Development of MEMS Inertial Sensors

The intelligent autonomous control of representative unmanned systems such as drones, unmanned vehicles, unmanned ships and robots is a research hotspot in the field of automatic control, and it is the core technology to improve the autonomy and intelligence level of unmanned systems. The autonomous navigation technology uses the corresponding autonomous navigation system to acquire the position, speed and attitude information of the unmanned system itself, which is an indispensable technical guarantee for the intelligent autonomous control of the unmanned system. In many navigation technologies such as radio navigation, terrain matching navigation, inertial navigation, satellite navigation, magnetic navigation and visual navigation, inertial navigation technology that does not need to rely on external information is one of the most powerful technical means to achieve autonomous navigation of unmanned systems. . MEMS inertial navigation technology based on MI-based inertial sensor (MEMS) inertial sensor is an important branch of inertial navigation technology. Its system has the advantages of low cost, small size, low power consumption and strong impact resistance. Therefore, research on MEMS inertial sensors and their navigation technology has important supporting significance for the rapid development of unmanned system autonomous navigation technology and meeting its increasing application requirements.

MEMS inertial sensor

1 Classification of MEMS inertial sensors

MEMS inertial sensors include MEMS gyroscopes and MEMS accelerometers. There are many ways to classify them. According to the accuracy from low to high, they can be divided into consumer grades (zero offset >100°/h) and tactical grades (zero offset 0.1°/h). ~ 10°/h).

According to the way of sensing angular velocity, MEMS gyroscopes can be divided into vibrating arm type, vibrating disc type and ring resonant type. The vibrating arm MEMS gyroscope obtains the angular velocity by measuring the torsional vibration amplitude and the torsional vibration phase, and is typically represented by the ENV-05A series tuning fork type gyro. The vibrating disc MEMS gyroscope obtains the angular velocity by measuring the change in capacitance between the component and the bottom, typically represented by Honeywell's HG1940 inertial measurement unit. The ring-resonant MEMS gyroscope acquires angular velocity by measuring changes in the magnetic field, typically represented by the SiIMU02 gyroscope.

According to the way of sensing acceleration, MEMS accelerometers can be divided into displacement type, resonance type and electrostatic suspension type. Displacement MEMS accelerometers measure acceleration by detecting changes in capacitance, typically represented by SiACTM by Northrop Grumman. Resonant MEMS accelerometers measure acceleration by measuring changes in resonant frequency with high accuracy, typically represented by Honeywell's SiMMA. The electrostatic suspension MEMS accelerometer measures the capacitance to obtain the position of the disk or sphere in the suspended state to measure the acceleration. The theoretical precision is high. The typical representative is the SupersTAR accelerometer of ONERA of France.

According to the sensing principle, MEMS accelerometers can be divided into three types: piezoresistive, piezoelectric and capacitive. The piezoresistive accelerometer can convert the acceleration information into the electric signal output by converting the resistance on the corresponding cantilever beam into a voltage output, which has the advantages of small volume, simple processing technology, high precision, fast response speed, strong anti-electromagnetic interference and the like. . The piezoelectric MEMS accelerometer estimates the external acceleration by measuring the relationship between the internal varistor resistance change and the measured acceleration. It has a large measuring range, small weight, small volume, strong anti-interference ability, simple structure and high measurement accuracy. advantage. The capacitive MEMS accelerometer derives the external acceleration by detecting the amount of change in the capacitance value, and has the advantages of high measurement accuracy, high sensitivity, good stability, and low power consumption.

2 Overview of the development of MEMS inertial sensors

Since the development of MEMS gyroscopes and accelerometers, the performance of MEMS gyroscopes and accelerometer devices has been significantly improved with the development of MEMS technology.

