Aiming at the shortcomings of single sensor positioning in indoor unknown environment, such as large cumulative error and limited by environment, a multisensor nonlinear fusion positioning system was designed to improve the positioning accuracy of mobile robot autonomous navigation. In the system, Gauss-Newton equation was used to fuse and optimize the pose information measured by the lidar, inertial measurement unit and wheel odometer to compensate the positioning error caused by the low positioning accuracy of a single sensor under indoor environmental information. The experimental results show that the mobile robot using the multisensor fusion positioning system can stabilize the positioning error within 0.106 3 m after running a total distance of 12.8 m when the indoor is 6 m long and 3 m wide while facing a tortuous and complex path and various noise disturbances. And the average relative error is stable about 0.716%. Compared with existing methods, using this method improves the accuracy and robustness of indoor mobile robot positioning.