Evolutionary Optimization Methods for Morphological Design of Unmanned Vehicles: A Desk Study

Abstract

The morphological design of unmanned vehicles involves complex multi-objective optimization across aerodynamic performance, structural integrity, payload capacity, energy efficiency, and manufacturing constraints—a challenge well-suited to evolutionary computation approaches. This desk study comprehensively examines evolutionary optimization methods applied to unmanned vehicle morphology, spanning genetic algorithms (GA), evolution strategies (ES), genetic programming (GP), and multi-objective evolutionary algorithms (MOEAs) including NSGA-II, NSGA-III, and MOEA/D. We systematically analyze the representation schemes for encoding vehicle morphology, from parametric models with fixed topology to generative encodings enabling radical design variations, including L-systems, cellular automata, and compositional pattern-producing networks (CPPNs). The study evaluates fitness evaluation strategies, examining the integration of computational fluid dynamics (CFD), finite element analysis (FEA), and multi-body dynamics simulations within evolutionary loops, addressing the computational expense through surrogate modeling, reduced-order models, and parallel evaluation architectures. Particular attention is devoted to multi-objective formulations balancing competing design criteria: lift-to-drag ratio versus structural weight for UAVs, hydrodynamic efficiency versus maneuverability for UUVs, and terrain traversability versus energy consumption for UGVs. We analyze constraint handling techniques for ensuring design feasibility including penalty functions, repair mechanisms, and constraint-domination principles. The desk study synthesizes case studies demonstrating evolutionary design of UAV wing configurations, UUV hull shapes, and UGV suspension systems, identifying common patterns in successful optimization campaigns. Advanced techniques are examined including co-evolution of morphology and control, developmental encodings producing scalable designs, and quality-diversity algorithms (MAP-Elites) generating diverse design portfolios. We critically assess the interpretability of evolved designs, investigating whether evolutionary processes discover known engineering principles or generate non-intuitive solutions. The study concludes by identifying methodological best practices, discussing the integration of additive manufacturing to realize complex evolved geometries, and highlighting open challenges in incorporating manufacturing constraints and lifecycle considerations.