Pedestrian inertial navigation systems (PINSs) based on zero velocity update (ZUPT) are extensively used in pedestrian positioning. However, because of sensor drift, traditional foot-mounted PINSs show poor precision. We design a dual MIMU single-board PINS and propose a new constraint method. A constant three-dimensional (3D) position difference constraint can be formed on the basis of the layout of the two sensors on the circuit board. The constraints of (linear) velocity, angular velocity, and constant position difference are constructed on the basis of the Coriolis theorem. To obtain the optimal velocity and position, a linear Kalman filter is constructed. Rectangle and stair experiments are performed to verify the proposed method. The results show that the proposed method can improve the positioning accuracy compared with the classical ZUPT and dual-sensor foot-to-foot constraint methods. Compared with the constraint in the two-step maximum distance scheme, the proposed position and velocity constraints are clearer and can be applied to the whole gait phase. Furthermore, the proposed PINS in this study has the advantages of easy installation, small size, high-speed communications, and low power consumption.