AERO: A Neo-Darwinian Approach to Time-Energy Optimal Behavioral Robotics

Abstract

Multi-rotor Autonomous Aerial Vehicles (AAVs) have seen a rapid increase in availability and deployment. As such, these systems are on the cusp of demonstrating their full potential. However, despite such promise, AAV systems continue to face poor mission resilience due to limited onboard energy reserves offset by weight, cost, and the high energy consumption needed to maintain lift. As a result, this thesis proposes an adaptive multi-variable optimizing control framework for AAVs called AERO (Another Evolutionary Robotic Optimizer). AERO consists of a layered software structure comprising Neo-Darwinian “directed evolution” techniques that generate time/energy tuned Behavior Tree parameters that then drive the AAV flight controller actions and attitude postures. AERO computes, through multi generational sampling of each AAV’s mission segment preferred velocity so as to converge to an optimal balance of both minimal transit time and low energy consumption across the overall mission profile. This approach allows for a self-aligning AAV layered control mechanism to dynamically incorporate numerous and varied behavioral responses while critically leaving the vehicle’s core flight controller and system logic unchanged.