Combined navigation system for substation inspection robot based on DGPS/DR multi-sensor fusion

To realize the routine inspection work of the capacitor tower inspection operation robot on the capacitor tower in the substation, and to guarantee the accuracy and safety of the robot’s movement in the substation, a multi-sensor combination navigation system and the corresponding Kalman fusion filter are designed. Combined with the overall structure of the substation and the working environment of the inspection robot, the navigation and positioning requirements of the robot are analyzed and compared, and the multi-sensor combination positioning technology using a combination of DGPS, attitude sensors and velocity sensors is determined. The Kalman fusion filter based on the constant velocity model is used to fuse the positioning information of multiple sensors, and more accurate and reliable navigation information is obtained compared with the single-sensor positioning system. According to the working environment of the substation, an outdoor planar robot mobile navigation experiment is carried out to collect the positioning information of multiple sensors and carry out data fusion processing to verify the accuracy and robustness of multi-sensor combination navigation and positioning.


Introduction
The power industry plays a crucial role as one of China's basic energy industries.Safety concerns necessitate regular maintenance and inspection of substation capacitors and other operational tasks.In proposing solutions, automation and intelligent technical means have been considered.A high-voltage substation maintenance robot is thus recommended to efficiently and reliably complete the above tasks, replacing manual labor [1][2][3].During the automatic operation of the capacitor tower maintenance robot, the robot must carry out positioning, navigation, and movement within the grid array of the capacitor tower site.To ensure precise and secure movement, the robot must continuously obtain and process realtime information about its condition and surrounding environment, including navigation and positioning data.This ensures the inspection robot can reach the operation target accurately and steadily.
Currently, several navigational methods exist for robots including black-and-white line trajectory navigation, magnetic signal navigation, and visual image navigation.Each has distinct pros and cons, but black-and-white line trajectory and magnetic signal navigation are vulnerable to the external conditions surrounding the substation and electromagnetic interference.Visual image navigation incurs significant computational expenses and poses challenges for industrial implementation [4][5].
The paper employs a DGPS/DR multi-sensor combination for navigation and localization.Electromagnetic interference is present in the substation.Simultaneously, the abundant distribution of utility poles causes obstruction, which may lead to interruptions and discontinuities in the GPS base station antenna transmission to the mobile station differential signal.As a result, GPS navigation is susceptible to random interference, leading to positioning accuracy limited to only a meter level [6][7].Therefore, it is insufficient to rely solely on a singular GPS navigation system in the power station equipment area due to the need for precise and dependable continuous operation in the inspection robot substation.As an integral aspect of navigation, trajectory projection employs inertial navigators and odometer sensors to track the displacement and angle of the robot over a certain period.This methodology serves as a means of estimating position.The current position is determined through analysis of the mobile robot's previous positional velocity and orientation.This method allows for high accuracy of continuous localization in a short period but generates a significant error offset over time [8][9].Both navigation algorithms possess complementary strengths and weaknesses, making it feasible to merge them to accrue the benefits of both techniques, thus providing uninterrupted, high-bandwidth, and highly accurate navigation parameters over both extended and short-time intervals.Multiple sensors are typically employed in integrated navigation to acquire localization data.To achieve integrated multisensor navigation, it is essential to apply data fusion and filtering methods, particularly when dealing with sensors of varying accuracy [10].

Multi-sensor combined navigation system design
The multi-sensor combination navigation experimental system comprises DGPS (differential global positioning system), IMU (inertial measurement unit), odometer, signal acquisition board, CAN communication bus, main controller, robot system, and upper computer.The hardware composition is illustrated in Figure 1.The experimental system was tested by using a mobile platform in the form of a capacitive tower-operating robot.The navigation system on the platform operates as follows: multiple sensors gather localization data from different sources and transmit it to the signal acquisition board through a serial communication protocol.The signal acquisition board then filters and processes the signals before sending them on to the main controller via the high-speed CAN field bus.Finally, the main controller utilizes this data to issue motion commands to the robot system and host computer for visualization.The motion commands are given by the main controller to the robot system, which sends them to the upper computer for visualization.The DGPS module and INS inertial navigation system function independently in this system.The position and velocity of the carrier are outputted when the received signal is processed through RF front-end, baseband digital signal processing, and navigation and localization solving, and the output rate amounts to 10 Hz.The inertial navigation system operates at a high speed, with the IMU measuring the carrier's acceleration and angular rate and compensating for any errors.Furthermore, the IMU continuously gauges the acceleration and angular rate data of the carrier to acquire position, velocity, and attitude data at a high rate of 1000 Hz following initial alignment and mechanical arrangement solving through inertial guidance.

