Abstract: Underground pipe rack,which is complex and changeable,is a semi-enclosed narrow space.Due to the emergence of non-line-of-sight situation,the sensor node positioning is very difficult,and the accuracy is not good. In view of the special situation of the underground pipe rack,like the multipath interference,this paper proposes the least squares non-convex estimation (LSNCE )algorithm.Combined the received signal strength(RSSI) and angle-of-arrival(AOA) information,we obtain the position estimation of the unknown node by using the least squares method.And then,we first introduce an auxiliary variable X and an auxiliary vector z of the unknown node.By applying the semidefinite programming relaxation technique,the position estimation is transformed into convexity,and we also obtained an Epigraph variable.Then,the optimal solution of the position estimation is obtained by using the CVX toolbox (included in Matlab),That is,the final location of the unknown node.Through the simulation,compared with conventional SHWA and SWLS algorithm,LSNCE algorithm can effectively improve the positioning accuracy,at the same time shows the good anti-jamming ability.