Sinusoidal-perturbation-based two-dimensional source extremum searching methods have the drawbacks of poor adaptability and mutual restriction between rapidity and accuracy. Aiming at this problem

a gradientestimation-based parameter adaptive extremum seeking algorithm is proposed. On the basis of traditional extremum seeking algorithm

this algorithm estimates gradient of the local area with three historical sample points and adjusts feedback gain parameter adaptively according to the gradient value. Moreover

the average value theory is used to theoretically analyze and prove the convergence of the proposed algorithm. Simulation comparison in different environments show that the proposed algorithm increase source search efficiency and adaptability to complicated gradient environments.