A mapping method for unmanned vehicles combining lidar and acceleration sensors

Abstract

This invention discloses a method for mapping unmanned vehicles using a combination of laser radar and accelerometers. The method involves installing a laser radar and accelerometer on the unmanned vehicle to collect information about the surrounding environment and the vehicle's own position. The vehicle's position information is converted to a world coordinate system, and an odometry mathematical model is constructed using the discrete information of the vehicle's position. The odometry model is subjected to particle filtering, and the predicted and measured position are calculated to determine the vehicle's position with minimal noise. A raster map is then generated based on this result. Frames of point cloud data are inserted into the generated raster map, and a quadtree search algorithm is used to find the locally optimal matching frame. The resulting locally optimal matching frame is then subjected to a point cloud-to-map match to obtain a more accurate global matching frame. These globally matching frames are then used to construct a more precise raster map. This method can enhance the mapping effect of corridor environments and similar environments, thereby improving mapping accuracy.

Description

Technical Field

[0001] The present invention relates to the field of unmanned vehicles, and in particular to a mapping method combining a laser radar and an acceleration sensor. Background Art

[0002] Over the past few decades, the development and application of mobile robots have progressed. Simultaneous localization and mapping involves sensors equipped with specific sensors modeling their environment while in motion, without prior knowledge of the environment, while simultaneously estimating their own motion. This is the foundation for robots to map in unknown environments. The classic SLAM framework consists of five main components: sensor information reading, front-end visual odometry, back-end (nonlinear) optimization, loop detection, and mapping. Laser SLAM is a method of mapping that uses lidar to perceive the current environment. LiDAR emits laser light from a transmitter toward the object being measured. Upon encountering the object, the laser light is diffusely reflected and received by a laser receiver. The radar module then processes the information based on the time interval to obtain position and velocity information.

[0003] At present, with the development of technology, many excellent algorithms have been applied in laser SLAM and have achieved good results in experiments.

[0004] However, relying solely on laser sensors cannot meet the needs of high-precision mapping. The reason is that laser sensors will drift when mapping corridor scenes, and it is impossible to determine whether the unmanned vehicle has passed through a location it has passed before. With current technology, fused sensors can be used to improve mapping accuracy. However, if only the fusion of lidar and sensors is considered, when the motion state changes significantly, the map constructed by the unmanned vehicle will have serious errors, and at this time it is necessary to introduce a filtering algorithm.

[0005] The essence of filtering algorithms is weighting, which reduces the impact of noise by reducing its weight. Currently, more mature filtering algorithms include particle filtering, Kalman filtering, and nonparametric filtering. Particle filtering is simpler than other filtering methods, requires fewer parameters, and exhibits remarkable robustness. However, debugging hyperparameters is cumbersome and requires careful selection. Even in the same scenario, hyperparameters can vary due to uncontrollable environmental variables, limiting their widespread application. Summary of the Invention

[0006] The purpose of the present invention is to provide a method for unmanned vehicle mapping that combines a laser radar with an acceleration sensor, thereby achieving grid map construction and overcoming the problems existing in the prior art.

[0007] In order to achieve the above tasks, the present invention adopts the following technical solutions:

[0008] A method for constructing a map based on the fusion of an acceleration sensor and a lidar, comprising the following steps:

[0009] Step 1: First, install an accelerometer in front of the unmanned vehicle to measure its position and posture; install a lidar on top of the unmanned vehicle to obtain radar point cloud data;

[0010] Place the unmanned vehicle on the road to be mapped and start the vehicle. As the vehicle moves, the lidar and accelerometer sensors acquire point cloud data and speed data in real time.

