· Shane Trimbur
Time-Constrained MPC: Real-Time Docking for Spacecraft with Edge-Grade Compute
A new paper introduces a model predictive control (MPC) framework that guarantees real-time docking maneuvers for spacecraft, even under tight compute constraints. Here's how time-bounded optimization unlocks autonomy in orbit.
Autonomy Under Pressure: Time-Constrained MPC for On-Orbit Docking
A spacecraft attempting to dock autonomously in orbit needs to solve an optimization problem—fast. But there’s a problem: space-grade hardware isn’t built for deep compute. The paper “Time-Constrained Model Predictive Control for Autonomous Satellite Rendezvous, Proximity Operations, and Docking” (Behrendt et al., 2025) answers this challenge directly by introducing a novel form of model predictive control that guarantees execution within a fixed time budget—no matter what.
This is mission-critical for systems with modest processors like the SpaceCloud iX10-101. The framework ensures 6-DOF control (position + orientation) is maintained during autonomous proximity maneuvers—all without sacrificing responsiveness.
Time-Bounded Optimization: A New Paradigm
Most MPC strategies assume enough time exists to converge to a solution. Not so in orbit. This system instead limits the number of optimization iterations per control loop, creating a predictable, stable control surface that still achieves the required terminal docking state. It’s a pragmatic trade-off: slightly suboptimal solutions, delivered deterministically, beat optimal ones that never complete.
Key Features
- Clohessy-Wiltshire dynamics for translation, Euler equations for rotation
- Bounded iterations per time step
- No reliance on convergence—just control feasibility
- Verified across 200 randomized initial conditions
- Runs on real flight hardware
Why This Matters for the Future of Space Autonomy
Autonomous on-orbit servicing, refueling, and satellite assembly all require fast, safe decision-making under strict hardware constraints. Behrendt et al. show how to bridge that gap without scaling down mission goals or overloading compute. Their method is flexible, resilient, and ready for future space applications like debris capture, microgravity robotics, and modular spacecraft.
What’s Next
- Extend to multi-agent coordination in formation flying
- Fuse with sensor-fault detection for higher reliability
- Apply adaptive iteration schemes for even tighter performance
- Pair with vision-based navigation for fully perception-integrated control
This is the kind of edge-AI paradigm we’ll need to scale the orbital economy—and it’s ready now.
