CN116659500A

CN116659500A - Mobile robot positioning method and system based on laser radar scanning information

Info

Publication number: CN116659500A
Application number: CN202310604731.6A
Authority: CN
Inventors: 宋凯; 翟凯; 华逢彬; 马辰; 程瑶
Original assignee: Shandong New Generation Information Industry Technology Research Institute Co Ltd
Current assignee: Shandong New Generation Information Industry Technology Research Institute Co Ltd
Priority date: 2023-05-26
Filing date: 2023-05-26
Publication date: 2023-08-29

Abstract

The invention discloses a mobile robot positioning method and system based on laser radar scanning information, belongs to the technical field of robot positioning, and aims to solve the technical problem of how to improve the defects of global positioning and follow positioning of a robot particle filter positioning algorithm in a large-scale map. The method is applied to a mobile robot configured with a perception sensor, wherein the perception sensor comprises an odometer, an IMU and a laser radar and is used for collecting pose information and environment information, the method comprises a pre-processing module, a global initialization positioning module and a follow-up positioning module, a scanning matching process based on a multi-resolution grid map and a high-confidence region is adopted, the high-confidence region and particle genetic mutation are introduced into the robot follow-up positioning module, a likelihood domain observation model is applied to update particles in the resampling process, and meanwhile, a high-confidence region reference is added, so that a particle set is converged to the vicinity of an execution track, the particle convergence effect is good, and the following accuracy of the robot is ensured.

Description

Mobile robot positioning method and system based on laser radar scanning information

Technical Field

The invention relates to the technical field of robot positioning, in particular to a mobile robot positioning method and system based on laser radar scanning information.

Background

In recent years, mobile robots have developed in ripening, and are widely used in different fields such as industrial inspection, logistics distribution, cleaning service, etc., and are often used to replace manual execution of simple, repeated and complex work. In consideration of the requirements of reliability, safety and the like of mobile robots, enterprises and scientific research institutions often conduct the research of robot positioning navigation by taking a laser radar with excellent characteristics such as high measurement precision, high stability and the like as a sensing terminal.

In the fields of machine rooms and industrial inspection, the mobile robots finish point-to-point operation mostly according to a path tracking control algorithm, and the robots finish noise, temperature, humidity, pointer values and other abnormal information detection by setting fixed points, so that the reliability of the tail end detection module is inseparable from the path following precision and the arrival precision of the robots; in the field of logistics distribution, a mobile robot becomes an important component in the whole assembly line, and often cooperates with a product line and a mechanical arm to finish tasks such as accurate assembly and transportation, and the quality of the positioning performance of the robot directly influences the whole cooperation process, so that the assembly line efficiency is influenced; in the field of cleaning services, mobile robots are often used in dynamic and noisy environments such as markets, living rooms, hotels and the like, the effective characteristics of the mobile robots based on laser radars are suddenly reduced, the robots execute motion planning by error positioning, and the positioning recovery efficiency after the characteristics are recovered determines the tasks such as reliable cleaning, meal delivery and the like of the robots. Therefore, the improvement of the positioning performance of the mobile robot is an effective means for promoting the intelligent and efficient catalysts of the industry and the efficient landing of products in various application scenes.

At present, the indoor mobile robot positioning based on laser radar scanning information mostly adopts a particle filtering algorithm with excellent performance of a nonlinear and non-Gaussian system, and the algorithm is relatively mature and stable, but still has certain limitations. In the global positioning process of the mobile robot, the global positioning can be relatively quickly completed in a small-scale map so as to follow the positioning later, but in a larger-scale map, fewer particles cannot iterate out the optimal pose efficiently, and the real-time requirement of the robot is influenced by higher time complexity of more particles; in the following positioning process of the mobile robot, when the prior environment changes or the robot kidnapping occurs, particles degenerate, and the robot positioning cannot be effectively recovered after the algorithm resampling, so that the robot navigation task fails.

How to improve the defects of global positioning and follow-up positioning of a robot particle filtering positioning algorithm in a large-scale map is a technical problem to be solved.

