Confronted with an array of challenges in the realm of underground positioning technology, including multipath, complex architecture structures, obstructions of positioning signals, and diverse positioning environments (such as underground parking (rectangular planes and long planes) and underground roadways (mostly narrow and long)). Due to the needs of social production and development, high-accuracy positioning services and location techniques have become indispensable. It is crucial to determine the precise location of people and objects in underground environment. Conventional positioning technologies and algorithms are difficult to obtain high-precision position information of underground carriers or personnel. UWB (Ultra-Wideband) has the advantages of penetration power, low power consumption, high data rate and well anti-multipath ability, and low complexity of the system, which can provide better positioning results than most sensors. In the underground experiment or complex architectural structure experiment, although UWB can obtain good positioning results in the LOS (Line of Sight) condition, the positioning technology is greatly influenced by the layout and number of the base stations. Therefore, to explore the optimal layout in multiple underground experiments, a positioning technology based on TDOA (Time Difference of Arrival) and AOA (Angle of Arrival) information is proposed in this article. First, according to the angle positioning and maximum likelihood method, the initial position of the object is obtained. Second, a performance index optimal model is determined the iteration step size and direction of iteration by using the minimization function of the second order expansion of Taylor series, and determined the final position of object by using a positive definite performance index defined by the sum of squares of distance errors. Finally, there are 10 groups of random objects in the underground experiment to verify the proposed algorithm, and the layouts of base stations are set in the forms of square, Y-type and diamond-shaped, indicating that the Y-type has better positioning than the other two. We discuss the positioning results with time errors and ranging errors in the case of Y-type deployment stations, and the Cramero boundary be used as an accuracy evaluation index. The 1000 Monte Carlo experimental simulations were used to explore the effects of time errors on positioning accuracy with small angle errors. We can determine the final position results and the proposed algorithm has obvious advantages, which demonstrates that the deviations between the estimated results of the proposed algorithm and the actual results are small and they are close to the Cramero boundary.