Sensor-Uncertainty-Aware Conservative Robust Route Selection for Autonomous Robot Path Planning Under Occupancy-Grid...

Autonomous robotic navigation in dynamic environments depends on sensor-derived occupancy maps that are often degraded by occlusion, localization error, dynamic blockage, incomplete observation, and perception noise. These uncertainties can make a nominally short route unsafe after deployment, motivating conservative route selection for collision-aware path planning under sensor-derived map uncertainty. We formulate Conservative Robust Route Selection (CRRS) as a finite-scenario robust optimization and route-selection framework for autonomous robotic path planning under this uncertainty. CRRS constructs a heterogeneous portfolio of candidate routes, scores each route using nominal and plausible-world information only, and applies a validation-frozen conservative override rule that defaults to the scenario ensemble unless a candidate route satisfies predefined feasibility, risk, clearance, and cost-ratio guards. The evaluation protocol separates implementation auditing, candidate-pool expansion, validation-based selector design, frozen confirmation, public-benchmark validation, simulated sensor-model validation, and a controlled validation-held-out mismatch stress test. On the generated 30-domain MovingAI-format benchmark, candidate-pool expansion finds strict-safe-superior candidates in 113/150 matched groups, and the frozen selector reduces plausible collision from 0.1250 to 0.0807 and held-out collision from 0.3053 to 0.2937. On an official long-distance MovingAI subset with 260 queries and a minimum start-goal distance of 100 cells, CRRS reduces the held-out collision from 0.9648 for the scenario ensemble to 0.8822, with 141 wins, zero losses, and 119 ties. In an additional LiDAR/SLAM-inspired simulated sensor-model validation on 200 routed official-query problems, CRRS reduces the held-out collision from 0.4059 to 0.3768 relative to the scenario ensemble. A validation-held-out mismatch stress ablation isolates the conservative override rule: CRRS differs from CVaR-only on 73/260 problems, obtains a lower or equal held-out collision in every comparison, and avoids the 18 harmful held-out losses incurred by CVaR-only relative to the scenario ensemble. The resulting claim is deliberately scoped: CRRS improves aggregate route robustness over a strong scenario-ensemble default on the evaluated robotic path-planning benchmarks, while real-time deployment, live sensor integration with calibrated sensors, physical robot validation, family-level variation, and benchmark-specific uncertainty models remain limitations.