2.1 Development history of foreign MEMS gyroscopes

In 1954, C.S. Smith discovered the piezoresistive effect and provided a theoretical basis for the development of miniature pressure sensors. In 1967, the surface sacrificial layer process technology was proposed, and on this basis, a cantilever beam technology with a high resonance frequency was developed. In 1989, Draper Laboratories of the United States developed the first vibrating micro-electromechanical gyroscope, which is a major revolution in the field of inertial technology. In 1993, the laboratory developed a tuning fork vibrating micro-electromechanical gyroscope that will be used in the gyroscope. Development has taken a big step forward. In 1997, at the University of California, Berkeley, the first surface micromechanical Z-axis gyroscope was designed with a resolution of 1°/s. In 1999, the Yokohama Technology Center proposed a MEMS gyroscope with a decoupling design with a resolution of 1°/h. In 2001, Draper Laboratories of the United States designed a single crystal silicon tuning fork MEMS gyroscope with a temperature drift of 1°/(h/°C). In 2002, the American adi company developed the world's first monolithic integrated commercial gyroscope ADXRS. In 2004, HSG of Germany designed a surface micromechanical X-axis gyroscope with a sensitivity of 8mV/(°/s). In 2006, K. Maenska of Hyogo University of Japan reported a novel piezoelectric vibrating solid-state micromachined gyroscope composed of only a lead zirconate titanate prism with electrodes. In 2013, the French Electronics and Information Technology Laboratory designed a 3D condenser tuning fork gyro with a lateral suspension design.

2.2 Development history of foreign MEMS accelerometers

In the late 1960s, research and development work on MEMS accelerometers began. The main research and development units were Draper Labs, Stanford University, and the University of California at Berkeley. In the 1970s, the integrated MEMS process and piezoresistive effect, the piezoresistive accelerometer appeared, and the commercialization of MEMS accelerometer was realized for the first time. In the late 1980s, with the combination of surface MEMS technology and sensing technology, capacitive MEMS accelerometers were rapidly developed and first used in the automotive industry. In 1989, the American ADI company developed an ADXL50 accelerometer with a 50g range. Since the 21st century, with the rapid development of the integrated circuit and computer industry, MEMS accelerometers are more used in automotive airbags, and play an increasingly important role in the electronic consumer industry such as mobile phones and computers. Future MEMS accelerometers will develop in the direction of lightweight, high precision and economy.

2.3 Development history of domestic MEMS inertial devices

The research on MEMS inertial devices in China started in the late 1990s. Since 1995, it has received strong support from the Ministry of Science and Technology, the Ministry of Education, and the National Natural Science Foundation of China. The development of domestic MEMS gyroscopes has achieved remarkable results. In 1998, Tsinghua University developed the nation's first tuning fork MEMS gyroscope with a resolution of 3°/s. In 2006, 49 Electronics Group and the Russian Institute of Applied Physics collaborated to develop a gyro with a resolution of 70°/h. In 2010, the State Key Laboratory of Sensor Technology of the Chinese Academy of Sciences reported a micromechanical vibration ring gyroscope with a high symmetrical structure. In 2012, Chun-Wei TsAI of Taiwan University produced a double decoupled micromachined gyroscope with a wide driving frequency. After more than 20 years of development, China's existing technology has formed a series of systems from design to production and testing. The accuracy of the devices of many famous MEMS inertial device companies in China has also been significantly improved.

Key technologies of MEMS inertial navigation

The software design of MEMS inertial navigation system is mainly navigation algorithm, including initial alignment, inertia solution and error compensation. Its hardware design mainly includes circuit and structure design, inertial navigation sensor (gyroscope, accelerometer) and navigation. Computer selection, etc. System accuracy is not only related to hardware, but also has a lot to do with software. Under the premise of slow development of hardware processing technology, the error compensation algorithm in the system is especially important. For applications with high navigation accuracy requirements, MEMS inertial navigation errors are prone to divergence due to the long-haul characteristics of the system, and combined navigation is often used to suppress the error divergence of the inertial navigation system. This section mainly introduces the error analysis and compensation of MEMS inertial sensors and the design of MEMS integrated navigation algorithms.

1 MEMS inertial sensor error analysis and compensation

The inertial sensor is the core component of the inertial navigation system. Its accuracy determines the accuracy of the inertial navigation system. Therefore, one of the main tasks of the inertial navigation system is to compensate the inertial sensor error. There are two ways to improve the accuracy of the inertial navigation system. The first one is to improve the accuracy of the inertial sensor from the process. However, this method is technically difficult and requires high processing conditions and materials. The second is to use error compensation. The way to compensate for the error of the system.

The error analysis and compensation methods of MEMS inertial sensors are roughly divided into three types: the first one is to compensate by means of error compensation algorithm, that is, the error is compensated by algorithm fitting method; the second is to use rotation modulation technology to make IMU ( The inertial measurement unit is rotated by a rotating mechanism to eliminate the constant error (called rotational modulation) by rotation; the third is to use Allan analysis of variance to compensate for the random error of the system.