[0011] Step 2: Based on the collected speed data, calculate the position information of the unmanned vehicle in the lidar coordinate system, and then use the coordinate transformation matrix to convert the position information in the lidar coordinate system to the world coordinate system; in the world coordinate system, establish the odometry model:

[0012] Step 2.1: The velocity data of the unmanned vehicle at the current moment i collected by the acceleration sensor is used to calculate the position of the unmanned vehicle in the lidar coordinate system according to the following formula:

[0013] S x =S x(i-1) +V x *Δt+a x *Δt 2

[0014] In the above formula, V x represents the speed of the unmanned vehicle projected onto the x-direction of the laser radar coordinate system, a x represents the acceleration of the unmanned vehicle projected in the x direction, and Δt represents the measured V x and a x The instantaneous time interval, S x(i-1) represents the coordinate projected onto the x-axis at time i-1, S x represents the coordinate projected onto the x-axis at time i; the coordinate projected onto the y-axis at time i is S y , then Thus, the position of the unmanned vehicle at time i in the laser radar coordinate system is obtained (S x ,S y ,S θ );

[0015] The obtained S x ,S y ,S θ Convert from the lidar coordinate system to the world coordinate system through coordinate transformation:

[0016] The position S of the unmanned vehicle in the radar coordinate system is (S x ,S y ,S θ) indicates that its position in the world coordinate system is U=(U x ,U y ,U θ ) T :

[0017]

[0018] Matrix T U is the orthogonal rotation matrix in Indicates the attitude angle of the unmanned vehicle, U x Indicates the horizontal coordinate of the unmanned vehicle in the world coordinate system, U y Indicates the vertical coordinate of the unmanned vehicle in the world coordinate system, U θ Indicates the angle between the unmanned vehicle and the horizontal coordinate in the world coordinate system;

[0019] In step 2.2, the distance Δs and the angle Δθ of the vehicle’s movement during the sampling period can be expressed as:

[0020]

[0021] In the formula, Δs r is the distance traveled by the right powered wheel of the unmanned vehicle at the sampling interval Δt, Δs l is the distance traveled by the left powered wheel at the sampling interval Δt; therefore, according to recursion, the mathematical expression of the estimated position of the unmanned vehicle from time k to time k+1 can be obtained, that is, the odometer model is as follows:

[0022]

[0023] where Z x(k+1) ,Z y(k+1) ,Z θ(k+1) U represents the estimated horizontal coordinate, vertical coordinate and the angle between the horizontal coordinate and the vertical coordinate of the unmanned vehicle in the world coordinate system at time k+1; x(k) ,U y(k) ,U θ(k) represents the position of the unmanned vehicle in the world coordinate system at time k;

[0024] Step 3: Perform particle filtering on the odometer model. The specific steps are as follows:

[0025] Step 3.1: The position of the unmanned vehicle in the world coordinate system at time k can be obtained by formula 1: Estimate the estimated position Z of the unmanned vehicle at time k k , according to formula 2 and formula 3, Z k It can be expressed as

[0026] Step 3.2: According to the following formula, when using particle filtering to predict the change of environmental information, under the condition that the unmanned vehicle estimates the posture at time k, the probability p(U k |Z k ):

[0027]

[0028] in It represents the i-th particle extracted by the unmanned vehicle at time k. A total of N particles are extracted. The Dirac function is represented by σ(.). Represents the important weight of the normalized particle,

[0029] Step 3.3, according to the probability p(U k |Z k ) and U obtained in step 3.1 k and Z k The posture P of the unmanned parking space at time k after filtering is calculated by the following formula: k :

[0030] P k =A*U k +B*(Z k -p(U k |Z k )*U k ) (5)

[0031] In the above formula, A represents the hyperparameter of the measured value, and B represents the hyperparameter of the difference between the estimated value and the measured value processed by the measured probability; compared with U k , P after filtering k Closer to the actual position of the unmanned vehicle; P k It can be expressed as

[0032] Step 3.4: Repeat steps 3.1, 3.2, and 3.3 to generate a set of filtered poses of the autonomous vehicle at each moment. These poses are connected according to the time of movement to construct a grid map.

