuum bell.
Threat observation and deflection

Why research and multiple attempts matter

M = ⁴⁄₃πr³ρ Δvkinetic = β · m · U / M miss ≈ 3 · Δv · lead time Eimpact = ½Mv²

A random threat begins as a point of light. Brightness cannot separate size from reflectivity, and a short observation arc does not fix the orbit. The game therefore blocks mission design until another instrument is chosen. Uncertainty in size, density, and momentum enhancement is rolled at intercept.

Assumption: rectilinear miss estimate. The factor-of-three along-track approximation is useful for teaching why early action wins. Operational planetary defense uses full orbit determination and n-body propagation.
Assumption: learning after contact. If at least one spacecraft reaches the target, the next turn tightens size uncertainty to ±8% and density uncertainty to ±15% and treats composition as characterized. This is a teaching proxy for encounter imaging, radiometric tracking, and measuring the orbit change. A launch campaign that delivers nothing teaches you about the rocket, not the asteroid, so it does not improve those target estimates.

When a first mission falls short but the window remains open, its measured orbital change carries into the next turn. The game shows impact only after the clock closes.

Primary sources
Program › Architecture › Launches › Surface
Expedition

Plan the next giant leap

1. Destination
Flight planEarth departure to destinationCommit a mission to begin flying the manifest.

2. What goes there

Build capability in order. A crew cannot arrive before power, communications, and a place to live. Robotic missions can fly alone.

3. Transportation
Recovery method
Mission timeline
Campaign flight log
Program › Engine laboratory › Flight test

Develop a rocket engine

Choose a propellant, feed cycle, chamber pressure, thrust class, and nozzle. The nozzle equations are ideal-gas physics; development time follows NASA technology-readiness steps. A new engine is a years-long campaign asset, not a last-minute defense trick.

Engine concept
Nozzle model

Rocket garden

Tap a rocket to see it, then launch it. All five are drawn to the same scale, so you can see how much bigger some of them really are.

Pick a rocket

Rocket designer

Two stages. Pick the engines, decide how much propellant each stage carries, and how fat the body is. Everything else — how tall it ends up, whether the engines physically fit, how much it can lift — falls out of the physics. It will tell you when a design cannot work, and why.

Drawn to scale. A Falcon 9 is shown beside it for comparison.

Looking up at the first-stage engines. This is the layout that decides how wide the rocket has to be.

Name
First stage

Second stage

Strap-on boosters

How the shape is derived
Someone gave you a code?
Building…
1 Brief › 2 Observe › 3 Design › 4 Launch › 5 Result

1 Brief › 2 Observe › 3 Design › 4 Launch › 5 Result

Observe the target

Right now we have a blurry survey detection and not much else. We do not know how big it is, what it is made of, or exactly where it will be. Observing costs time off the clock, and time is the thing that makes deflection easy.

These all point at the same rock at the same time, so they run together. What you pay is the slowest one you pick, not the total — which means adding a fast one next to a slow one costs you nothing.

Longest campaign
0days
Warning left
Size known to
Composition
Unknown
Why observing matters

Deflection depends on the target's mass (M = ⁴⁄₃πr³ρ) and its momentum enhancement factor β. Both come from size and composition. If you skip characterization, the game rolls your real values out of the uncertainty band you left yourself, so a mission that works on paper can still miss. This is a real problem: before DART flew, β for Dimorphos was known only to within a factor of several.

1 Brief › 2 Observe › 3 Design › 4 Launch › 5 Result

1. Pick your method

Different tools win in different situations. The clock usually decides.

Emergency powers

Everything below is normal-times rules: range safety, launch queues, months of pad turnaround. If the object is big enough, none of that survives contact with the news.

2. Pick your rocket

A rocket's famous payload number is to low Earth orbit. Escaping Earth costs another 3.2 km/s, and every bit of extra speed eats into the mass you can deliver.

Throw it away, or bring it home?

Launch site

Earth spins eastward at 465 m/s at the equator. Launch east from near the equator and you get that speed for free. Launch from high latitude, or south over open ocean, and you buy it out of your own tanks.

The "tonnes to orbit" number is not what you get to send. Saturn V is quoted at 140 t to low orbit, but it only ever sent about 48 t to the Moon — because that 140 t figure is mostly the S-IVB stage and the propellant still inside it. Getting to orbit and leaving Earth are two different jobs, and the second one is paid for out of the first one's answer. Watch what happens to every vehicle here as you raise C₃.
Departure model

One impulsive burn from a 200 km circular parking orbit (vcirc 7.784 km/s, vesc 11.007 km/s) using the vehicle's real upper stage, subtracting its dry mass. Assumption: ignores gravity losses on the departure burn, staging above the second stage, and launch-window geometry. Lands within roughly 20% of published payload-vs-C₃ curves.

