CN-122015830-A - Multi-sensor data fusion positioning navigation system for intelligent campus unmanned vehicle

CN122015830ACN 122015830 ACN122015830 ACN 122015830ACN-122015830-A

Abstract

The invention relates to the technical field of urban planning management, in particular to a multi-sensor data fusion positioning navigation system for an intelligent campus unmanned vehicle, which is used for collecting vehicle-mounted multi-line laser radar point cloud, IMU inertial navigation data and GNSS satellite positioning data, generating a three-dimensional point cloud map by adopting an improved SLAM algorithm and converting the three-dimensional point cloud map into a two-dimensional grid map when a map is established, adaptively selecting a positioning mode when navigating, solving an optimal pose by the multi-sensor data fusion algorithm, generating mixed navigation data comprising a global path sequence and a local obstacle avoidance track by a path planning module based on the pose, the map and a target point, constructing a complete navigation closed loop from environment perception to motion execution by the process through multi-source data depth fusion and time-space synchronization, effectively aiming at positioning drift and dynamic obstacle in a complex campus scene, and providing reliable technical support for high-precision autonomous navigation of the intelligent campus unmanned vehicle.

Inventors

MA YUNQIANG
GAN QUAN
MA MINGJIE
WANG LEI

Assignees

安徽机电职业技术学院

Dates

Publication Date: 20260512
Application Date: 20260320

Claims (10)