DGPS positioning system
The DGPS utilized in this investigation is a dual-antenna high-precision differential positioning and orienting module, autonomously developed by Beili Electronics.It is founded on a recently designed, domestic high-performance GNSS SoC chip (UM982 inside); this device is capable of supporting multisystem and multi-frequency RTK positioning, dual-antenna high-precision orienting, and BeiDou navigation and positioning.A complete RTK system consists of two GRTK modules, one for the mobile side and one for the base station side, as shown in Figures 2 and 3.The base station and mobile station communicate via a separate link.This link employs a 433 M serial port for GFSK wireless data transmission, while the DGPS module utilizes a TTL serial port configuration to output NMEA protocol positioning data at a 10 Hz rate.The output localization data is provided in formats such as GPGGA, GPRMC, GPHDT, and KSXT.In this study, the primary data source utilized is GPGGA, which enables the robot's real-time three-dimensional position information, ground rate, and ground heading to be obtained directly.Considering that the capacitor tower maintenance operation robot is a two-dimensional planar motion inside the substation, the data processing part pre-converts the original WGS-84 coordinate system obtained from the GPS receiver to under the right-angle coordinate system.We define the station center right-angle coordinate system origin   is located in the observation station, the axis coincides with the ellipsoid normal of the   point, the   axis is perpendicular to the   axis pointing to the short axis of the ellipsoid, and the   axis is perpendicular to the       plane, which constitutes the lefthanded coordinate system.

Inertial measurement unit
Currently, robots predominantly rely on MEMS inertial measurement units (IMU) to acquire the realtime angular velocity of the robot.Subsequently, the angular velocity is integrated to determine the corresponding real-time attitude angle of the robot's body [11].The Yesense YIS10 attitude sensor, as shown in Figure 4, which utilizes the Y-fusion algorithm in this study, employs the characteristics of the sensor signal to minimize angular drift caused by a cumulative error in real time.As a result, it achieves an attitude angle accuracy of 0.5° and a heading angle accuracy of 1° after joint error correction.Following secondary filtering, the obtained data can be directly applied to control the attitude of the capacitor tower maintenance robot.When implemented, the attitude sensor is mounted parallel to the vehicle body.The robot's primary controller obtains attitude information output via serial communication at a high update frequency of 100 Hz.This guarantees the projection of the trajectory with high accuracy.

Wheeled odometer
In this paper, the researchers use a wheel with an encoder as an odometer to track the robot's movement, as depicted in Figure 5.The odometer runs parallel to the rear wheel of the robot, and the encoder outputs the angle of the wheel's rotation to calculate the distance and speed of the robot's motion.The encoder boasts a precision of 13 bits, thereby enabling millimeter-level high-precision displacement information and improving the localization accuracy of the multi-sensor combination navigation [12].

Kalman filter construction
The Kalman filter is a linear filter comprising a series of recursive mathematical equations.This filter provides an efficient and reliable method to estimate the state, minimizing the estimation mean square error [13].The Kalman filter, along with its related algorithms, has a broad range of applications in multi-sensor information fusion [14][15].The fused Kalman filter has been implemented in the multisensor combined navigation system adopted to achieve effective and trustworthy navigation and localization.
Robot operation site substation site leveling, the traveling speed of the capacitor tower maintenance operation robot is slow and constant.Based on this condition, the following definitions are made: We define the state space vector   = [        ], where   and   are the x-and y-direction coordinates under the GPS data transformed into the station-centered rectangular coordinate system;   is the angular component of the capacitance tower maintenance operation robot, with the horizontal axis positively oriented at 0, and positively oriented in the counterclockwise direction; and   is the traveling speed of the capacitance tower maintenance operation robot in the longitudinal direction.According to the constant velocity model, the prediction matrix  +1 is obtained The external observation vector   = [   ] and the measurement matrix H between the external observation vector and the state vector is a constant matrix: The measured noise variance array is: (3 where  1 2 ,  2 2 ,  3 2 , and  4 2 denote the variance of the GPS receiver x-and y-direction localization, attitude sensor, and velocity sensor measurement noise, respectively.

