Orion Proxima builds autonomous AI for the final metres of rendezvous and proximity operations. We supply perception, planning and control for capture, refuelling and inspection in orbit, as Software-as-a-Service and Robot-as-a-Service.
AI-powered black-box testing for spacecraft GNC.
Discover failure cases faster than Monte Carlo testing. ProximaAssure actively searches for the conditions that break your guidance, navigation and control software, so problems surface on the ground and not in orbit.
In our benchmarks, ProximaAssure uncovers over five times as many edge cases as Monte Carlo testing, spread across a far wider range of failure modes.
Test your GNC as a black box, without exposing source code or proprietary algorithms. ProximaAssure only sees inputs and responses.
Evidence-based assurance for autonomous RPO, docking and proximity operations, ready to support mission reviews and qualification.
Generates test scenarios aimed at likely weak points.
Only inputs and responses cross the interface. Your code stays private.
Each outcome feeds back so the next search targets new failures.
Orion Proxima is a UK deep-tech company building the AI-driven autonomy and capture systems that close every Rendezvous and Proximity Operation in orbit. The last mile is where in-space missions succeed or fail. It is the final few metres where a servicer must perceive, approach, align with, and physically engage a target that was never designed to be approached.
We work as a technology partner to mission primes, satellite operators and government programmes, supplying the autonomous AI software and the capture hardware it commands.
Our portfolio spans multi-DoF robotic arms, hyper-redundant continuum manipulators, refuelling end-effectors and bespoke capture interfaces. Each is driven by the same hardware-agnostic AI autonomy core.
Perception, planning and control transfer across form factors, so each new mission builds on proven software rather than starting from scratch.
A capture system is more than a robotic arm. Different missions, different targets and different orbits demand different end-effectors. Our AI autonomy stack is hardware-agnostic by design. The same perception, planning and control AI drives a six-DoF arm, a continuum manipulator threading through a propellant deck, or a refuelling nozzle making contact under thrust.
Six- and seven-DoF manipulators for life-extension docking, debris capture and component assembly. Autonomy-led grasping on cooperative and non-cooperative targets, from launch adapter rings to legacy bus features.
Hyper-redundant, snake-arm class manipulators for confined-access inspection, internal servicing and operations around delicate appendages where a rigid arm cannot reach without collision. Real-time motion planning through constrained envelopes.
Active propellant-transfer interfaces with closed-loop alignment, seal verification and flow control under residual relative motion. Designed for standardised docking plates and bespoke client valves alike.
Custom grippers, magnetic capture interfaces, soft-contact tooling and inspection booms. Co-designed with mission primes when off-the-shelf hardware doesn't fit the target geometry. The autonomy stack adapts to the actuator, not the other way around.
We support primes, operators and agencies through hardware supply, software licensing or per-mission delivery, depending on what each programme needs.
Direct procurement of capture-system hardware and a perpetual licence to the underlying autonomy stack. Suited to primes building their own fleets of dedicated servicers.
Subscription access to our AI perception, planning and control stack. Mission primes integrate the Orion Proxima autonomy layer into their own servicers without bespoke development or full IP transfer.
Capture systems delivered on a per-mission basis. Orion Proxima retains hardware ownership while partners integrate for a defined mission window. Suited to debris removal and one-off servicing campaigns.
AI perception in space fails differently than on Earth. Lighting is unforgiving, targets are reflective and ill-textured, sensors degrade, and the cost of a bad command is the mission itself. Our stack is engineered around that reality.
AI-driven 6-DoF pose estimation on non-cooperative targets using fused monocular, stereo and time-of-flight sensing. Robust through eclipse transitions, glare and sensor noise.
Adaptive AI control and motion planning that closes on a grapple feature in real time, compensating for relative-motion drift, plume disturbances and contact dynamics. Hardware-agnostic core.
Detection of degraded, deceptive or anomalous sensor inputs before they reach the controller. A requirement for the dual-use future of orbital robotics.
The same perception and control core drives every end-effector in our portfolio. Mission primes get a productised autonomy layer they can integrate, not a bespoke build per programme.
Our autonomy choices are grounded in lessons from operational in-orbit servicing missions, not simulator-only assumptions about how capture will behave.
Architected from day one to meet both commercial servicing requirements and the resilience standards of sovereign orbital security programmes.
Our founders have led perception and computer-vision work on operational in-orbit debris removal missions, and bring machine-learning research and robotics engineering from industry and academia directly into the capture stack.
Whether you are preparing a servicing, refuelling or inspection mission, or specifying autonomy for an upcoming demonstrator, we would like to hear from you.