Disclosure of Invention

The technical task of the invention is to provide a mobile robot positioning method and a mobile robot positioning system based on laser radar scanning information to solve the technical problem of how to improve the defects of global positioning and follow positioning of a robot particle filtering positioning algorithm in a large-scale map.

In a first aspect, the invention provides a mobile robot positioning system based on laser radar scanning information, which is applied to a mobile robot provided with a perception sensor, wherein the perception sensor comprises an odometer, an IMU and a laser radar and is used for acquiring pose information and environment information, the positioning system comprises a pre-processing module, a global initialization positioning module and a following positioning module,

the pre-processing module is used for constructing a priori grid map based on pose information and environment information, generating a multi-resolution grid map based on the priori grid map, and dividing high-low confidence areas in the multi-resolution grid map according to task information set by a user to obtain high-confidence areas and low-confidence areas, wherein the task information is used for providing motion control and navigation planning for the mobile robot;

the global initialization positioning module is used for controlling the mobile robot to execute a rotating action of rotating 360 degrees in situ, estimating the posture of the mobile robot based on point cloud information acquired by the laser radar and pose information acquired by the IMU, constructing a full-view point cloud map based on the laser point cloud information under a laser radar coordinate system, and performing branch-and-bound scanning matching based on the multi-resolution grid map and a high-confidence region to obtain global positioning;

The following positioning module is used for carrying out motion updating based on gesture information acquired by the IMU and position information acquired by the odometer to obtain a predicted value of the pose, carrying out laser radar observation updating based on the predicted value of the pose, the high-confidence-degree region and the particle genetic mutation to obtain an observed value of the pose, and carrying out resampling based on the observed value of the pose to obtain the pose.

Preferably, the pre-processing module interacts with the user through a pre-processing interface, and is configured to perform the following steps:

constructing an a priori grid map: regulating and controlling a mobile robot scanning operation environment, and constructing a priori grid map with a specified resolution based on graph optimization according to environmental information acquired by a sensing sensor;

constructing a multi-resolution grid map: removing invalid points in the prior grid map through a drawing tool to obtain a preprocessed prior grid map, and performing sliding window calculation on the prior grid map based on the multiple of the resolution of the prior grid map as the length of a search frame to obtain a multi-resolution grid map;

dividing the high-low confidence regions: under the condition of configuring high-low confidence region division, task information set by a user is written into a multi-resolution grid map in a grid form, and a likelihood domain space based on a track corresponding to the task information is generated, wherein a likelihood domain space calculation formula is as follows:

Wherein, (x) _free ,y _free ) Is a free space point in a multi-resolution grid map, (x) _line ,y _line ) For points in the track corresponding to the task information, dist_arg is used as a likelihood domain setting width parameter, score is the distance from a blank grid to a set route in the multi-resolution grid map, a likelihood domain space in the dist_arg range is used as a high confidence interval, and scores exceeding the set width of the likelihood domain are all-1, and the score is used as a low confidence area.

Preferably, the global initialization positioning module is configured to perform the following:

constructing a full scenic spot cloud map: controlling a mobile robot to execute a rotation action of rotating in situ for 360 degrees, recording the posture information of the IMU at a preset frequency, estimating the posture according to the time stamps of each point of the laser point cloud and the posture information of the IMU under each time stamp by linear interpolation, sequentially reading the laser point cloud information under a laser radar coordinate system, and splicing the laser point cloud information to obtain a full-scenic point cloud map;

global initialization pose search: performing global initial pose search in a mode of scanning and matching based on the multi-resolution grid map and the high-confidence region, or performing global initial pose search in a mode of searching based on the full-confidence region;

