Particle filter optimization parameter-selection conjectures

Let $k_{\max}$ be the maximum number of iterations, let $\lambda$ be the unscented-transform scaling factor, let $N$ be the number of particles, and let $Q$ be the state-transition covariance. The parameter analysis considers these quantities at high, low, or moderate levels, and compares the resulting optimization solutions.

Parameter-selection conjecture. It is likely to reach a better solution with lower $k_{\max}$ and $\lambda$ when $N$ and $Q$ are high. The results suggest a correlation between $k_{\max}$ and $N$, and between $\lambda$ and $Q$.

This claim is based on random-sampling experiments for parameter selection in single-variable optimization and is suggested to extend to relatively higher dimensions. Its status is not established beyond the reported experiments.