Topics covered in this article:
The flight sensing modules of drones are mainly used for two purposes. One is to provide the flight control system. Since the main function of the flight control system is to control the drone to reach the desired attitude and spatial position, this part of the sensing technology mainly measures state-related physical quantities. The modules involved include gyroscopes, accelerometers, magnetic compasses, barometers, GNSS modules, and optical flow modules.
Another purpose is to provide autonomous navigation systems for drones, that is, path and obstacle avoidance planning systems. So it needs to perceive the state of the surrounding environment, such as the location of obstacles. Related modules include ranging modules, object detection, and tracking modules, etc...
At present, MEMS technology gyroscopes are commonly used in commercial drones because of their small size and low price. They can be packaged in the form of ICs. MEMS gyroscopes are used to measure the angular rate of the body rotating around its own axis. Commonly used models are 6050A (Invensense) and ADXRS290 (ADI). The indicators for measuring gyroscope performance include measurement range, sensitivity, and stability (drift) and the signal-to-noise ratio, and so on.
test results of the temperature drift of the gyroscope
The above is a graph of the test results of the temperature drift of the gyroscope. The test environment is heated from 25°C to 50°C. The gyroscope is kept still during the whole process, and the accurate output of the gyroscope should be a fixed value. But judging from the results, the actual output of the two sensors is affected by temperature changes. In comparison, the output value of ADXRS290 (ADI) has a small change range, basically around 0.5.
The accelerometer measures the linear acceleration of the body's motion. Due to the gravity of the earth, the measured value will also include the gravitational acceleration component, which needs to be subtracted in some use cases. Commonly used MEMS accelerometer sensor models are 6050A (Invensense) and ADXL350 (ADI). In order to improve chip integration, some sensor manufacturers package gyroscopes and accelerometers together, called six-axis sensors, such as 6050A (Invensense).
The physical quantity measured by the magnetic compass is the component of the earth's magnetic field strength along the axis of the airframe, and the heading angle of the airframe is calculated based on this. Commonly used MEMS magnetic compass sensor models are HMC5983L (Honeywell) and QMC5883L (Silicon Core), both of which have similar performance. The former has been discontinued. The main performance parameters of the magnetic compass include sensitivity, stability (drift), etc.
The physical quantity measured by the barometer is the atmospheric pressure value, from which the absolute altitude can be calculated. Commonly used barometer sensor models include MS5611 (MEAS), MS5607 (MEAS), and BMP180 (Bosch) [check this product's detail]. The problem with the barometer is that when flying near the ground, the "ground effect" will cause the air pressure distribution around the aircraft to be different from the atmosphere at rest, making it impossible to use the barometer to measure altitude. The usual solution is to use other sensors during takeoff or landing, such as ultrasonic sensors or laser rangefinders.
The physical quantities measured by the GNSS module are relatively rich, mainly including geographic coordinates (latitude and longitude), altitude, linear velocity, and heading angle (RTK system). Commonly used GNSS module manufacturers include U-BLOX of Switzerland and NOVATEL of Canada. When using the GNSS module, the placement of the satellite signal receiving antenna needs to avoid the shielding of electromagnetic interference. Some powerful complete machine manufacturers will customize the satellite signal receiving antenna according to the aircraft model.
The optical flow module is a special module that can be used to sense the motion state of the body, such as measuring the displacement speed in the horizontal direction, or it can be used to sense the surrounding environment for obstacle avoidance purposes. The more common optical flow module is the open-source PX4FLOW. Optical flow modules are usually used indoors, mainly to solve the problem of poor indoor satellite signals. In addition, a certain texture pattern is required for the ground to be photographed.
Here are five commonly used ranging modules: ultrasonic, infrared TOF, lidar, millimeter-wave radar sensor, and depth-sensing camera.
|Ultrasonic measurement||Infrared ToF||Laser measurement||Depth-sensing camera||Millimeter-wave radar|
Ultrasonic and infrared TOF have similar performance in all aspects. For example, the measurement distance is relatively close. The distance for ultrasonic measurement is generally about 4 meters. In addition, the use range of these two sensors is easily limited by the actual environment. For example, infrared TOF emits red light to the surface of the measured object and reflects it. If it encounters an object with low red light reflectivity like glass, it will fail. But these two kinds of sensors have one of the biggest advantages of low cost, and the module size is relatively small, so they have been widely used in consumer drones.
The measurement distance of lidar ranging is far enough. Most products can reach more than 100 meters, but the weather environment of heavy rain and fog will affect its measurement results. Another disadvantage is that the cost is relatively high: Velodyne is the strongest in the lidar industry. Its VLP-16, a miniaturized product suitable for drones, has also reached a price of more than $1,000. The cost is still relatively high.
Depth-sensing cameras can be divided into three types according to measurement technology. ①Stereo cameras are also called binocular vision technology. The representative product of this technology is DJI's Wizard 4. ②Structured light technology. It represents Microsof’s Kinect. ③ Time of flight technology (TOF). Because there are fewer manufacturers and higher costs, there are few applications on UAVs.
Depth-sensing cameras also have limitations when used. The disadvantage of binocular vision technology is that it cannot work normally in low-light environments, while structured light technology, on the contrary, cannot work normally under strong light. Therefore, some manufacturers combine the two technologies to make up for each other's shortcomings and expand the scope of their applicable environment.
Sensor calibration, including fine calibration and rough calibration. The fine calibration is better, but it requires expensive calibration equipment; the rough calibration does not need to rely on external equipment, only the sensor itself can be operated.
Taking the rough calibration of a magnetic compass as an example, since the geomagnetic field intensity at any position on the earth can be regarded as constant over a long time span when the magnetic compass is rotated, it can be assumed that the magnetic compass is fixed according to the relative motion. The geomagnetic field vector rotates accordingly, and the trajectory of the vector endpoint in space should be a standard sphere. However, due to the error of the sensor, the actual measured data is not strictly on the surface of the sphere. At this time, it needs to be based on the measured data. The numerical value and the known accurate value are used to calculate the conversion relationship between the two, which is the error model of the magnetic compass. When using this magnetic compass in the future, the measured value can be processed according to the error model obtained from the rough calibration, so that the error of the measured value is reduced.
Magnetic compass calibration (SGB sbgcenter)
There are many different types of sensor data fusion methods. The most common method used in the industry is EKF, which is the extended Kalman filter.
Taking the fusion method of calculating the aircraft attitude angle as an example, the EKF update process is mainly divided into two parts, prediction update and measurement update. The forecast update mainly uses the gyroscope to update the forecast state quantity, and at the same time calculates the covariance matrix of the state quantity. In the measurement update, the filter gain is calculated first, and then the filter gain is used to fuse the data of the predicted state quantity, accelerometer, and magnetic compass to become a fusion state quantity. At the same time, the covariance matrix of the fusion state quantity is calculated, which will be calculated in the next update cycle.
Fusion method flow for calculating attitude angle
The sensor redundancy design mainly combines multiple sensors of the same type. The processing method is to first remove the sensors with abnormal data, and then perform sensor fusion. The redundant design can not only improve the measurement accuracy but also improve the reliability of the entire system. When a certain sensor fails, the entire system can continue to work normally.