The global initial pose search is performed in a scanning matching mode based on the multi-resolution grid map and the high-confidence region, and the method comprises the following operations:

l1, selecting a grid map with the maximum resolution from the multi-resolution grid map as a target grid map;

l2, performing branch operation on the target grid map, dividing the search space into four small areas by halving the x and y coordinate directions of the target grid map, if the low confidence coefficient area ratio of one small area exceeds 2/3, skipping the scanning matching calculation of the area, and performing the scanning matching score calculation according to the following formula:

wherein M is _smoot h represents a bicubic interpolation function for smoothing gray information of a target grid map, T represents an attitude adjustment variable, z _i Measuring information for an ith laser spot;

l3, selecting a search space with the largest scanning matching score from the four small areas, performing a next round of branching operation on the selected search space, and repeatedly executing the step L2 with a lower-layer resolution grid map until the optimal pose is found;

if the optimal pose is not found in the steps L1-L3, the global initialization module is used for carrying out global initialization pose search in a full confidence region search mode, and the method comprises the following operations:

M1, selecting a grid map with the maximum resolution from the multi-resolution grid map as a target grid map;

m2, performing branch operation on the target grid map, dividing the search space into four small areas by halving the x and y coordinate directions of the target grid map, and performing scanning matching score calculation on each small area by the following formula:

m3, selecting a search space with the largest scanning matching score from the four small areas, performing a next round of branching operation on the selected search space, and repeatedly executing the step L2 with a lower-layer resolution grid map until the optimal pose is found;

if the pre-processing module is to execute the high-low confidence region division, the global initialization module is used for searching the global initial pose based on the mode of scanning and matching the multi-resolution grid map, and the method comprises the following steps:

s1, selecting a grid map with the maximum resolution from the multi-resolution grid map as a target grid map;

s2, performing branch operation on the target grid map, dividing the search space into four small areas by halving the x and y coordinate directions of the target grid map, and performing scanning matching score calculation on each small area by the following formula:

s3, selecting a search space with the largest scanning matching score from the four small areas, performing a next round of branching operation on the selected search space, and repeatedly executing the step L2 with a lower-layer resolution grid map until the optimal pose is found.

Preferably, the following positioning module is configured to perform the following:

motion update: based on the attitude information acquired by the IMU and the position information acquired by the odometer, carrying out motion update on the pose of the mobile robot based on an EKF algorithm to obtain a predicted value of each pose;

and (5) observing and updating: the method comprises the steps of taking a predicted value of a pose, laser point cloud information acquired through a laser radar and a minimum resolution grid map in a multi-resolution grid map as inputs, calculating and outputting an updated value of the pose in a mode of calculating particle weights through a likelihood domain observation algorithm, observing and updating the updated value of the pose based on a high confidence region and a particle genetic mutation algorithm to obtain an observed value of each pose, and constructing particle sets based on the observed value of each pose;

Resampling: and selecting a pose with the largest observation value from the particle set, outputting the pose if the observation value of the pose is larger than a threshold THR, if the observation value of the pose is smaller than or equal to the threshold THR, resampling based on a high confidence region, selecting the pose in the high confidence region, adding the pose into the particle set to obtain a new particle set, selecting the pose with the largest observation value from the new particle set, outputting the pose if the observation value of the pose is larger than the threshold THR, performing global random sampling based on the high confidence region and the low confidence region if the observation value of the pose is smaller than or equal to the threshold THR, randomly selecting the pose, adding the pose into the particle set to obtain the new particle set, selecting the pose with the largest observation value from the new particle set, if the observation value of the pose is larger than the threshold THR, selecting the particle with the highest weight in the particle set, and outputting the particle set to ensure the positioning real-time performance.

Preferably, before resampling, the following positioning module is configured to perform the following: according to the weight of the particles, the particles are divided into two groups, one third of the particles with small weight are selected for particle crossing and mutation to add new unknown particles, and the calculation formula is as follows:

In the method, in the process of the invention,respectively small weight particle sets, large weight particle sets, cross particle sets and mutant particle sets.

