Simulating a swarm of heterogeneous UAVs for Road Assessment during emergencies using the Yellowstone Supercomputer

During emergencies, it is paramount to assess the state of roads and other infrastructure in near-real time. We envision an autonomous swarm of UAVs, flying to the affected area to acquire comprehensive imagery, completely autonomous and self-organizing after taking off, even in case of UAVs losses. In this work, we present the use of the NCAR's massively parallel Yellowstone supercomputer to simulate the behavior of a swarm of UAVs as an autonomous entity. At the moment, we focus only on the supercomputer simulation.

UAVs are capable of autonomous flight to a destination using GPS or other navigation methods, but are however constrained by their power and radio range. When multiple UAVs are used in a collaborative environment to achieve a common goal, like for example maximizing the coverage of an area, navigation strategies should address how to control them cooperatively, taking into account the UAVs constraints, different characteristics and potential failures.

This research explores solutions to this problem by testing different optimization algorithms in a simulated environment. The use of the supercomputer is required because of the complexity of the search space, paired with the computational requirements associated with simulating each UAV as an individual agent.