3. Fly the trajectory

Two dials. Departure energy buys speed but costs mass. Lead time is how far ahead of the collision you hit it — the earlier you push, the more the miss distance grows on the way in.

Speed left over after escaping Earth: v = √C₃ =

Cruise needed: · Time you have:

Mass delivered
Closing speed U
β ejecta bonus
Asteroid Δv
Predicted miss distance

The orange line is Earth's radius plus a safety margin. The bar has to get past it.

The whole calculation
Optional

Send it to the politicians

Everything else on this page is physics, and physics does not have an opinion. Getting permission is the other half of the problem, and it is the half that is actually hard. This asks a real AI to play the National Security Council, the UN, and the public, and to decide whether your plan is allowed to fly — and how many days the arguing costs you.

Working it out…

Terminal count

All stations are go for launch.

T-plusT-00:08
Altitude0.0 km
Speed0 m/s
Downrange0 km
Vehicle
Payload
Flight model

A 2D point-mass gravity turn integrated at 0.1 s steps. Thrust and mass flow come from each stage's real propellant load, thrust and specific impulse; drag uses an exponential atmosphere (ρ = 1.225·e^(−h/8500)) with Cd 0.3 over the vehicle's frontal area; gravity falls off as 1/r² and the centrifugal term is included. Assumption: no steering losses beyond the pitch program, no throttling, no aerodynamic lift. It produces realistic-shaped ascents, not flight-certified ones.

1 Brief › 2 Observe › 3 Design › 4 Launch › 5 Result

Cruise

Coasting to intercept.

Range to target
Closing speed
Lead time

Impact

Energy
Crater
Severe blast
Earthquake

How worried should I actually be?

Short answer: not very. But the honest version is more interesting than either "we're all doomed" or "it'll never happen", so here it is properly.

How often things actually hit
1 m — a bright fireball
yearly
20 m — Chelyabinsk
~50 yr
50 m — Tunguska
500–1,000 yr
140 m — a region
~20,000 yr
1 km — global
100,000–500,000 yr
10 km — the dinosaurs
~100 M yr

Estimates differ between sources, which is why several of these are ranges rather than numbers. The small end comes from counting fireballs detected by satellites; the big end comes from counting craters and from how many objects the telescopes have found.

Two things make this much better than it sounds
  • The solar system got a lot emptier. Four billion years ago this was a shooting gallery. Most of the loose material has since been swept up, thrown out, or settled into stable orbits. What is left is the leftovers of the leftovers.
  • We have found nearly all the big ones. Over 90% of the near-Earth asteroids bigger than 1 km are catalogued, and their orbits are known well enough to say that none of them is going to hit us for centuries. The dinosaur-killer scenario is the one we have most thoroughly ruled out.
And three that keep it honest
  • The middle size is only partly mapped. Around 38% of the 140 m-and-up objects have been found. Congress asked NASA for 90% by 2020 and that target was missed, not through carelessness but because the objects are small and dark and there was never a telescope built for the job. NEO Surveyor is that telescope.
  • Comets arrive with almost no notice. They fall in from the outer solar system on long, steep orbits and often get discovered a year or two before they arrive, moving far faster than an asteroid. They are a small slice of the total risk, and they are the slice we can do least about. That is the mission in this game you probably could not win.
  • Orbits are not frozen. A close pass by a planet, or one asteroid nudging another out in the belt, can move something onto a path that was clear last time we looked. This is why the surveys keep running instead of finishing once.
For comparison: a big solar storm

In 1859 the Sun threw a cloud of plasma at Earth and set telegraph offices on fire. Estimates of how often that happens vary a lot — from roughly 1 in 200 per decade up to 1 in 8 per decade, depending on which statistical method you trust — so it is far more likely than a dangerous impact. It would also do far less harm: transformers, satellites and power grids, not craters. More probable, much less lethal. That is usually how risk works, and it is why "how likely" and "how bad" have to be held in your head at the same time.

So, the actual answer

In any given year the chance that something big enough to matter hits Earth is very small — smaller than almost anything else you already do not worry about. Over a hundred thousand years it becomes close to certain. Both of those are true at once, and neither is a reason to lie awake. It is a reason for a few hundred people to keep pointing telescopes at the sky, which is exactly what they do.

Sources: NASA CNEOS and the Planetary Defense Coordination Office for the survey numbers, the National Academies planetary defense report for impact frequencies, and the space-weather literature for solar storm rates, where the disagreement between studies is genuine and worth knowing about.

1 Brief › 2 Observe › 3 Design › 4 Launch › 5 Result

Flight debrief
    What happened vs. what you predicted