In a second aspect, the present invention provides a mobile robot positioning method based on laser radar scanning information, applied to a mobile robot configured with a perception sensor, the perception sensor including an odometer, an IMU and a laser radar, for acquiring pose information and environmental information, the method being used for positioning by the mobile robot positioning system based on laser radar scanning information according to any one of the first aspects, the method comprising the steps of:

pretreatment: constructing a priori grid map based on pose information and environment information, generating a multi-resolution grid map based on the priori grid map, and dividing high-low confidence areas in the multi-resolution grid map according to task information set by a user to obtain high-confidence areas and low-confidence areas, wherein the task information is used for providing motion control and navigation planning for a mobile robot;

global initialization positioning: controlling the mobile robot to execute a rotating action of rotating in situ by 360 degrees, estimating the rotating action of the mobile robot based on point cloud information acquired by a laser radar and pose information acquired by an IMU, constructing a full-view point cloud map based on the laser point cloud information under a laser radar coordinate system, and performing branch-and-bound scanning matching based on a multi-resolution grid map and a high-confidence region to obtain global positioning;

And (5) following and positioning: and carrying out motion updating based on the gesture information acquired by the IMU and the position information acquired by the odometer to obtain a predicted value of the pose, carrying out laser radar observation updating based on the predicted value of the pose, the high-confidence region and the particle genetic mutation to obtain an observed value of the pose, and carrying out resampling based on the observed value of the pose to obtain the pose.

Preferably, the phase pre-treatment comprises the following operations:

constructing an a priori grid map: the remote control mobile robot scans the working environment and constructs a priori grid map with specified resolution based on graph optimization according to the environmental information acquired by the sensing sensor;

Preferably, the global initialization positioning includes the following operations:

if the optimal pose is not found in the steps L1-L3, carrying out global initialization pose search in a full confidence region search mode, wherein the method comprises the following operations:

if the high-low confidence region division is not executed during the pre-processing, the global initial pose search is performed based on a mode of scanning and matching the multi-resolution grid map, and the method comprises the following steps:

Preferably, the following positioning includes the following operations:

Resampling: and selecting a pose with the largest observation value from the particle set, outputting the pose if the observation value of the pose is larger than a threshold THR, resampling the pose based on a high confidence region if the observation value of the pose is smaller than or equal to the threshold THR, selecting the pose with the largest observation value from the new particle set, outputting the pose if the observation value of the pose is larger than the threshold THR, performing global random sampling based on the high confidence region and the low confidence region if the observation value of the pose is smaller than or equal to the threshold THR, randomly selecting the pose, adding the pose into the particle set, obtaining the new particle set, selecting the pose with the largest observation value from the new particle set, outputting the pose if the observation value of the pose is larger than the threshold THR, and selecting the particle with the highest weight in the particle set to output so as to ensure the positioning instantaneity.

Preferably, the following operations are performed before resampling: according to the weight of the particles, the particles are divided into two groups, one third of the particles with small weight are selected for particle crossing and mutation to add new unknown particles, and the calculation formula is as follows:

The mobile robot positioning method and system based on laser radar scanning information have the following advantages:

(1) In the global initialization pose positioning module, constraint is formed by the robot and the following track, in practical application, the robot always performs tasks according to a given track or task point, and the robot is in the areas most of the time, so that the areas are considered to be high-confidence areas, when the robot performs global positioning, the scanning matching process based on the multi-resolution grid map and the high-confidence areas is adopted, the problems that the traditional positioning method of the mobile robot is poor in global initialization positioning effect, low in efficiency, incapable of solving the problem of robot kidnapping and the like in a large-scale and complex environment are solved, the particle convergence speed is enhanced, and the scene adaptability of the robot is improved;

(2) Introducing the ideas of high-confidence region and particle genetic mutation into a robot following and positioning module, and adding high-confidence region references while updating particles by using a likelihood domain observation model in the resampling process, so that a particle set is converged to the vicinity of an execution track, the particle convergence effect is good, and the robot following precision is ensured;

(3) In order to enhance the robustness of the original algorithm, the particle sets with small weights are crossed and mutated before resampling, and the diversity of the particle sets is ensured.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the embodiments or the description of the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings can be obtained according to these drawings without inventive effort for a person skilled in the art.

The invention is further described below with reference to the accompanying drawings.

