Optimisation by predefined upper bound of settling time convergent projected gradient flows
Sunidhi Pandey, Parijat Prasun, Shyam Kamal, Devender Singh, Thach Ngoc Dinh · International Journal of Systems Science · 2025
In this paper, we utilise the prospect of integrating the idea of predefined upper bound of settling time-based stable gradient flow dynamics and projected gradient techniques to solve convex optimisation problems with linear equality constraints. We propose a projected gradient flow algorithm with a predefined-time convergence property, where the upper bound of settling time can be selected a priori. A perturbed projected gradient flow is also studied within the framework of input-to-state stability. Examples based on linear least squares and resource allocation problems (RAP) are solved using the proposed algorithm and simulation results are also compared with finite/fixed-time stable projected gradient flows.