1.1 Inertial sensor temperature error compensation technology

The inertial device accuracy error caused by temperature is mainly due to the sensitivity of the inertial device itself to temperature and the temperature gradient or the cross product term of the temperature and temperature gradient. With the change of temperature, the structural material of the inertial device will form the interference torque due to thermal expansion and contraction. Therefore, it is necessary to study the temperature characteristics of the inertial device to obtain the law of the influence of temperature on the output performance of the inertial device, and establish the static temperature model of the accelerometer. And compensation for errors caused by temperature changes is an effective means to improve its accuracy.

The method of fitting the static temperature model of the gyroscope and the accelerometer generally adopts the least squares method to obtain the relationship between the mathematical model coefficients of the gyroscope and the accelerometer and the temperature and establish a static temperature error compensation model, thereby improving the device precision. Many domestic gyroscope and accelerometer production units have studied the temperature error compensation, which reduces the static error of the product before compensation by an order of magnitude.

1.2 Rotational modulation technique for inertial sensor constant drift error

The rotary modulation technique was initially applied to an electrostatic gyro system, and the drift error torque was automatically compensated by the rotation of the housing. Since the introduction of the laser gyro, the United States has rapidly carried out research on the rotary inertial navigation system. In 1968, some scholars first proposed to compensate the drift error of the inertial sensor by rotating the IMU. In the 1970s, Rockwell developed an electrostatic gyro detector with a rotating technology that allowed the ship's associated ship system to have long-term accuracy. In the 1980s, Sperry developed a single-axis rotary inertial navigation system that used the classic single-axis four-position forward and reverse stop solution, which is still widely used until now. In 1989, NATO's standard inertial navigation system, the MK49 dual-axis rotary laser gyro inertial navigation system, was equipped on submarines and surface ships. At the National Defense University of Science and Technology, the application of rotary modulation technology to optical gyros was first started. Today, the rotary modulation technology mainly adopts a single-axis rotation scheme on MEMS, and the two-axis rotation scheme is relatively less applied due to the complexity of the rotating mechanism.

Due to the need of rotation, the navigation system adopts the strapdown algorithm. In principle, the rotational modulation of the MEMS inertial navigation system can effectively offset the system constant error. The error propagation equation of the system is as follows:

In equation (1), since the systematic errors caused by the gyroscope and the accelerometer's own measurement error are σωbib and σfb, the two errors of Cnbσωbib and Cnbσfb in the equation are introduced due to the measurement error, so the error compensation mainly compensates for this. Two errors. Since both of these include Cnb, periodically changing the Cnb value can eliminate these two errors. Therefore, a rotating device is applied to the inertial navigation system to cancel the periodic error by rotation. This is the rotation modulation technology to improve the accuracy of the inertial navigation system. The principle.

The rotary modulation scheme needs to determine the number of rotating axes (single-axis, two-axis or multi-axis), the rotation rate, the angular acceleration of rotation, the number of times of stoppage, and the number of stop positions. The difference between the static base and the transfer plan under the moving base will affect the rotation modulation effect.

1.3 Allan Variance Analysis of Random Error of Inertial Sensor

At present, the commonly used random error modeling methods include time series analysis, Allan variance method and power spectral density analysis.

Since the error equations of inertial navigation are based on the white noise, in reality, the various noises contained in the output data of the MEMS inertial device will cause interference to the system, resulting in random errors in the calculation results. The random noise in the error of the gyro output value needs to be modeled to compensate, and the Allan ANOVA is one of the most common and widely used methods in random noise analysis. The random errors in MEMS devices are mainly divided into angular random walk, acceleration random walk, quantization noise and zero offset stability.

The Allan method was proposed by David Allan in 1966 and is mainly used to analyze oscillator phase and evaluate frequency stability. The Allan variance can reflect the fluctuation of the average frequency difference between two consecutive sampling intervals. The Alan variance estimation based on phase data and frequency data is

2 MEMS integrated navigation algorithm

MEMS inertial navigation systems have the advantages of low cost, small size, and low power consumption. However, due to the low precision of MEMS inertial devices, long-term use will lead to faster error dispersion and can not serve as a long-term navigation task. Therefore, multi-sensor fusion is generally used for navigation, which is to fuse MEMS inertial navigation with other navigation methods. The navigation information of other navigation systems is used to correct the error of the inertial navigation system, thereby improving the accuracy of the entire navigation system. To perform data fusion for multiple navigation systems, use methods such as filtering.