[0033] Step 4: Build a global map

[0034] Step 4.1: Insert the set number of frame point cloud data into the generated grid map. The point cloud data comes from the point cloud data collected by the radar in step 1. The quadtree search algorithm is used to simplify the grid map based on the depth-first principle. By introducing the occupancy probability value M of the grid point corresponding to the laser scanning point in the direction of the unmanned vehicle posture, the grid map is simplified. nearest , reducing the time complexity of the search algorithm; M nearestThe calculation method is as shown in formula 7. The mathematical expression of the above process is as follows:

[0035]

[0036]

[0037]

[0038] In formula 6, W represents the search space, that is, the generated grid map, M nearest represents the occupancy probability value of the grid point corresponding to the laser scanning point in the direction of the unmanned vehicle posture, ξ represents the grid, T ξ Represents a point cloud coordinate of the grid, where the subscript ξ∈W represents traversing all grids of the grid map; K represents the total time of constructing the grid map, h k represents T at time k ξ The weight value of the point cloud coordinates in the grid map, parameters a, b∈(0,1), ξ * Represents the local optimal matching frame of the grid map search;

[0039] In step 4.2, after finding the local optimal matching frame using the quadtree search algorithm, insert the point cloud data acquired by the lidar in step 1 into the obtained local optimal matching frame, register the point cloud data with the local optimal matching frame in the order of acquisition time, and thus obtain the global optimal keyframe;

[0040] Step 4.3, repeat steps 4.1 and 4.2. When the inserted point cloud data completely covers the grid, a more accurate grid map based on the global matching frame is constructed.

[0041] Furthermore, the acceleration sensor model is BWT61CL, the installation direction of the acceleration sensor is consistent with the direction of travel of the unmanned vehicle, and the operating voltage of the sensor is set to 5V DC.

[0042] Furthermore, in the particle filter, each particle represents a possible hypothesis of the sensor measuring the current moment, that is, a possible hypothesis of the posture measured by the unmanned vehicle at time k.

[0043] Furthermore, the value range of A is (1 / N, 1), and the value range of B is [0, 1].

[0044] A mapping module includes a processor, a memory, and a computer program stored in the memory; when the processor executes the computer program, the steps of the unmanned vehicle mapping method combining a laser radar and an acceleration sensor are implemented.

[0045] A computer-readable storage medium stores a computer program; when the computer program is executed by a processor, the steps of the unmanned vehicle mapping method combining a laser radar and an acceleration sensor are implemented.

[0046] Compared with the prior art, the improved invention has the following advantages:

[0047] This invention is suitable for map construction in complex scenarios, overcoming the shortcomings of traditional algorithms in map construction distortion in corridor environments. Compared with traditional algorithms, it simplifies the amount of calculation, improves the mapping accuracy, adds an acceleration sensor, collects the speed and angular velocity of the unmanned vehicle, and integrates the data transmitted by the speed sensor with the visual odometry, thereby improving the system's mapping stability and accuracy under rapid changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a flowchart of the operation process of the method of the present invention;

[0049] Figure 2 This is the Rviz simulation effect diagram of the method of the present invention;

[0050] Figure 3 Gazebo configuration environment map for the method of the present invention;

[0051] Figure 4 This is a real picture of the unmanned vehicle using the method of the present invention. DETAILED DESCRIPTION

[0052] A method for constructing a map based on the fusion of an acceleration sensor and a laser radar of the present invention comprises the following steps:

[0053] Step 1: First, install an accelerometer directly in front of the unmanned vehicle to measure its position and posture. A lidar (lidar) is installed on top of the vehicle to acquire radar point cloud data. In this embodiment, the accelerometer is a BWT61CL. The accelerometer's installation direction must align with the vehicle's direction of travel. The sensor's operating voltage is set to 5V DC.

[0054] Place the unmanned vehicle on the road and start it. During its movement, the point cloud data acquired by the lidar and the speed data (including linear velocity and angular velocity data) acquired by the acceleration sensor are transmitted to the host computer in real time.

[0055] Step 2: Based on the collected speed data, calculate the position information of the unmanned vehicle in the lidar coordinate system, and then use the coordinate transformation matrix to convert the position information in the lidar coordinate system to the world coordinate system; in the world coordinate system, establish an odometer model.

