Uncrewed aerial vehicles (UAVs) often need to penetrate wildfires, collapsed structures, or mine tunnels to collect vital data, yet falling debris can easily destroy them. Existing navigation stacks typically guarantee safety only in static or pre‑mapped scenes, leaving dynamic, unknown hazards unaddressed.
MIT researchers introduced SANDO (Safe AutoNomous trajectory planning for Dynamic unknOwn environments), a planner that requires merely the maximum speed of any obstacle to mathematically guarantee collision‑free flight in completely unmapped environments. No prior map is needed; knowing the speed bound suffices for universal safety.
The method builds a time‑varying safety corridor. First, dynamic obstacles are detected, clustered, and tracked. For each obstacle a sphere is drawn whose radius equals the maximum speed multiplied by a chosen time horizon, representing the farthest region the obstacle could occupy. The corridor consists of connected 3D regions that lie outside all such spheres, ensuring no obstacle can intersect the UAV’s future path.
A heat‑map planner then highlights “hot” zones with dense obstacles and steers the UAV away, improving efficiency. Within the corridor a hard‑constrained trajectory optimization finds the fastest route to the goal. As the UAV moves, the corridor and trajectory are continuously updated on‑board, enabling rapid replanning.
Simulations and twelve real‑flight experiments showed SANDO reaching the target faster than several state‑of‑the‑art baselines while maintaining zero collisions, confirming the theoretical safety guarantee. Future directions aim at further computational gains and integration with machine‑learning models for natural‑language instruction.
Review: By uniting spatiotemporal safety corridors with rigorous constrained optimization, SANDO delivers a provably safe solution for autonomous UAV navigation in dynamic, unknown settings, marking a significant step toward reliable real‑world deployment.