FIG. 1 is a block diagram of the operation of a mobile robot positioning system based on laser radar scanning information according to embodiment 1;

fig. 2 is a user task setting diagram based on a grid map in the mobile robot positioning system based on laser radar scanning information according to embodiment 1;

fig. 3 is a high-low confidence region division diagram based on a grid map in a mobile robot positioning system based on laser radar scanning information according to embodiment 1.

Detailed Description

The invention will be further described with reference to the accompanying drawings and specific examples, so that those skilled in the art can better understand the invention and implement it, but the examples are not meant to limit the invention, and the technical features of the embodiments of the invention and the examples can be combined with each other without conflict.

The embodiment of the invention provides a mobile robot positioning method and a mobile robot positioning system based on laser radar scanning information, which are used for solving the technical problem of how to improve the defects of global positioning and following positioning of a robot particle filtering positioning algorithm in a large-scale map.

Example 1:

the invention discloses a mobile robot positioning system based on laser radar scanning information, which is applied to a mobile robot provided with a sensing sensor, wherein the sensing sensor comprises an odometer, an IMU and a laser radar and is used for acquiring pose information and environment information. In this embodiment, the positioning system includes a pre-processing module, a global initialization positioning module, and a follow-up positioning module.

The pre-processing module is used for constructing a priori grid map based on pose information and environment information, generating a multi-resolution grid map based on the priori grid map, and dividing high-low confidence areas in the multi-resolution grid map according to task information set by a user to obtain high-confidence areas and low-confidence areas, wherein the task information is used for providing motion control and navigation planning for the mobile robot.

The pre-processing module is interacted with a user through a pre-processing interface and is used for executing the following steps:

(1) Constructing an a priori grid map: the remote control mobile robot scans the working environment and constructs a priori grid map with specified resolution based on graph optimization according to the environmental information acquired by the sensing sensor;

(2) Constructing a multi-resolution grid map: removing invalid points in the prior grid map through a drawing tool to obtain a preprocessed prior grid map, and performing sliding window calculation on the prior grid map based on the multiple of the resolution of the prior grid map as the length of a search frame to obtain a multi-resolution grid map;

(3) Dividing the high-low confidence regions: under the condition of configuring high-low confidence region division, task information set by a user is written into a multi-resolution grid map in a grid form, and a likelihood domain space based on a track corresponding to the task information is generated, wherein a likelihood domain space calculation formula is as follows:

In this embodiment, the pre-processing module is a unit for performing cooperative interaction with a user, and performs data fusion processing on visual information such as a given priori map and a user-set task, so as to enrich positioning reference information of the robot, thereby improving positioning efficiency of the robot. When the mobile robot based on laser radar scanning information works, the remote controller is used for controlling or pushing to complete the acquisition of attitude information and environment information through sensing sensors such as an odometer, an IMU, a laser radar and the like, so that the environment map construction, namely the SLAM process is completed. In order to ensure that the robot stably completes path planning and fixed-point operation according to an assumed route, deployment personnel often add task setting and auxiliary information in the obtained priori map, wherein the task setting and auxiliary information comprises charging point electricity for pile returning, a to-be-positioned working point of the robot, a key working point, a virtual wall, a set path and the like. The early preprocessing module consists of two parts of multi-resolution grid map generation and high-low confidence level region division, and the generation of the multi-resolution grid map is a calculation process with high time complexity, so that the current environment map is required to be calculated before task execution, low time complexity inquiry is provided for subsequent global positioning and follow-up positioning, and algorithm instantaneity is ensured; the high-low confidence region dividing part divides the prior grid map into a high-confidence region and a low-confidence region according to task setting and auxiliary information added by a user, wherein the high-confidence region represents that the robot exists in the region with higher probability, the low-confidence region represents that the robot is less likely to appear in the region, and in practical application, the robot can form higher constraint with an execution track, task points and auxiliary task points when the robot works normally, and even if the robot is positioned due to missing caused by shielding or sensor disconnection, the robot is parked near the task setting information.

