Nature inspired collaborative team autonomy (NICiTA)

Abstract

Collaborative autonomy of unmanned air-vehicles (UA) in harsh or contested environments is a challenging and increasingly important problem. While prior attempts have been made to perform collaborative autonomy, the high complexity of UA systems in general and the planning in particular, have resulted in centralized solutions that tend to be brittle and lack scalability, or in distributed approaches that incur high overheads and require over-simplifying assumptions. In contrast, we propose an approach that is an inherently scalable (linear in the number of tasks, per UA (parallelizable over the UAs) per update) and performs distributed joint task assignment and trajectory planning using local laws analogous to those in natural processes. More specifically, trajectories are planned using the artificial potential force method, with goals and neighbors modeled as attractive forces and hazards modeled as repulsive forces. Additionally, task assignments are regulated using dynamic “valences” that are associated with the goal positions of trajectory planning. This approach allows the incorporation of dynamics and feedback laws to guarantee equilibrium stability. When applied to a scenario of simulated communication networking in harsh environment, our results show that NICiTA is able to build a relay UA data network that has near perfect data-delivery ratio despite the presence of a nearby interferer.