In recent years, the micro-electro- mechanical-system(MEMS) IMUs (Inertial Measurement Units) have been widely used in UAVs(Unmanned Aerial Vehicles), UGVs(Unmanned Ground Vehicles) and Robotics due to their small size, light weight, low power consumption and low cost.MEMS IMU contains 3-axis accelerometer, 3-axis gyroscope, 3-axis magnetometer, thermometer, etc. in a single microchip, but MEMS gyroscope’s accuracy tends to be low, especially by temperature influence, and they generally require error compensation in practical applications.Especially, the zero-bias is greatly changed by temperature variations.Considering the simplification and real-time requirement, we present the hardware and software design methods for thermostatic control of MEMS IMU using a peltier plate.Then the Neural Adaptive PID (Proportion Integration Differentiation) controller is used to stabilize the inner temperature of MEMS IMU enclosure by controlling the amount of heating and cooling of a peltier plate.As shown in the experiment, the transient time is less than 3 min , the rising time is less than 5s/℃ and the falling time is less than 10 s/℃ so the property of the dynamic response in the thermostatic control system is improved much better than in other methods.The overall cost is lower than 100$ and it is easy to implement hardware and software.Thus this can be widely used in UAVs, UGVs and Robotics