MIT Unveils SANDO, a UAV Path-Planning Method to Avoid Collisions in Unknown Environments

MIT researchers on October 7, 2026, introduced SANDO, a path-planning method that mathematically guarantees unmanned aerial vehicles (UAVs) will avoid collisions in unknown dynamic environments. The system creates spatial safety corridors and accounts for obstacles’ future positions based on their maximum speeds. It also avoided collisions in 12 flights with a real UAV. The research was published in IEEE Transactions on Robotics.

According to MIT, SANDO offers a formal guarantee that a UAV will not collide with obstacles in an unmapped environment, provided their maximum speeds are known. This is relevant to users who want to assess whether a UAV can fly during search-and-rescue operations based on conditional safety guarantees rather than intuition.

SANDO’s novelty is not in localization or path optimization in unknown environments, but in mathematically guaranteeing collision avoidance in unmapped areas with moving obstacles. By contrast, using a UAV’s cameras and sensors to navigate while replanning onboard is an extension of established approaches in robotics research. The focus of this announcement, then, is on bringing perception, mapping and replanning together as a constrained optimization problem with formal safety guarantees. There is no connection to our previous coverage, but the announcement is significant because it offers a concrete solution to a gap in research: the difficulty of guaranteeing safety in dynamic environments. The design relies only on obstacles’ maximum speeds; rather than fully predicting each object’s behavior, it encloses the area an obstacle could reach in the future with a sphere and draws safety corridors around it. This narrows the required inputs to make the guarantee possible. The approach can be seen as prioritizing safety constraints over search speed in applications such as sending UAVs into buildings with rubble or into wildfire zones. For the industry, the key issue is how to connect sensor inputs, onboard computing, dynamic-obstacle modeling and safety guarantees. SANDO uses a heat map to avoid congested areas while continually reoptimizing its trajectory, making ongoing replanning—not one-time path generation—a core part of the system. In real-world operations, computing demands, replanning frequency and sensor tracking accuracy are therefore likely to affect performance. Questions to watch include the obstacle densities and speeds under which the guarantee holds, the conditions covered by the 12 real-world flights, and whether integration with machine-learning-based natural-language instructions can work without compromising the safety guarantee.

SANDO plans paths using information from a UAV’s onboard cameras and sensors. Unlike conventional methods for unknown static environments, or methods that only steer clear of dynamic obstacles without formal guarantees, the research presents an algorithm that theoretically guarantees collision avoidance with moving objects. The system establishes safety corridors made up of connected regions in three-dimensional space and uses obstacles’ maximum speeds to represent their reachable areas as spheres. It then detects and tracks obstacles as groups and combines this with heat-map-based planning to avoid congested “hot” areas. In simulations, the researchers reached destinations faster than several state-of-the-art systems. In real-world tests, the UAV repeatedly replanned using an onboard computer and sensors.

In Japan, formal safety guarantees and onboard replanning in unknown environments are implementation issues for UAV use in disaster response and infrastructure inspections. The ability to integrate sensors, embedded computers and control software will be a key factor in moving research results into real-world operations.