Luis Úbeda-Medina,Á. F. García-Fernández,J. Grajal
Abstract:A novel particle filter for multiple target tracking with track-before-detect measurement models is proposed. Particle filters efficiently perform target tracking under nonlinear or non-Gaussian models. However, their application to multiple target tracking suffers from the curse of dimensionality. We introduce an efficient particle filter for multiple target tracking which deals with the curse of dimensionality better than previously developed methods. The proposed algorithm is tested and compared to other multiple target tracking particle filters.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is about the lower bound estimate of the heat kernel on Riemannian manifolds. Specifically, the author hopes to generalize Cheeger - Yau's lower bound estimate of the heat kernel for Ricci curvature to the case of integral Ricci curvature.
### Problem Background
In geometric analysis, the heat kernel is a very important quantity, which describes the diffusion process of heat on manifolds. For a Riemannian manifold \((M, g)\), the heat kernel \(E(x, y, t)\) is the kernel function of the Laplace - Beltrami operator and satisfies the heat equation:
\[
\frac{\partial E}{\partial t}=\Delta E
\]
where \(\Delta\) is the Laplace - Beltrami operator. The upper and lower bound estimates of the heat kernel are very important for understanding the geometric and topological properties of manifolds. The classical Cheeger - Yau estimate gives the lower bound estimate of the heat kernel when the pointwise Ricci curvature has a lower bound. However, this estimate is no longer applicable when dealing with more general integral Ricci curvatures.
### Main Contributions of the Paper
The main contribution of this paper is to generalize Cheeger - Yau's lower bound estimate of the heat kernel to the case of integral Ricci curvature. Specifically, the author introduces the following definitions to measure the integral Ricci curvature:
- For each \(x\in M\), let \(r(x)\) denote the minimum eigenvalue of the Ricci tensor on \(T_xM\).
- Define \(\rho(x) = |\min\{0, r(x)-(n - 1)\lambda\}|\).
- Let \(k(p,\lambda, R)\) and \(\overline{k}(p,\lambda, R)\) be respectively:
\[
k(p,\lambda, R)=\sup_{x\in M}\left(\int_{B(x, R)}\rho^{p}\right)^{1 / p}
\]
\[
\overline{k}(p,\lambda, R)=\sup_{x\in M}\left(\frac{1}{\text{vol}B(x, R)}\int_{B(x, R)}\rho^{p}\right)^{1 / p}
\]
These quantities measure the degree to which the Ricci curvature is lower than \((n - 1)\lambda\) in the integral sense. In particular, when \(\overline{k}(p,\lambda, R) = 0\), it means that \(Ric\geq(n - 1)\lambda\).
### Main Results
The main result of the paper is Theorem 1.1, which gives the lower bound estimate of the heat kernel under the condition of integral Ricci curvature:
**Theorem 1.1**: Let \(n>0\) be an integer, \(p>(n + 1)/2\), \(\lambda\leq0\) be a real number, and \(D>0\). Then there exists an explicitly computable \(\epsilon_0=\epsilon(n, p,\lambda, D)\) such that for any Riemannian manifold \((M, g)\) with diameter not exceeding \(D\) and \(\overline{k}(p,\lambda, D)\leq\epsilon_0\), we have:
\[
E(x, y, t)\geq E_{\lambda}(x, y, t)-k(p,\lambda, D)C(n, p,\lambda, D)t^{-\frac{n - 1}{2}+ 1}
\]
where \(E_{\lambda}(x, y, t)\) is the heat kernel of the space of constant curvature.
### Methods and Techniques
To obtain this result, the author adopts a method similar to that of Cheeger - Yau, that is, transplanting the heat kernel of the model space to the manifold by the Duhamel principle and making comparisons. The new difficulty lies in controlling an error term, which will disappear when the pointwise Ricci curvature is bounded. The author overcomes this difficulty by the following methods:
1.