报告时间:2026年11月28日(周六)上午 09:20-10:00
报告地点:苏州大学天赐庄校区精正楼1楼报告厅
报告人:程伟 教授,南京大学
报告摘要:
Singularities of viscosity solutions of Hamilton–Jacobi equations are not defects to be avoided. They encode where minimizing characteristics cease to be unique, where the value function loses differentiability, and where the smooth picture of optimal transport breaks down. This talk presents a unified view of this singular structure centered on the notion of strict singular characteristic (SSC). For a semiconcave solution, an SSC is the Lipschitz curve obtained by selecting at every point the supergradient of minimal energy and then following the associated Hamiltonian flow. In the optimal control setting, where the solution is a value function, this means that an SSC is the trajectory generated by the feedback control selected by minimizing the Hamiltonian over the superdifferential, it arises as a limit of classical optimal trajectories, and it is therefore the characteristic flow of the value function continued through its singular set. I will review the construction of SSCs, their existence from every point, their stability under approximation of the data, and their global propagation, and explain their relation to cut loci and to the singular set. I will then discuss how the same semiconcave and variational tools extend beyond weak KAM theory, with applications to optimal transport (Kantorovich potentials, cut loci, and singular optimal plans) and to stochastic Hamilton–Jacobi equations in the vanishing-viscosity regime (small-noise asymptotics and the zero-noise limit of singular characteristics). This talk is based on joint work with Piermarco Cannarsa, Konstantin Khanin, Jiahui Hong, and Kaizhi Wang, Tianqi Shi, as well as my PhD students Wenxue Wei, and Zhi-Xiang Zhu.
邀请人:赵云