2.1 Kalman filter algorithm

Kalman filtering is a filtering algorithm that estimates the amount of state by obtaining information from the extracted observed signals. Kalman filtering is a real-time recursive algorithm. The processing object is a random object. According to the system noise and the observed noise, the output of the observation value of the system is used as the filter input, and the state quantity to be estimated is taken as the output, that is, through the previous moment. The observations estimate the amount of system state at the next moment, so it is essentially an optimal estimation method.

Conventional Kalman filtering is suitable for linear Gaussian models, and most inertial navigation systems are nonlinear systems. Therefore, conventional Kalman filtering cannot meet the requirements, and a filtering algorithm suitable for nonlinear systems must be established. Therefore, an extended Kalman filtering method is developed, which linearizes the nonlinear function of the nonlinear system by Taylor series and saves the high-order term to obtain a linear system model.

Since the extended Kalman filter linearizes the nonlinear function, linearization errors are inevitably brought about, thereby developing an unscented Kalman filter. The filtering method approximates the probability density for a nonlinear function, and estimates the posterior probability density of the state using the determined samples, without approximating the nonlinear function. Compared to extended Kalman filtering, the statistics of the unscented Kalman filter not only have higher precision, but also have higher stability.

2.2 Complementary Filtering Algorithm

The traditional extended Kalman filter has a Jacobian matrix, which has the disadvantages of large computational complexity and white noise conditions cannot guarantee the time is established. However, the complementary filtering algorithm can reduce the calculation amount and improve the system measurement accuracy, and it does not need to be under white noise conditions. Can also be established. The complementary characteristics of the gyroscope and the accelerometer in the frequency domain can improve the data fusion precision of the gyroscope and the accelerometer, and achieve high-precision fusion.

2.3 Neural Network

Machine neural networks are based on biological neural networks. Neural network is a kind of machine learning. The model parameters are trained through the network system. The neural network is mainly composed of input layer, output layer and hidden layer. From MP neuron and Hebb learning rules in the 1940s, to the Hodykin-Huxley equation in the 1950s, to perceptron models and adaptive filters, to self-organizing mapping networks in the 1960s, neurocognitive machines, adaptive resonance Network, many neural network computing models have developed into a classic method in the field of computer vision, signal processing, and so on, with far-reaching effects.

The neural network has two kinds of forward neural networks and reverse neural networks. Neural networks have parallel processing, distributed storage, high redundancy, non-linear operations, and good fault tolerance. With the development of neural network technology, its application field is also expanding, and now plays a vital role in the fields of inertial navigation and image processing. Neural network algorithms have a wide range of theoretical foundations, including neural network structure models, network communication models, and memory models. The learning algorithm shows that the big data analysis based on neural network algorithm has good performance and application prospects, provides decision-making basis in sensor data fusion, and makes important contributions to the autonomous navigation of unmanned systems. Fuzzy neural network has superior performance in data fusion and data mining. It can make good use of language, and the form of knowledge expression is easy to understand. However, it has the disadvantages of weak self-learning ability and difficult to use numerical information. Therefore, artificial neural network and fuzzy system can be used. Combine.

MEMS inertial navigation application

MEMS inertial navigation technology is widely used in many unmanned systems, such as drones, unmanned vehicles, unmanned ships and robots, due to its small size, low power consumption, light weight and low cost.

1 drone field

In recent years, micro-miniature drones have played an increasingly important role in the military and civilian fields, and in order to achieve the positioning and positioning problems of the drones, the attitude monitoring and control system plays a vital role. The attitude measurement and control system is mainly composed of a GPS antenna, a GPS receiver board, a strap-down magnetic sensor, an inertial measurement unit, a high airspeed sensor, and a conditioning unit. The accuracy of the sensor directly determines the accuracy of the position of the drone. The data collected by the sensor calculates the position and attitude information of the drone through the navigation algorithm. At present, the navigation of the UAV mainly adopts the means of combining the MEMS inertial navigation system with the GPS, so that the accuracy of the system can be improved and the initial alignment time can be shortened. Nowadays, the accuracy of the navigation system mounted on the drone is consumer grade. For example, the accuracy of the Invensense MP6500 is 2°/s. With the improvement of the precision of the MEMS device and the cost reduction, the navigation accuracy of the UAV will be improved in the future.

2 Unmanned vehicle field

The unmanned vehicle senses the external environment through the on-board sensor, and acquires the vehicle position, posture information, and obstacle information, thereby controlling the vehicle traveling speed, steering, and starting and stopping. At present, companies such as Google and Baidu are developing the development of unmanned vehicles and have carried out road experiments. When an unmanned vehicle walks under a tall building and the GPS is blocked and cannot work normally, the accuracy of the inertial navigation system mounted on the unmanned vehicle in a short period of time can satisfy the demand of the vehicle itself. The MEMS inertial navigation system on the unmanned vehicle has high requirements for general accuracy.