[0056] In step 2.1, the point cloud data acquired by the LiDAR is in the LiDAR coordinate system. The speed data of the unmanned vehicle at the current moment i collected by the accelerometer is used to calculate the position of the unmanned vehicle in the LiDAR coordinate system according to the following formula:

[0057] S x =S x(i-1) +V x *Δt+a x *Δt 2

[0058] In the above formula, V x represents the speed of the unmanned vehicle projected onto the x-direction of the laser radar coordinate system, a x represents the acceleration of the unmanned vehicle projected in the x direction, and Δt represents the measured V x and a x The instantaneous time interval, S x(i-1) represents the coordinate projected onto the x-axis at time i-1, S x Indicates the coordinate projected onto the x-axis at time i. The initial coordinate S x(0) Set to 0, indicating that the unmanned vehicle starts from the origin of the laser radar coordinate system. Similarly, the coordinate S projected onto the y-axis at time i can be obtained. y , according to the tangent theorem, we can get Thus, the position of the unmanned vehicle at time i in the laser radar coordinate system is obtained (S x ,S y ,S θ ).

[0059] The obtained S x ,S y ,S θ Convert from the lidar coordinate system to the world coordinate system through coordinate transformation:

[0060] The position S of the unmanned vehicle in the radar coordinate system is (S x ,S y ,S θ ) indicates that its position in the world coordinate system is U=(U x ,U y ,U θ ) T :

[0061]

[0062] The coordinate system transformation of the unmanned vehicle is generally solved by calculating the orthogonal rotation matrix. The matrix T U is the orthogonal rotation matrix The function of this formula is to transform the motion of the mobile unmanned vehicle into the world coordinate system, where Indicates the attitude angle of the unmanned vehicle, U x Indicates the horizontal coordinate of the unmanned vehicle in the world coordinate system, U y Indicates the vertical coordinate of the unmanned vehicle in the world coordinate system, U θ Represents the angle between the autonomous vehicle and the horizontal coordinate in the world coordinate system.

[0063] Step 2.2, as can be seen from the analysis and deduction of step 2.1, the position of the unmanned vehicle can be expressed as U = (U x ,U y ,U θ ) T The odometer mathematical model can be expressed as Δt, which replaces speed information with measured distance data. The position information of the mobile unmanned vehicle is calculated using the equation. Because the input control system of the unmanned vehicle is discretized and has a set sampling interval Δt, the discrete mathematical model of the mobile unmanned vehicle's position information can be estimated by measuring the distance the two powered wheels have rotated at a fixed time interval using the odometer. The change in the mobile robot's position at the previous moment is calculated. Then, the distance Δs and the angle Δθ of the unmanned vehicle's movement during the sampling period can be expressed as:

[0064]

[0065] In the formula, Δs r is the distance traveled by the right powered wheel of the unmanned vehicle at the sampling interval Δt, Δs l is the distance traveled by the left powered wheel at the sampling interval Δt. Therefore, based on recursion, we can obtain the mathematical expression of the estimated position of the unmanned vehicle from time k to time k+1, that is, the odometer model is as follows:

[0066]

[0067] Among them U x(k) Z represents the horizontal coordinate of the unmanned vehicle in the world coordinate system at time k (here is the coordinate of the unmanned vehicle projected by the radar measured by the sensor, and the position obtained by coordinate transformation), x(k+1) ,Z y(k+1) ,Z θ(k+1) U represents the estimated horizontal coordinate, vertical coordinate and the angle between the horizontal coordinate and the vertical coordinate of the unmanned vehicle in the world coordinate system at time k+1; x(k) ,U y(k) ,U θ(k) Represents the position of the unmanned vehicle in the world coordinate system at time k.