Kalman filter workflow and improvement
Based on the establishment of the Kalman filter state and observation equations, the recursive system of difference equations for a linear discrete Kalman filter can be expressed as shown in Figure 6 below: If the GPS differential signal becomes lost, the accuracy of the GPS receiver's positioning significantly decreases.Hence, it is crucial to assess the consistency between the current sensor measurement data and the filter's predicted value for each correction calculation.If the consistency criteria are met, the observed data is used to rectify the prediction.In the case of non-fulfillment of the consistency criterion, the observation data gets discarded, and the prediction gets directly replaced in the next recursive loop without any modification.
The Horizontal Dilution of Precision (HDOP) coefficient of the GPS receiver output indicates the level of horizontal positioning accuracy of the current GPS receiver.If the HDOP value is large, the positioning error is also large, and if it is small, the positioning error is also small.This study automatically adjusts the size of R and Q based on the output parameters of the GPS positioning system, such as HDOP, to adaptively adjust the performance of the combined navigation system model.

Results
The Kalman filter fusion algorithm is used to test the adopted multi-sensor combined navigation system in real experiments.The system software design is developed based on the Windows QT C++ platform, including three modules data acquisition, data processing, and data sending.The data acquisition module obtains the localization packet information of multiple sensors from the stm32 signal acquisition board via wireless LAN through serial communication.The data processing module inputs the navigation data from the lower computer into the established Kalman filter for data fusion processing to obtain the accurate navigation and localization information of the robot.The data-sending module can visualize and present the fusion-computed robot localization information on the host computer, and at the same time send it down to the motion control main controller mounted on the robot to perform the next motion task.
Firstly, a section of the moving path experiment in the substation is simulated on the outdoor flat ground, so that the robot can complete a section of right-angle turning action.The data results obtained for the two experiments are shown in Figure 7 below.The blue trajectory is the set trajectory of the tracking performed by the robot, the green scatter data is the raw data from the DGPS receiver, and the black path is the robot's movement trajectory computed based on the trajectory extrapolation method combined with the inertial measurement unit (IMU) and the wheeled odometer.In both experiments, the data from the DGPS localization system presents a scattered distribution around the actual value, which is affected by environmental interference, and the data is more randomly distributed and noisy.It can be seen that, in the robot path calculated according to the trajectory projection method, the heading angle error is small in the early stage, but with time, the yaw error caused by the integral calculation accumulates, and at the same time, the direction of error accumulation is uncertain, which leads to the different offset directions presented in the two experiments.From the above experimental results, it can be seen that it is difficult to achieve a more accurate positioning effect by using DGPS positioning information or the DR trajectory projection method alone, and the two methods have their advantages and disadvantages.According to the characteristics of the two schemes, we fused the experimental data.The experimental results of the fusion of DGPS and trajectory projection algorithm are shown in Figure 8 below, the red part is the Kalman filtering results, which can be intuitively seen that the fused experimental results are basically close to the setup path, which is a more effective enhancement compared to the effect of the two localization schemes alone, eliminating the random noise generated by DGPS due to the environmental interference, and at the same time making up for the random noise caused by the accumulation of time in the DR trajectory projection method.It eliminates the random noise caused by DGPS due to environmental interference, compensates for the cumulative heading error caused by time accumulation of DR trajectory, complements the two navigation and localization schemes, and obtains the robot localization information that is more accurate in heading information and more convergent in position information through the fusion algorithm.

Conclusion
During the movement of the maintenance robot for the capacitor tower, we developed an adaptive Kalman filter model that is based on the constant velocity model.This model is used for real-time data fusion of the multi-sensor combination navigation system.This fusion system smooths the DGPS localization data by combining DGPS signals with the inertial measurement unit, odometer, and other multi-sensor systems.We found this fusion essntial in effectively eliminating signal noise and instability in the DGPS dynamic localization.The system corrects the accumulated inertial measurement unit error in real time.According to the experimental results, the creation of a multi-sensor navigation system with a Kalman fusion filter can enhance robot positioning accuracy.This yields more precise robot heading and position information, thereby achieving sub-meter range positioning accuracy which can be stably maintained.