Enhanced Beetle Antennae Search with Zeroing Neural Network for online solution of constrained optimization

This paper proposes a continuous-time enhanced variant of Beetle Antennae Search (BAS), a metaheuristic algorithm that mimics the food searching nature of beetles. Beetles register the smell of the food on their two antennae, and based on the intensity of smell, they move left or right. Likewise, discrete-time BAS computes the value of objective function three times and moves toward the optimal solution. However, the computation of objective function three times in each iteration known as a “virtual particle,” makes it computationally expensive, inefficient, and time-consuming, especially while dealing with complex systems, e.g., redundant manipulators. Our proposed, Enhanced Beetle Antennae Search with Zeroing Neural Network (BASZNN) algorithm overcomes this problem by introducing a delay factor in objective function and input. This delay allows BASZNN to compute objective function value once, making it computationally robust and efficient. BASZNN includes the flexible random searching nature of BAS and the parallel processing nature of ZNN, making it computationally fast and less time-consuming, especially for complex problems. As a testbed, we employed BASZNN on two types of problems: Unconstrained (unimodal, multimodal) and Constrained (real-world) and compared the results with state-of-the-art metaheuristic algorithms.