[0068] Step 3: Perform particle filtering on the odometry model. This step aims to improve the accuracy of the odometry model and measure the difference between the probability distribution of the estimated odometry and the probability distribution of the true value. The specific steps in the particle filtering process are as follows:

[0069] Step 3.1: The position of the unmanned vehicle in the world coordinate system at time k can be obtained by formula 1: Estimate the estimated position Z of the unmanned vehicle at time k k , according to formula 2 and formula 3, Z k It can be expressed as

[0070] Step 3.2, the purpose of particle filtering is to minimize the noise impact caused by the accuracy error of the measurement equipment. Due to the error of the sensor, the final unmanned vehicle posture also has errors. In addition to the error in the measured unmanned vehicle posture, the estimated posture Z of the unmanned vehicle can be calculated by formula 3. k There is a difference between the two. The essence of filtering is weighting. According to the size of the set hyperparameter value, the unmanned parking space posture P that is closer to the true value is finally obtained. k .

[0071] According to the following formula, when using particle filtering to predict changes in environmental information, the position U measured by the unmanned vehicle at the current k moment is described k (measured value) and estimated pose Z k The mathematical relationship between the estimated values ​​is that under the condition that the unmanned vehicle estimates the pose at time k, the probability p(U k |Z k ):

[0072]

[0073] in represents the i-th particle extracted by the unmanned vehicle at time k, and a total of N particles are extracted (each particle here represents a possible hypothesis of the sensor measuring the current moment, that is, a possible hypothesis of the unmanned vehicle measuring the posture at time k). The Dirac function is represented by σ(.), Represents the important weight of the normalized particle, usually taken as (1 / N is also selected in this step).

[0074] Step 3.3, according to the probability p(U k |Z k ) and U obtained in step 3.1 k and Z k The posture P of the unmanned parking space at time k after filtering is calculated by the following formula: k :

[0075] P k =A*U k +B*(Z k -p(U k |Zk )*U k ) (5)

[0076] In the above formula, A represents the hyperparameter of the measured value, and B represents the hyperparameter of the difference between the estimated value and the measured value processed by the measured probability; in this scheme, the value range of A is (1 / N, 1), and the value range of B is [0, 1]; compared with U k , P after filtering k Closer to the actual position of the unmanned vehicle; P k It can be expressed as

[0077] Step 3.4: Repeat steps 3.1, 3.2, and 3.3 to generate a set of filtered unmanned vehicle poses at each moment. These pose sets are connected according to the time of movement to construct a grid map.

[0078] Step 4: Build a global map. This step is done after the odometry model is processed with particle filtering in step 3. Then, the global map is updated by matching the inserted frame point cloud data and performing back-end optimization.

[0079] Step 4.1, insert the set number of frame point cloud data into the generated grid map. The point cloud data comes from the point cloud data collected by the radar in step 1. The quadtree search algorithm is used with the depth-first principle. In order to narrow the search range of the grid map, the grid map needs to be simplified. By introducing the occupancy probability value M of the grid point corresponding to the laser scanning point in the direction of the unmanned vehicle posture, the grid map is simplified. nearest , reducing the time complexity of the search algorithm. nearest The calculation method is as shown in Formula 7. The mathematical expression of the above process is as follows:

[0080]

[0081]

[0082]

[0083] In formula 6, W represents the search space, that is, the generated grid map, M nearest represents the occupancy probability value of the grid point corresponding to the laser scanning point in the direction of the unmanned vehicle posture, ξ represents the grid, T ξ Represents a point cloud coordinate of the grid, where the subscript ξ∈W represents traversing all grids of the grid map; K represents the total time of constructing the grid map, h k represents T at time k ξ The weight value of the point cloud coordinates in the grid map is calculated as shown in Formula 8. The parameters a, b∈(0,1) are random values ​​between 0 and 1. *The purpose of Equation 6 is to find the local optimal matching frame in the search space W so that the global optimal key frame can be obtained by inserting point cloud data in the next step.

[0084] In step 4.2, after using the quadtree search algorithm to find the local optimal matching frame, insert the point cloud data acquired by the lidar in step 1 into the local optimal matching frame. Register the point cloud data and the local optimal matching frame in the order of acquisition time to obtain the global optimal keyframe. The main purpose of obtaining the global optimal keyframe is to generate a more accurate raster map when the inserted point cloud data is sufficient to cover the entire raster map.

