Assessing Collision Risk for Autonomous Drone Swarms in Confined Fluids via Lattice Boltzmann Simulation

This paper presents a simulation framework for assessing collision risk in autonomous drone swarms operating in confined gaseous environments. The framework couples a three-dimensional 19-velocity (D3Q19) Lattice Boltzmann Method (LBM) fluid solver with explicit rigid-body models of self-propelled drones, using a single-relaxation-time Bhatnagar–Gross–Krook (BGK) operator to enable two-way momentum exchange between drones and fluid and to model inter-drone contacts as perfectly elastic. We simulate swarms of 30 to 60 drones in cubic grids with periodic boundary conditions. The grid sizes reached \((130^3)\) cells, and we tested at different fluid densities (\((\rho = 1.0, 1.8, 2.93)\), and \((4.55)\)). For each setup, we ran multiple simulations as part of an ensemble. We analyzed the outcomes using nonparametric statistical tools—specifically Spearman’s rank correlation \(( r_s)\), Kendall’s \((\tau)\), and associated confidence intervals—to understand how often collisions occurred. The results show a strongly nonlinear increase in collision frequency with swarm size, while fluid density has only a minor effect within the tested range. When we increased the domain size from \((100^3)\) to \((130^3)\), the total number of collisions decreased. However, collision frequency still showed a convex relationship with swarm size. In our simulations, the drones lack built-in collision-avoidance or traffic-coordination mechanisms. As a result, the collision statistics reported here likely represent a conservative upper bound on the collision frequency at a given density, assuming no behavioral safeguards are in place. The framework thus provides a physics-based tool for studying fluid–swarm interaction and for benchmarking collision-avoidance strategies for autonomous multi-robot systems in confined fluids.