Quantum-Limited Sensing-to-Action for GPS-Denied Navigation in Autonomous Robots

Abstract: This study investigates a sensing-to-action pipeline for GPS-denied autonomy that explicitly couples navigation uncertainty to safety-constrained decision-making. GPS denial is intrinsic to subsea operations and frequently arises in space and lunar environments where satellite navigation is unavailable or unreliable, and where perception is degraded by scattering, aliasing, and extreme illumination. A factor-graph formulation implemented in GTSAM is used to express multi-sensor navigation as probabilistic inference, enabling posterior covariances and estimator-health diagnostics to be produced alongside state estimates. Quantum-limited sensing is introduced as a physically grounded way to bound inertial uncertainty by clamping process noise and bias evolution to experimentally demonstrated performance envelopes from quantum accelerometry and quantum rotation sensing. The empirical foundation uses two real-world, public datasets that capture GPS-denied conditions in domains relevant to space and marine robotics. The AQUALOC underwater harbor sequences combine monocular imagery, inertial measurements, and pressure-derived depth, thereby capturing feature intermittency and scale/altitude challenges typical of underwater SLAM. The DLR S3LI planetary analog dataset was recorded on Mount Etna in volcanic terrain using stereo vision, solid-state LiDAR with limited field of view and minimum range, and inertial sensing, thereby capturing perceptual aliasing and sparse geometry. Published benchmark results on these datasets are consolidated into inline tables and figures to quantify accuracy–robustness trade-offs that directly affect safety. Results show (i) consistent underwater accuracy gains from pressure-aided fusion but sensitivity to tracking loss, (ii) pronounced completion–accuracy trade-offs on planetary-analog terrain, and (iii) orders-of-magnitude differences in inertial drift bounds when quantum-limited accelerometry and rotation sensing envelopes are used as uncertainty floors. A risk proxy based on chance-constraint tails illustrates how reduced uncertainty can justify smaller safety buffers for a fixed probability-of-violation target, improving mission feasibility in constrained environments. The resulting blueprint treats uncertainty as a first-class output, validates it through consistency tests, and consumes it in a planner that enforces probabilistic safety constraints. Keywords:GPS-denied navigation; factor graphs; GTSAM; uncertainty quantification; quantum inertial sensing; risk-aware planning; underwater SLAM; planetary robotics; safety constraints.