Distribution-Allowed Noise-Resistant Neural Dynamics for Constrained Time-Dependent Quadratic Programming With <i>k</i>WTA Application

Existing computational models for addressing time-dependent quadratic programming (TDQP) problems encounter some challenges, such as generating lagging errors, lack of noise immunity, and inability to be distributed. To handle these challenges, this article proposes a distribution-allowed noise-resistant neural dynamics (DANRND) model to solve TDQP problems with equality and inequality constraints by introducing auxiliary variables rather than by using the nonlinear complementary problem (NCP) function. The proposed model is able to effectively eliminate the hysteresis error and suppress the influence of noises. Specifically, the proposed model is capable of implementation in a distributed manner, which extends its scope of applications. Then, theoretical analyses are provided to prove the global convergence in both noise-free and noisy conditions. Simulative examples and comparison results with existing methods are offered, demonstrating the superiority of the proposed DANRND model. Finally, a distributed cooperative task based on the k-winner-take-all (kWTA) operation is performed on a multirobot platform to further verify the distributed implementation of the proposed DANRND model.