🤖 AI Summary
To address the challenge of autonomous rendezvous with interstellar objects (ISOs) characterized by high orbital inclinations, high velocities, and uncertain orbital altitudes, this paper proposes an end-to-end real-time navigation and control framework. The method innovatively integrates a spectrally normalized deep neural network with a pointwise minimum-norm tracking controller and introduces a stochastic incremental stability analysis—yielding, for the first time, a rigorous exponential probabilistic bound on delivery error that explicitly accounts for state uncertainty and the local nature of nonlinear estimation. By jointly optimizing model predictive control (MPC) trajectory tracking loss and constructing a stochastic supermartingale, the framework satisfies the theoretical error bound across all 100 candidate ISO trajectories. Validation via high-fidelity spacecraft simulation and distributed reconfiguration experiments with 20 UAVs demonstrates millisecond-level, robust, sub-meter-accuracy autonomous guidance.
📝 Abstract
Interstellar objects (ISOs) are likely representatives of primitive materials invaluable in understanding exoplanetary star systems. Due to their poorly constrained orbits with generally high inclinations and relative velocities, however, exploring ISOs with conventional human-in-the-loop approaches is significantly challenging. This paper presents Neural-Rendezvous—a deep-learning-based guidance and control framework for encountering fast-moving objects, including ISOs, robustly, accurately, and autonomously in real time. It uses pointwise minimum norm tracking control on top of a guidance policy modeled by a spectrally normalized deep neural network, where its hyperparameters are tuned with a loss function directly penalizing the model predictive control state trajectory tracking error. We show that Neural-Rendezvous provides a high probability exponential bound on the expected spacecraft delivery error, the proof of which leverages stochastic incremental stability analysis. In particular, it is used to construct a non-negative function with a supermartingale property, explicitly accounting for the ISO state uncertainty and the local nature of nonlinear state estimation guarantees. In numerical simulations, Neural-Rendezvous is demonstrated to satisfy the expected error bound for 100 ISO candidates. This performance is also empirically validated using our spacecraft simulator and in high-conflict and distributed unmanned aerial vehicle swarm reconfiguration with up to 20 unmanned aerial vehicles.