3 unmanned boat area

Unmanned ship technology has developed rapidly due to the dangers and high cost of adopting ordinary ship equipment for tasks such as border patrols and water quality exploration. Obtaining the position and attitude information of the unmanned ship is an important prerequisite for the unmanned ship to carry out its work independently. The sensors currently installed on unmanned ships mainly include GPS, MEMS inertial navigation systems and obstacle avoidance radars. With the improvement of the accuracy of MEMS inertial navigation system, the inertial navigation system plays a vital role in the acquisition of position and attitude information of unmanned ships. The MEMS inertial navigation system mounted on the unmanned ship can meet the demand in general and low-precision consumption level.

4 robot field

A mobile robot is an automated device that can work autonomously in a fixed or time-varying environment. In recent years, it has been widely used in the fields of service industry, home and industry. Wheeled robots are similar in application to unmanned vehicles, and they collect data through sensors such as vision cameras, MEMS inertial sensors, laser radars, and odometers. Domestic universities such as National Defense University of Science and Technology, Tsinghua University, Shanghai Jiaotong University, Harbin Institute of Technology and other universities have started research on wheeled robots earlier. In the navigation process of wheeled robots with inertial sensors and odometers, MEMS inertial sensors provide precise attitude angles, and most of them are combined with visual odometers and MEMS inertial navigation due to the impact of wheel slip on inertial navigation and odometers. Navigation, by extending the Kalman filtering algorithm for data fusion, thereby improving system accuracy.

5 other fields

In addition to the above areas, MEMS inertial sensors are also used in electronic devices such as cell phones, tablets, game consoles, cameras, vr glasses, and individual navigation for indoor positioning. At present, the fire safety of firefighters in high-rise buildings and the personal safety of elderly people with limited mobility is a common concern in the society. If the MEMS inertial navigation system is placed on the detection personnel for navigation, real-time position and posture information can be obtained, which can improve The safety factor of the person being monitored. There are several ways to use the MEMS inertial navigation system for indoor personnel positioning: one is to use the MEMS accelerometer to detect and identify the state of the personnel's pace, and then use the magnetometer to detect the direction of the person's movement, thereby performing the directional positioning of the indoor personnel. Another method is to use two or more MEMS inertial navigation systems, installed at the foot and waist of the person, and positioned by multiple MEMS inertial navigation system correction methods.

Prospects for the development of MEMS inertial navigation

1 MEMS inertial navigation device

In recent years, MEMS inertial sensors have developed rapidly and their accuracy has been continuously improved. Although there is still a big gap compared with fiber optic gyro and laser gyro, its low price, small size and light weight make MEMS inertial navigation system play an important role in the inertial navigation system. In the future, with the continuous development of MEMS material process and manufacturing process, the accuracy of MEMS inertial navigation system will continue to increase, and its cost will continue to decrease. Therefore, the replacement of fiber optic gyroscope with strategic high-precision MEMS gyroscope is an important development trend. With the continuous advancement of micro-machining technology, MEMS inertial sensors will develop in the direction of light weight and miniaturization.

2 MEMS integrated navigation algorithm

Although the accuracy of MEMS inertial sensors is constantly improving, the error of tactical MEMS inertial navigation system is still diverging with time. In many cases, the requirements of high precision cannot be met. Therefore, MEMS inertial navigation and GPS integrated navigation are still the main navigation methods. Therefore, it is also an important development direction to study the accuracy and efficiency of the algorithm with higher robustness and more support for integrated navigation system in software.

3 MEMS inertial navigation applications

In the decades of MEMS technology development, MEMS inertial navigation technology has been widely used in the electronics, automotive, and home service industries. With the increasing accuracy and stability of MEMS inertial navigation, the future MEMS inertial navigation technology will play an important role in the unmanned system, such as spacecraft, satellite, robot and other unmanned systems.


MEMS inertial navigation technology has the advantages of miniaturization and low cost. It has developed rapidly in the past few decades and has been used more and more in the field of unmanned systems. As the main development direction of future inertial navigation, it is showing Strong potential and good application prospects. This paper reviews the development history of MEMS inertial navigation system, summarizes its key technologies, and forecasts the application and development of MEMS inertial navigation technology, and provides reference for the research of MEMS inertial navigation system.