Flight Structure Optimization of Modular Reconfigurable UAVs

TL;DR

To address heterogeneity in modular UAVs, the paper formulates flight-structure optimization as a genetic algorithm operating on a tree-based connectivity representation and a force-decomposition dynamic model. The approach yields over-actuated configurations with improved controllability and energy efficiency, demonstrated on swarms of 30 and 37 modules with substantial reductions in computation time compared with enumeration, aided by dual representations and a force-allocation framework using $\mathbf W$ and $\mathbf J_{\mathcal S}$. A dual representation framework using the connectivity aim $\mathbf A(\mathcal S)$ and docked-module positions $\mathbf d_i$, together with a force-decomposition model and the allocation matrix $\mathbf W$, enables efficient fitness evaluation via the total inertia $\mathbf J_{\mathcal S}$. Simulation results show better trajectory tracking and lower thrust energy for optimized structures, indicating practical potential for scalable, reconfigurable UAV platforms and online adjustments.

Abstract

This paper presents a Genetic Algorithm (GA) designed to reconfigure a large group of modular Unmanned Aerial Vehicles (UAVs), each with different weights and inertia parameters, into an over-actuated flight structure with improved dynamic properties. Previous research efforts either utilized expert knowledge to design flight structures for a specific task or relied on enumeration-based algorithms that required extensive computation to find an optimal one. However, both approaches encounter challenges in accommodating the heterogeneity among modules. Our GA addresses these challenges by incorporating the complexities of over-actuation and dynamic properties into its formulation. Additionally, we employ a tree representation and a vector representation to describe flight structures, facilitating efficient crossover operations and fitness evaluations within the GA framework, respectively. Using cubic modular quadcopters capable of functioning as omni-directional thrust generators, we validate that the proposed approach can (i) adeptly identify suboptimal configurations ensuring over-actuation while ensuring trajectory tracking accuracy and (ii) significantly reduce computational costs compared to traditional enumeration-based methods.

Figures (7)

• Figure 1: The optimal structure configuration with five modular uav with different installed equipment. Each module is equipped with either a manipulator, an RGBD camera, a Lidar, or a computing unit, resulting in different weights and inertia parameters. The proposed algorithm efficiently produces an over-actuated flight structure with optimal dynamical properties.
• Figure 2: Coordinate systems and configuration representations of a flight structure. Each quadcopter module can connect to four others with its docking faces, and each module serves as an omni-directional thrust generator after connections, making the flight structure over-actuated. Different types of representation of the flight structure are incorporated to support subsequent optimization.
• Figure 3: The fitness values for different flight structures composed of five same-weighted modules. Configuration (a) has $-\text{Inf}$ fitness value as it is under-actuated, while (f) has the maximum fitness value (optimal dynamics property) due to its symmetric configuration.
• Figure 4: Steps of a crossover operation. The crossover operation divides the original tree into two small trees and reconnects them to build a new one. Through the crossover operation, all the feasible configurations can be acquired.
• Figure 5: The optimization process of generating an optimal flight structure for a large modular UAV system swarm with different-weighted modules. Two cases with $n=30$ and $n=37$ are presented to illustrate the evolution of the optimization process.