Artificial life-inspired morphology/learning codesign framework: Towards conceptual design of intelligent hybrid quadcopters

In this paper, a novel co-design process is proposed to solve the coupled objectives of maximizing flight range and collision avoidance robustness of an autonomous quadcopter UAV. The co-design process combines the morphology and intelligence designs into a single on-the-line learning/optimization process. The UAV features the H-shaped physical structure, aerodynamic wing-alike BWB, and a neuron network based collision avoidance intelligence maneuver model. A case study involving two flight scenarios, evaluating the endurance and maneuverability, is performed in the form of a dual-objective optimization process. Strategies include surrogate modeling, combined high-and-low fidelity modeling, PSO based neuron network training, inheritance-driven Lamarkian learning, and parallel computing are applied to reduce the computational cost of the case study. The optimization results captured a Pareto frontier of the trade-offs between the flight range and the collision avoidance capacity. An analysis of the results shows strong couplings among the morphology and intelligence design parameters. The case study results indicate the novel UAV design reaches a good balance between endurance and maneuverability, and the proposed morphology-intelligence co-design framework is effectively co-optimizing coupled parameters to solve the conflicted objectives, while demanding limited computational cost.