[0085] Step 4.3, repeat steps 4.1 and 4.2. When the inserted point cloud data completely covers the grid, a more accurate grid map based on the global matching frame is constructed.

[0086] Simulation experiment:

[0087] The algorithm was simulated using the Gazebo simulation environment. The computer operating system was Ubuntu, the mobile robot operating system version was ROS Kinetic, and the Rviz visualization environment was used.

[0088] First, configure the mobile unmanned vehicle simulation model, including speed, acceleration, steering angle and other parameters; in ROS, subscribe to pre-published topics for the lidar, accelerometer, and encoder. Create a simple experimental environment in Gazebo, set up models such as walls and obstacles, write the launch file, and start the ROS node. The simulation effect is as follows: Figure 3 shown.

[0089] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.

Claims

1. A method for constructing a map based on the fusion of an acceleration sensor and a lidar, characterized in that: The following steps are involved: Step 1: First, install an accelerometer in front of the unmanned vehicle to measure its position and posture; install a lidar on top of the unmanned vehicle to obtain radar point cloud data; Place the unmanned vehicle on the road to be mapped and start the vehicle. As the vehicle moves, the lidar and accelerometer sensors acquire point cloud data and speed data in real time. Step 2: Based on the collected speed data, calculate the position information of the unmanned vehicle in the lidar coordinate system, and then use the coordinate transformation matrix to convert the position information in the lidar coordinate system to the world coordinate system; in the world coordinate system, establish the odometry model: Step 2.1: The velocity data of the unmanned vehicle at the current moment i collected by the acceleration sensor is used to calculate the position of the unmanned vehicle in the lidar coordinate system according to the following formula: S x =S x(i-1) +V x *Δt+a x *Δt 2 In the above formula, V x represents the speed of the unmanned vehicle projected onto the x-direction of the laser radar coordinate system, a x represents the acceleration of the unmanned vehicle projected in the x direction, and Δt represents the measured V x and a x The instantaneous time interval, S x(i-1) represents the coordinate projected onto the x-axis at time i-1, S x represents the coordinate projected onto the x-axis at time i; the coordinate projected onto the y-axis at time i is S y , then Thus, the position of the unmanned vehicle at time i in the laser radar coordinate system is obtained (S x ,S y ,S θ ); The obtained S x ,S y ,S θ Convert from the lidar coordinate system to the world coordinate system through coordinate transformation: The position S of the unmanned vehicle in the radar coordinate system is (S x ,S y ,S θ ) indicates that its position in the world coordinate system is U=(U x ,U y ,U θ ) T : Matrix T U is the orthogonal rotation matrix in Indicates the attitude angle of the unmanned vehicle, U x Indicates the horizontal coordinate of the unmanned vehicle in the world coordinate system, U y Indicates the vertical coordinate of the unmanned vehicle in the world coordinate system, U θ Indicates the angle between the unmanned vehicle and the horizontal coordinate in the world coordinate system; In step 2.2, the distance Δs and the angle Δθ of the vehicle’s movement during the sampling period can be expressed as: In the formula, Δs r is the distance traveled by the right powered wheel of the unmanned vehicle at the sampling interval Δt, Δs l is the distance traveled by the left powered wheel at the sampling interval Δt; therefore, according to recursion, the mathematical expression of the estimated position of the unmanned vehicle from time k to time k+1 can be obtained, that is, the odometer model is as follows: where Z x(k+1) ,Z y(k+1) ,Z θ(k+1) U represents the estimated horizontal coordinate, vertical coordinate and the angle between the horizontal coordinate and the vertical coordinate of the unmanned vehicle in the world coordinate system at time k+1; x(k) ,U y(k) ,U θ(k) represents the position of the unmanned vehicle in the world coordinate system at time k; Step 3: Perform particle filtering on the odometer model. The specific steps are as follows: Step 3.1: The position of the unmanned vehicle in the world coordinate system at time k can be obtained by formula 1: Estimate the estimated position Z of the unmanned vehicle at time k k , according to formula 2 and formula 3, Z k It can be expressed as Step 3.2: According to the following formula, when using particle filtering to predict the change of environmental information, under the condition that the unmanned vehicle estimates the posture at time k, the probability p(U k |Z k ): in It represents the i-th particle extracted by the unmanned vehicle at time k. A total of N particles are extracted. The Dirac function is represented by σ(.). Represents the important weight of the normalized particle, Step 3.3, according to the probability p(U k |Z k ) and U obtained in step 3.1 k and Z k The posture P of the unmanned parking space at time k after filtering is calculated by the following formula: k : P k =A*U k +B*(Z k -p(U k |Z k )*U k ) (5) In the above formula, A represents the hyperparameter of the measured value, and B represents the hyperparameter of the difference between the estimated value and the measured value processed by the measured probability; compared with U k , P after filtering k Closer to the actual position of the unmanned vehicle; P k It can be expressed as Step 3.4: Repeat steps 3.1, 3.2, and 3.3 to generate a set of filtered poses of the autonomous vehicle at each moment. These poses are connected according to the time of movement to construct a grid map. Step 4: Build a global map Step 4.1: Insert the set number of frame point cloud data into the generated grid map. The point cloud data comes from the point cloud data collected by the radar in step 1. The quadtree search algorithm is used to simplify the grid map based on the depth-first principle. By introducing the occupancy probability value M of the grid point corresponding to the laser scanning point in the direction of the unmanned vehicle posture, the grid map is simplified. nearest , reducing the time complexity of the search algorithm; M nearest The calculation method is as shown in formula 7. The mathematical expression of the above process is as follows: In formula 6, W represents the search space, that is, the generated grid map, M nearest represents the occupancy probability value of the grid point corresponding to the laser scanning point in the direction of the unmanned vehicle posture, ξ represents the grid, T ξ Represents a point cloud coordinate of the grid, where the subscript ξ∈W represents traversing all grids of the grid map; K represents the total time of constructing the grid map, h k represents T at time k ξ The weight value of the point cloud coordinates in the grid map, parameters a, b∈(0,1), ξ * Represents the local optimal matching frame of the grid map search; In step 4.2, after finding the local optimal matching frame using the quadtree search algorithm, insert the point cloud data acquired by the lidar in step 1 into the obtained local optimal matching frame, register the point cloud data with the local optimal matching frame in the order of acquisition time, and thus obtain the global optimal keyframe; Step 4.3, repeat steps 4.1 and 4.