1. The multi-sensor data fusion positioning navigation system for the intelligent campus unmanned vehicle is characterized by comprising a data acquisition module, a preprocessing module, a map building and positioning module, a path planning module and a motion control module; The data acquisition module is used for acquiring environment point cloud data output by the vehicle-mounted multi-line laser radar, inertial navigation data output by the inertial measurement unit and satellite positioning data output by the global satellite navigation system in real time; the preprocessing module is configured to receive environmental point cloud data, inertial navigation data and satellite positioning data, perform time point synchronous alignment on the data according to a preset time reference, unify the sensor data under a vehicle body coordinate system based on a pre-calibrated external parameter matrix, and output multi-source sensor data streams after space-time registration; The map building and positioning module is used for receiving the multi-source sensor data flow, generating a three-dimensional point cloud map by utilizing an improved simultaneous positioning and map building algorithm in a map building stage, converting the three-dimensional point cloud map into two-dimensional grid map data, adaptively selecting a positioning mode according to the current environmental characteristics in a navigation stage, and calculating optimal pose estimation data of the unmanned vehicle in the two-dimensional grid map data based on a multi-sensor data fusion algorithm; The path planning module generates mixed navigation data comprising a global path sequence and a local obstacle avoidance track according to the optimal pose estimation data, the two-dimensional grid map data and the target point data; the motion control module receives the hybrid navigation data, and converts the hybrid navigation data into a chassis motor driving instruction through the calculation of a kinematic model to control the unmanned vehicle to run.
2. The intelligent campus unmanned vehicle-oriented multi-sensor data fusion positioning navigation system of claim 1, wherein the mapping and positioning module performs the following data processing steps when generating two-dimensional raster map data: Acquiring the environmental point cloud data after space-time registration, and removing outlier noise data by using a hybrid filtering algorithm; performing ground point segmentation on the filtered point cloud data based on a preset height threshold value and a gradient threshold value, and separating out ground point cloud data and obstacle point cloud data; projecting the obstacle point cloud data to a two-dimensional plane, and counting the point cloud density characteristic values in each grid unit in the projection process; And comparing the point cloud density characteristic value with a preset occupancy probability threshold value, marking the corresponding grid as an obstacle occupancy state when the characteristic value exceeds the threshold value, and generating three-dimensional to two-dimensional grid map data containing occupancy state information, wherein the two-dimensional grid map data is configured as a basic map layer for subsequent navigation path searching.
3. The intelligent campus unmanned vehicle-oriented multi-sensor data fusion positioning navigation system according to claim 1, wherein the map building and positioning module comprises an environmental state judging unit, wherein the environmental state judging unit is used for monitoring covariance matrix eigenvalue of the satellite positioning data in real time; when the covariance matrix eigenvalue is smaller than a preset precision threshold, judging an outdoor environment mode, and enabling a system to activate a first fusion positioning channel optimized based on a factor graph; And when the covariance matrix eigenvalue is larger than or equal to a preset precision threshold, judging an indoor or weak signal environment mode, and enabling the system to activate a second fusion positioning channel based on self-adaptive particle filtering.
4. The intelligent campus unmanned vehicle-oriented multi-sensor data fusion positioning navigation system of claim 3, wherein the processing procedure of the first fusion positioning channel on the data comprises the following steps: Constructing a factor graph model comprising the unmanned vehicle pose vertexes; adding the satellite positioning data into the factor graph model as an absolute position constraint factor; Pre-integrating the inertial navigation data to generate a relative motion constraint factor and adding the factor graph model; Performing scanning matching on the environment point cloud data and the three-dimensional point cloud map to generate a laser odometer constraint factor and adding the factor map model; And carrying out nonlinear optimization solution on pose vertexes in the factor graph model by minimizing a joint error function of all factors, and outputting outdoor high-precision pose data as the optimal pose estimation data.
5. The intelligent campus unmanned vehicle-oriented multi-sensor data fusion positioning navigation system of claim 3, wherein the processing procedure of the second fusion positioning channel on the data comprises the following steps: initializing a particle set, and updating prior pose distribution of each particle in the particle set by using the inertial navigation data as prediction input; performing likelihood domain matching on the environmental point cloud data of the current frame and the two-dimensional grid map data, and calculating a weight value of each particle; resampling operation is carried out according to the particle weight value, low-weight particles are removed, and high-weight particles are duplicated; and calculating a weighted average value of the resampled particle set, and outputting indoor pose estimation data as the optimal pose estimation data, wherein the indoor pose estimation data and the inertial navigation data form a closed-loop correction relation.
6. The intelligent campus unmanned vehicle-oriented multi-sensor data fusion positioning navigation system of claim 1, wherein the path planning module comprises a global path planning unit that performs the following operations: in the two-dimensional grid map data, the current optimal pose estimation data is taken as a starting point, target point data is taken as an end point, and an improved A star algorithm is utilized to search and generate initial path data comprising a series of discrete road points; extracting turning key point data in the initial path data, and performing smooth interpolation calculation on road sections before and after the key point data by using a Bezier curve algorithm; Global optimum path data with continuous curvature is generated and transmitted to a local path planning unit.
7. The intelligent campus unmanned vehicle-oriented multi-sensor data fusion positioning navigation system of claim 6, wherein the path planning module further comprises a local path planning unit that performs the following operations: constructing local cost map data according to the environmental point cloud data, wherein the data reflects dynamic and static obstacle distribution around the unmanned vehicle in real time; sampling in a speed space based on a dynamic window algorithm to generate a plurality of groups of simulation speed pairs, and deducing a plurality of pieces of corresponding predicted track data; Each piece of predicted track data is scored by constructing an evaluation function, and the evaluation function comprehensively considers the fitting degree of the track and the global optimal path data, the distance between the track and an obstacle in the local cost map data and the current running speed; and selecting a speed pair corresponding to the predicted track data with the highest score, and outputting the speed pair as local control instruction data comprising the linear speed and the angular speed.
8. The intelligent campus unmanned vehicle-oriented multi-sensor data fusion positioning navigation system according to claim 1, wherein the preprocessing module takes acquisition and transmission time of the environmental point cloud data as a reference time axis when executing time point synchronous alignment; for the inertial navigation data with higher frequency, calculating an inertial data value corresponding to the moment of the reference time axis by adopting a linear interpolation method; for the satellite positioning data with lower frequency, adopting a nearest neighbor matching method to correlate a satellite positioning data frame closest to the moment of a reference time axis; And packaging the synchronized data into a multi-sensor data frame containing a uniform time tag, and ensuring that the data processed by a subsequent fusion algorithm are in the same time section.
9. The intelligent campus unmanned vehicle-oriented multi-sensor data fusion positioning navigation system according to claim 1, wherein the system adopts a cloud edge end cooperative architecture for data interaction and processing: The three-dimensional point cloud map construction and global path planning calculation in the map construction and positioning module is deployed on a cloud server, a large-scale map data optimization task is processed by cloud computing power, and optimized global map data are issued; The real-time pose resolving and local path planning computing in the positioning module are deployed on a vehicle-mounted edge computing terminal, and the vehicle-mounted edge computing terminal utilizes locally acquired sensor data and global map data issued by a cloud to carry out real-time matching and obstacle avoidance decision; the motion control module is deployed at the vehicle chassis execution end and directly responds to the control instruction output by the edge computing terminal.
10. The intelligent campus unmanned vehicle-oriented multi-sensor data fusion positioning navigation system according to claim 9, wherein the motion control module is embedded with a differential kinematics model which decouples the linear velocity and the angular velocity in the local control instruction data; calculating target rotation speed data of the left driving wheel and the right driving wheel according to the wheel spacing parameter and the wheel radius parameter of the unmanned vehicle; and utilizing a PID control algorithm to adjust the motor voltage in a closed loop manner, so that the actual wheel rotating speed follows the target rotating speed data, and driving the unmanned vehicle to travel along a planned path.

