DecABLS: Decentralized Asynchronous Bitwise Level Set Planner

This paper presents DecABLS: a decentralized, asynchronous, multi-agent trajectory planning framework that generates persistently feasible, safe, and optimal trajectories. Persistent feasibility and safety is critical for fixed-wing aircraft because they are unable to avoid collisions by hovering like rotary-wing aircraft. Within parameterized communication and computation delay allowances, DecABLS provides this feasibility and safety guarantee using a decentralized deconfliction strategy that is analogous to the Time Division Multiple Access (TDMA) method from telecommunications. However, DecABLS allows planning to execute asynchronously, which enables the use of optimal control planners ranging from application-specific analytic methods to general purpose commercial solvers. The DecABLS concept is extensible to heterogeneous dynamics, but only Dubins-modeled fixed-wing agents with various parameters are explored in this paper. DecABLS simulations with up to 32 agents show that multi-agent safety is maintained, and normalized mean time-of-flight scales well with increasing swarm size. Several indoor hardware demonstrations validate the DecABLS safety guarantees. Lastly, DecABLS is used to successfully coordinate and deconflict two live-virtual flight tests using UAS.