2. When the inserted point cloud data completely covers the grid, a more accurate grid map based on the global matching frame is constructed.

2. The method for constructing a map based on the fusion of acceleration sensor and laser radar according to claim 1, characterized in that: The acceleration sensor model is BWT61CL. The installation direction of the acceleration sensor is consistent with the direction of travel of the unmanned vehicle. The operating voltage of the sensor is set to 5V DC.

3. The method for constructing a map based on the fusion of acceleration sensor and laser radar according to claim 1, characterized in that: In particle filtering, each particle represents a possible hypothesis of the sensor measuring the current moment, that is, a possible hypothesis of the posture measured by the unmanned vehicle at time k.

4. The method for constructing a map based on the fusion of an acceleration sensor and a laser radar according to claim 1, characterized in that: The value range of A is (1 / N, 1), and the value range of B is [0, 1].

5. A mapping module comprising a processor, a memory, and a computer program stored in the memory; characterized in that: When the processor executes the computer program, it implements the steps of the method for constructing a map based on the fusion of an acceleration sensor and a lidar as described in any one of claims 1 to 4.

6. A computer-readable storage medium storing a computer program; wherein: When the computer program is executed by a processor, the steps of the method for constructing a map based on the fusion of an acceleration sensor and a lidar as described in any one of claims 1 to 4 are implemented.

Images