Abstract: To solve the problem of location inaccuracy caused by abnormal error registration, a variable step-size constrained total least squares algorithm is presented. By building a nonlinear function relationship between step size and steps, algorithm step size is relatively large in initial stage and is very small during steady state. So this algorithm can achieve faster convergence rate and smaller steady-state imbalance noise at the same time. This algorithm reduces the convergence steps and improves the stability after convergence. Therefore this algorithm reduces the impact of abnormal error registration on location result. Meanwhile, this algorithm also makes location result more accurate. Theoretical derivation and simulation results show that variable step-size constrained total least squares algorithm is superior to constrained total least squares algorithm. Simulation results also verify the feasibility of the variable step-size constrained total least squares algorithm.

Key words: abnormal error registration, step-size, convergence rate, location result