Vel-v4.3: A Decentralized Swarm Protocol for Autonomous Volume Dominance

This paper presents the comprehensive empirical validation of the Vel-v4.3 protocol, a decentralized swarm coordination system for autonomous robotic navigation, introducing a critical architectural upgrade: the replacement of center-outward bloom repulsion with Map-Aware Gradient Repulsion (MAGR). Validated across five systematic parameter sweeps comprising 105,000 simulation trials, this architecture is characterized across four independent protocol dimensions: agent population (N), repulsion factor (G), momentum coefficient (μ), and MAGR scan radius (r). The Population Sweep (N ∈ \50, 150, 300\) confirms that MAGR successfully restores the monotonic efficiency scaling previously eliminated by the original center-bloom architecture. The integer G Sweep (G ∈ \0, 2, 4, 6, 8\) establishes $G = 2$ as the coarse optimum and reveals that repulsion mechanics account for over 97% of total protocol efficiency, a configuration further refined by the Fine G Sweep (G ∈ [1.0, 3.0], Δ G = 0.25) to $G = 1.5$ at $N = 300$, yielding an average efficiency of η̄ = 1.796. Additionally, the Momentum Sweep (μ ∈ [0.70, 0.99]) identifies μ = 0.99 as the empirical optimum, delivering an average efficiency of η̄ = 2.696, which marks a 49.6% performance gain over the original baseline assumption of μ = 0.88. Finally, the Scan Radius Sweep (r ∈ [2, 10]) demonstrates that while MAGR efficiency is fundamentally radius-invariant (Δη = 0.006 across all tested values), the computational overhead scales by 17.6× from $r = 2$ to $r = 10$, thereby establishing $r = 2$ as the optimal deployment radius. Ultimately, the global operational efficiency optimum is achieved at $N = 300$, $G = 1.5$, μ = 0.99, and $r = 2$, delivering a performance profile of η̄ = 2.696, a low rolling variance of σ = 0.089, and an average convergence time of T̄ = 35 steps. This represents a 78.7% improvement over the initial baseline, effectively superseding the original operational benchmark of η > 0.105 with an expected target of η > 2.0 for fully-optimized MAGR configurations.