As a specific implementation of the pre-processing module, the pre-processing module is configured to perform the following operations:

step A1: priori environmental map construction

Before executing the operation task, the mobile robot with a perception energy sensor such as a laser radar needs to be controlled to scan the operation environment, so that the prior environment map construction based on the map optimization is completed, the grid map with the specified resolution is obtained, and the grid resolution is 5cm/Pixel selected by the method described in the fff text.

Step A2: multi-resolution grid map construction

In order to minimize the influence of the prior environment map on scanning matching, firstly, invalid points such as noise points, miscellaneous points and the like appearing in the grid map are erased through a drawing tool; because the grid map is a reflection of a real environment, the higher the resolution and the accuracy are, the more accurate the environment description is, but the problems of large search scale, difficult matching and the like exist when the scanning matching is performed under the environment, in order to improve the efficiency and the accuracy of the scanning matching of the subsequent modules, the step is to finish the calculation of a sliding window based on the original 5cm grid map obtained in the step A1 respectively by the search frame length of 10cm, 20cm, 40cm, 80cm, 160cm and 320cm, the 5cm grids with the specified number are stored in the window, and all grid values are covered by the gray maximum value in the window, so that the 7-layer resolution grid map of 5cm, 10cm, 20cm, 40cm, 80cm, 160cm and 320cm is obtained.

Step A3: high and low confidence region partitioning

In the actual task operation process of the robot, information such as a task starting point, a task end point, a special point, a specified path and the like is required to be manually given, the information is specified on an upper computer interface by an operator, as shown in fig. 2, red dots are specified task points of the robot in the whole environment, most of the specified task points are fixed work points for assembly, inspection and distribution of a production line, a yellow origin is a standby point of the robot, a return point after the robot executes a task is used for waiting for next operation, a wine red origin is a robot charging point for state recovery and material supplement of the robot, and a green line is a track tracked by the robot.

If the high-low confidence region division is not configured in the parameters, the multi-resolution grid map and the set task information are issued to a global initialization positioning module and a follow positioning module; on the contrary, the module generates high-low confidence regions based on the set task information, firstly writes the set route into the map in a grid form, and generates a likelihood domain space based on the track, wherein the calculation formula is as follows:

in (x) _free ,y _free ) Is a free space point in the grid map, (x) _line ,y _line ) The method comprises the steps of setting a route midpoint for a grid, setting dist_arg as a width parameter of a likelihood domain, setting score as the distance from a blank grid in a grid map to the set route, wherein the score is higher when the distance from the blank grid to the set route is nearer, the unit refers to likelihood domain space in the range of dist_arg as a high confidence degree area, and the score exceeding the set width of the likelihood domain is-1 and is a low confidence degree area. In order to minimize the time complexity, the whole calculation process is an off-line calculation process, the robot is only required to be subjected to table lookup calculation when being positioned, the shaded part shown in fig. 3 is a high-confidence region of the whole map, and the rest space is a low-confidence region.

The global initialization positioning module is used for controlling the mobile robot to execute a rotating action of rotating 360 degrees in situ, estimating the posture of the mobile robot based on point cloud information acquired by the laser radar and pose information acquired by the IMU, constructing a full-view point cloud map based on the laser point cloud information under a laser radar coordinate system, and performing branch-and-bound scanning matching based on the multi-resolution grid map and a high-confidence region to obtain global positioning.

The global initialization positioning module is used for executing the following steps:

wherein M is _smoot h represents two standsSquare interpolation function for smoothing gray information of a target grid map, T representing an attitude adjustment variable, z _i Measuring information for an ith laser spot;

In this embodiment, the global initialization positioning module is a process of acquiring the global pose under the map coordinate system after being observed by the robot self-perception module. Under the normal condition, due to the influences of factors such as map size, perception observation range and the like, the robot is mostly used for acquiring global positioning from places with priori information such as charging piles, standby points and the like, so that accurate priori pose is provided for follow positioning, but when a mobile robot is used for generating a robot kidnapping or is positioned in an environment without priori features, the robot has the condition of low positioning efficiency and even incapability of positioning. The global initialization positioning module optimizes the existing positioning mode, adopts a panoramic point cloud map with more abundant environmental characteristics as sensing observation information, and performs scanning matching based on a multi-resolution map and a high-confidence region in the pre-processing module so as to acquire global initialization pose in order to improve global positioning efficiency and precision in a large-scale and complex environment. However, interference such as measurement noise still exists in the matching process, so that the module generates a Gaussian pose sequence based on the obtained estimated pose and is used as an initial input particle of particle filtering or a random particle in resampling after particle degradation, the particle set is changed from uniform distribution to Gaussian distribution, and the algorithm robustness is enhanced.

