Research on AGV real-time location based on low cost IMU and RFID technology
1 Introduction
With the development of intelligent manufacturing, China's industrial production will further develop in the direction of intelligent, flexible and highly integrated called AGV as a mobile robot to realize the automatic transportation of materials in intelligent factories, real-time positioning is the key to the accurate distribution of materials.
In recent years, a variety of indoor positioning methods based on infrared, ultrasonic, Bluetooth, radio frequency (RFID), ultra-wide band (UWB) and other technologies have been rising, widely used in smart home and other fields, and have high positioning accuracy. Literature reports an accurate positioning method of multi-vision and laser combined navigation, but the cost is relatively high. In the literature, ultrasonic sensors installed on both sides of the AGV are used to realize its positioning, but this method is seriously affected by visual distance. The above positioning technologies all present such problems as weak independence and visual distance transmission requirements, which need to be assisted by external devices and hardware devices for data transmission. Once the external environment cannot satisfy the conditions, the positioning methods will be invalid and reluct. For complex factory processing environment, the above methods are difficult to achieve accurate positioning, and the inertial navigation positioning method based on its own sensor positioning becomes the key to solve the problem.
Inertial positioning does not rely on external information, nor does it need radiation energy. Because of its good independence, it has absolute advantages in the face of complex indoor environment. But mature inertial navigation devices on the market cost is very high, if the use of low-cost sensors, it will collect the data of zero static acceleration is serious, motion acceleration noise is large, different measurement speed measured deflection Angle accuracy is different and calculated displacement drift serious problems; In addition, using the inertial positioning method alone to calculate the position of AGV, the errors obtained will have a cumulative effect over time. After a certain time, positioning errors will exceed the acceptable range and positioning will lose significance. In order to solve the above problems, two methods are used: one is to refer to the rail transit widely used transponder method, using RFID technology to inertial navigation positioning position correction, through the combination of the two methods of data to solve the AGV position, to solve the cumulative error of inertial navigation; Second, make a set of real-time positioning error correction method based on low-cost IMU, alleviate the low cost inertial sensor measurement error in the test and other problems, make the positioning results meet the requirements of industrial indoor positioning.
2 Indoor positioning system model and improvement
2.1 Principle of inertial positioning
The advantage of the strapdown inertial navigation system lies in eliminating the complicated fixed inertial platform and installing the sensor on the moving body. A strapdown inertial measurement unit includes an accelerometer and a gyroscope to track the translation and rotation of the moving body. For AGVs in the workshop environment, two-dimensional planar motion is mainly considered. The acceleration of the AGV's movement (in the X and Y directions) is collected by the accelerometer, and the yaw Angle of the AGV (the Angle of rotation about the Z axis) is collected by the gyroscope.
In the positioning model, AGV is taken as the carrier coordinate system b, and the geographic coordinate system is taken as the navigation coordinate system n, and the acceleration information of AGV in the carrier coordinate system b is measured by INS(inertial navigation system)
Theta k yaw Angle information, according to the attitude algorithm method for AGV under navigation coordinate system n acceleration information as well as the Angle deflection | 8].
Acceleration information under the system; ψ k-agv deflection Angle ψk= θk-θ0 if the initial Angle is θ0
Through the above transformation, the acceleration information in b series is converted to that in n series.
2.2 Pose update algorithm based on Kalman filter
The data collected by IMU is doped with noise, and the real-time location of AGV requires the algorithm to process the generated data in real time. The real-time performance of the commonly used filtering algorithm cannot be guaranteed. Kalman filtering algorithm (UKF), as a pure time domain estimation algorithm, solves this problem well.
Equation of state: In the INS positioning system, the position and velocity updates of AGVs in the n system are obtained by the information obtained by the sensors in the b system. The model is as follows:
Where: a plus acceleration; ak, the acceleration at time k and K-1 under the AK-1 -- n system; Velocity at Vk, vK-1 -- k and k-1; Displacement at Sk, SK-1 -- k time and k-1 time; △T -- Sampling interval.
UKF equation of state :Xk = AXk-1 +Bkuk (4)
UKF measurement equation :Zk =HXk +Vk (5)
K moment state: xk= [Sk,x Sk,y Vk,x ak,x Vk,y]
According to the pose updating model and the equation of state, the state transition matrix is as follows:
