On 10 November 2025, Protolaunch — a UK propulsion company working out of the Westcott Space Cluster — announced it had been selected by the European Space Agency to design, develop and verify a multivariable control system for bi-propellant liquid rocket engines. Note what is being bought: not an engine, but the layer of sensing, estimation and valve commanding that sits between a throttle command and the hardware, and that decides whether an instruction for half thrust actually produces half thrust, stably, twice in a row.
Why a controller and not an engine
Protolaunch's own framing is a capability gap. It says European liquid rocket engines generally lack three things: deep throttlability, multiple restart functionality and extended operational life — the traits it ties to reusable launchers and to landers for autonomous planetary exploration. That is the company's characterization of the market, published in its own announcement, not an ESA assessment.
It is also why a control system is the near-term lever. Throttle range and restart behaviour are as much about how an engine is run as about how its chamber is designed. You can build a perfectly good engine and still be unable to land with it, because landing requires walking the operating point down and back up without losing mixture ratio, combustion stability or the thermal margin.
What closing the loop actually changes
The conventional approach is a schedule: for a commanded thrust level, set the main valves to positions found during testing. That is open loop — it assumes the engine, the feed system and the propellant conditions are the ones you characterized. The controller under this contract is described as closed loop and multivariable, dynamically monitoring chamber pressure, mixture ratio and overall engine health, rather than following a fixed schedule.
Two things get harder as you throttle down, and both argue for feedback.
First, the pressure drops across the injector and the valves shrink. Flow through a restriction scales with the square root of the pressure difference across it, so a fixed absolute pressure disturbance — regulator droop, a feed line oscillation, a bit of combustion roughness — produces a larger fractional flow error as that difference falls. Something that is a rounding error at full thrust can be a real mixture-ratio excursion at low thrust.
Second, chamber pressure stops being negligible compared with the injector pressure drop. At full thrust the chamber is a small back-pressure on a large driving head; throttled deep, the two are comparable, and the engine's own operating point starts feeding back into its own flow rates. That is precisely the regime where you want the controller to measure rather than assume.
The coupling is the whole problem
A bipropellant engine has two dominant actuators — the oxidizer valve and the fuel valve — and two dominant outputs: chamber pressure and mixture ratio. They are not independent. Opening either valve raises chamber pressure and raises that propellant's flow, shifting mixture ratio; trimming mixture ratio by moving one valve disturbs chamber pressure again. A pair of single-loop controllers chasing each other through that coupling will fight, overshoot, or settle slowly.
"Multivariable" is not decoration here. It means the controller carries a model of the cross-couplings and moves both valves together to hit both targets. Doing that reliably, without a test campaign's worth of hand-tuning at every throttle point, is the engineering content of the project.
The third variable, engine health, is the least defined. The announcement names it but does not say which signals constitute it — sensor plausibility, stability margins, thermal limits, or something else. That matters, because health monitoring is what would let a controller be trusted to run a descent burn without a human in the loop.
Two engines, one controller
The verification plan is the interesting part. Cross-engine compatibility is to be checked on Protolaunch's own 500 N engine, Arctic-FOx, and Airborne Engineering Ltd's 1,500 N Lambda — a factor of three in thrust class. Testing is split between Protolaunch's in-house E-Site facility and Airborne Engineering's propulsion test facility, both inside the Westcott cluster.
Controllers are usually tuned to a plant. Valve time constants, sensor ranges, sample rates and thermal lag all differ between a 500 N engine and a 1,500 N one. Proving one controller across that step without a full retune would say something real about the architecture — gain scheduling, normalization, model-based design — rather than about a single engine.
The caveat is scope. Both engines sit in one UK space cluster, so what is being tested is engine class, not geography, test-stand infrastructure or propellant supply chain. And the announcement does not say whether either engine is pressure-fed or pump-fed, which changes the control problem substantially.
There is heritage to build on: Protolaunch previously delivered cavitating venturi valves for an autonomous planetary lander application, and says it wants to extend throttling experience to cryogenic propulsion. A cavitating venturi is an elegant open-loop flow setter — it holds flow steady against downstream pressure changes, provided it stays cavitating, which requires enough upstream pressure relative to chamber pressure. Deep throttle is exactly where that margin gets thin.
The 24-month box, and what is not said
The work runs 24 months under ESA's General Support Technology Programme, funded through it. If the clock started around the announcement, completion lands in late 2027. No test results are reported, because this is work in progress.
Protolaunch also says it will consult industry stakeholders on control priorities and invites reusable-launcher and lander developers to submit theirs during development. Read plainly, that means requirements are still being set — reasonable for a technology programme, but it does mean the success criteria are not yet public.
Which leaves a long list of unquantified items: what "deep" means numerically, the minimum chamber pressure, the mixture-ratio accuracy band, response time and settling behaviour, how many hot-fire tests, and whether any of it touches cryogenics.
The bounded reading is this. The deliverable is a closed-loop multivariable controller regulating chamber pressure, mixture ratio and engine health, and the credibility test it sets itself is portability across two thrust classes at two facilities inside 24 months. If that lands, the claim is a control architecture that transfers. It is not yet a demonstration that European engines can throttle deeply — and when results appear, the numbers to look for are the lowest stable throttle point and the mixture-ratio error held there.