As a specific implementation of the global initialization positioning module, the module is configured to perform the following operations:

step B1: robot panoramic point cloud map construction

Under the conditions of complex confusion and relative spaciousness, the environmental characteristic description of single-frame laser radar data is less, a robot cannot calculate the current position through the characteristics, in order to acquire the maximized point cloud information, the step is to construct a robot all-scene point cloud map, the robot executes self-rotation action at a fixed angular speed, at the moment, the laser radar has inherent measurement distortion, so the laser radar needs to execute motion distortion elimination work when measuring, the rotation angular speed of the robot is constant, the IMU is adopted for distortion correction, the step is to record the attitude angle data of the IMU at a higher frequency, and the correct attitude estimation is completed according to the attitude information of the IMU under the timestamp of each point timestamp of the laser point cloud in a linear interpolation mode, so the environmental description accuracy is improved; and after the rotation action is finished, sequentially reading laser point cloud information based on a laser radar coordinate system to finish the splicing action, and constructing a panoramic point cloud map of the robot.

Step B2: scanning matching process based on multi-resolution grid map and high-confidence region

If the pre-processing module completes the division configuration of the high-low confidence coefficient region, the step firstly executes a branching operation on the grid map with the maximum resolution generated in the pre-processing module, and halving the map x and y so as to divide the whole search space into four minimum regions, if the ratio of the low confidence coefficient region of a certain small region exceeds 2/3, the scanning matching calculation of the region is skipped, and the scanning matching score calculation formula is as follows:

m in the formula _smoot Is a bicubic interpolation function for smoothing gray information of the grid map, T is an attitude adjustment variable, z _i Is a laserPoint measurement information. The whole function is meant to calculate the maximum score of the occupied grid of the all-scene point cloud.

Then, the step finds out a search space with the largest score in the layer of resolution grid map, then completes a new round of branching operation under the range of the search space, and repeats the above operation with the lower layer of resolution grid map until the optimal pose is found;

if the optimal pose is not found in the steps, a full confidence region searching mode is adopted, and global initial pose searching is conducted again. If the pre-processing module does not perform the high-low confidence region division, the step only performs a scanning matching process based on the multi-resolution grid map to calculate the global initialization pose.

The following positioning module is used for carrying out motion updating based on gesture information acquired by the IMU and position information acquired by the odometer to obtain a predicted value of the gesture, carrying out laser radar observation updating based on the predicted value of the gesture, the high confidence degree region and the particle genetic mutation to obtain an observed value of the gesture, and carrying out resampling based on the observed value of the gesture to obtain the gesture.

Wherein, follow the location module and be used for carrying out as follows:

Before resampling, the following positioning module is configured to perform the following steps: according to the weight of the particles, the particles are divided into two groups, one third of the particles with small weight are selected for particle crossing and mutation to add new unknown particles, and the calculation formula is as follows:

