Adaptive ZNN Model and Solvers for Tackling Temporally Variant Quadratic Program With Applications

As the zeroing neural network (ZNN) approach needs human intervention to handle temporally variant quadratic program (TVQP) problem, which results in less flexibility and convenience, especially considering the situation of tolerated precision given. Aiming at these weaknesses, in this article, we propose a type of adaptive ZNN (AZNN) approach on solving TVQP problem. Specifically, we design and propose a novel continuous-time AZNN model and two adaptive time-discretization techniques with different precision to obtain two discrete-time AZNN (DAZNN) solvers, making them more readily implemented on digital computers. The convergence analyses for these model and solvers are theoretically proved. Comparative experiments among the proposed DAZNN solvers and four existing solvers are conducted to verify the efficacy and superiority of the DAZNN solvers. Furthermore, simulative and physical experiments for the position tracking of JACO2 mechanical arm are successfully performed. Finally, based on the AZNN approaches, a novel adaptive kinematics planner and two adaptive solvers are proposed and applied for the pose tracking of the Franka Emika Panda mechanical arm, further substantiating their efficacy and effectiveness.