Description

Multi-sensor data fusion positioning navigation system for intelligent campus unmanned vehicle Technical Field The invention belongs to the technical field of urban planning management, and relates to a multi-sensor data fusion positioning navigation system for an intelligent campus unmanned vehicle. Background The multi-sensor data fusion positioning navigation technology for the intelligent campus unmanned vehicle has important significance and necessity for realizing accurate, continuous and robust autonomous operation in a semi-structured environment of dynamic, man-vehicle mixed running. The intelligent campus scene has indoor and outdoor communication areas, dense building groups and high-frequency people stream activities, and a single sensor such as GNSS is prone to be blocked and invalid, the laser radar is affected by weather, and the vision sensor is interfered by illumination, so that multi-source data such as the laser radar, the IMU and the GNSS can be fused, stable pose estimation can be provided under various conditions through advantage complementation, and the intelligent campus scene is a basic stone for guaranteeing unmanned vehicle safety obstacle avoidance, path planning and high-efficiency completion of logistics, connection and other tasks. However, the prior art has obvious defects that the external parameter calibration is an off-line static process, mechanical drift in long-term operation cannot be compensated, so that a coordinate unified foundation is not aligned, a fusion algorithm (such as Kalman filtering) adopts a fixed noise model, self-adaptive adjustment is difficult to cope with environmental fluctuation such as severe GNSS signal change and the like, and the practicability and reliability of the system are limited. Disclosure of Invention In view of the problems in the prior art, the invention provides a multi-sensor data fusion positioning navigation system for an intelligent campus unmanned vehicle, which is used for solving the technical problems. In order to achieve the above and other objects, the present invention adopts the following technical scheme: The invention provides a multi-sensor data fusion positioning navigation system for an intelligent campus unmanned vehicle, which comprises a data acquisition module, a preprocessing module, a map building and positioning module, a path planning module and a motion control module; The data acquisition module is used for acquiring environment point cloud data output by the vehicle-mounted multi-line laser radar, inertial navigation data output by the inertial measurement unit and satellite positioning data output by the global satellite navigation system in real time; the preprocessing module is configured to receive environmental point cloud data, inertial navigation data and satellite positioning data, perform time point synchronous alignment on the data according to a preset time reference, unify the sensor data under a vehicle body coordinate system based on a pre-calibrated external parameter matrix, and output multi-source sensor data streams after space-time registration; The map building and positioning module is used for receiving the multi-source sensor data flow, generating a three-dimensional point cloud map by utilizing an improved simultaneous positioning and map building algorithm in a map building stage, converting the three-dimensional point cloud map into two-dimensional grid map data, adaptively selecting a positioning mode according to the current environmental characteristics in a navigation stage, and calculating optimal pose estimation data of the unmanned vehicle in the two-dimensional grid map data based on a multi-sensor data fusion algorithm; The path planning module generates mixed navigation data comprising a global path sequence and a local obstacle avoidance track according to the optimal pose estimation data, the two-dimensional grid map data and the target point data; the motion control module receives the hybrid navigation data, and converts the hybrid navigation data into a chassis motor driving instruction through the calculation of a kinematic model to control the unmanned vehicle to run. As described above, the multi-sensor data fusion positioning navigation system for the intelligent campus unmanned vehicle provided by the invention has at least the following beneficial effects: The invention adopts an improved simultaneous positioning and mapping algorithm in the mapping and positioning link, can efficiently construct and optimize a three-dimensional environment model in the mapping stage, and intelligently converts the three-dimensional environment model into a light two-dimensional grid map suitable for real-time navigation. Particularly, in the navigation process, the scheme innovatively designs a mechanism for adaptively selecting an optimal positioning mode (such as feature matching positioning, inertial dominant positioning or fusion positioning) according to the multiple