In this embodiment, the following positioning module is a process of predicting, observing and updating, by means of non-stop motion, the real-time pose of the computing robot under the map coordinate system after the global initialization positioning module gives the initial pose. The module mainly comprises three units of mobile robot pose motion update based on EKF, laser radar observation update based on high confidence coefficient region and particle genetic mutation, resampling based on high confidence coefficient region, wherein the pose motion update unit adopts high-robustness high-precision EKF algorithm to complete pose integral calculation, and provides more reliable initial information for the subsequent observation update unit; the laser radar observation updating unit based on the high confidence region and the particle genetic mutation takes the high confidence region consisting of the back pile charging point electricity, the to-be-positioned working point and the set path in the pre-processing module as a necessary condition for particle weight observation updating, improves the positioning accuracy of the robot and the calculation efficiency at the same time, introduces a genetic mutation algorithm, solves the particle degradation problem inherent to the particle filtering algorithm, and further improves the positioning robustness of the robot; the particle filtering algorithm has a full-image random resampling process when particle degradation occurs, when the map scale is large or the particle number is small, the robot positioning calculation efficiency and accuracy cannot be guaranteed, and in order to solve the problem, a resampling unit based on a high-confidence region is introduced by the following positioning module.

As a specific implementation of the following positioning module, the module is configured to perform the following operations:

step C1: EKF-based mobile robot pose motion update

When the mobile robot navigates in the known global initial pose, IMU pose and Odom position information are subjected to robot dead reckoning through an EKF algorithm, the robustness of a dead reckoning method is enhanced while the inputness error of a positioning algorithm is reduced, and then dead reckoning pose predicted values, current frame laser radar point clouds and prior obstacle grid maps are input into particle filtering to be unfolded and calculated to output real pose nearby values.

Step C2: laser radar observation updating based on high confidence region and particle genetic mutation

The module changes the mode that the original particle filtering algorithm calculates the particle weight through a likelihood domain observation model, and simultaneously applies a high confidence region to change the particle weight, so that the particles are converged on the tracking track after resampling, and the robot and the tracking track form stronger constraint, so that the positioning precision of the robot in the path tracking process and the rapid convergence of the particles are ensured; in order to slow down the degradation of particles and enhance the robustness of particle filtering, the module applies the idea of genetic mutation to enhance diversity on the premise of ensuring the consistency of particles, the module is divided into two groups according to the weight of the particles before resampling, one third of the small weight particle groups are selected for particle crossing and mutation to add new unknown particles, and the calculation formula is as follows:

Example 2:

the invention relates to a mobile robot positioning method based on laser radar scanning information, which is applied to a mobile robot provided with a perception sensor, wherein the perception sensor comprises an odometer, an IMU and a laser radar and is used for acquiring pose information and environment information, and the method is used for positioning through a positioning system disclosed in an embodiment 1 and comprises the following steps:

s100, pretreatment: constructing a priori grid map based on pose information and environment information, generating a multi-resolution grid map based on the priori grid map, and dividing high-low confidence areas in the multi-resolution grid map according to task information set by a user to obtain high-confidence areas and low-confidence areas, wherein the task information is used for providing motion control and navigation planning for a mobile robot;

s200, global initialization positioning: controlling the mobile robot to execute a rotating action of rotating in situ by 360 degrees, estimating the rotating action of the mobile robot based on point cloud information acquired by a laser radar and pose information acquired by an IMU, constructing a full-view point cloud map based on the laser point cloud information under a laser radar coordinate system, and performing branch-and-bound scanning matching based on a multi-resolution grid map and a high-confidence region to obtain global positioning;

S300 follows the positioning: and carrying out motion updating based on the gesture information acquired by the IMU and the position information acquired by the odometer to obtain a predicted value of the pose, carrying out laser radar observation updating based on the predicted value of the pose, the high-confidence region and the particle genetic mutation to obtain an observed value of the pose, and carrying out resampling based on the observed value of the pose to obtain the pose.

Wherein, the pre-processing in step S100 includes the following operations:

(1) Constructing an a priori grid map: regulating and controlling a mobile robot scanning operation environment, and constructing a priori grid map with a specified resolution based on graph optimization according to environmental information acquired by a sensing sensor;

(2) Constructing a multi-resolution grid map: removing invalid points in the prior grid map through a drawing tool to obtain a preprocessed multi-resolution grid map, and performing sliding window calculation on the prior grid map based on the multiple of the resolution of the prior grid map as the length of a search frame to obtain the multi-resolution grid map;

Step S200 of global initialization positioning includes the following operations:

(1) Constructing a full scenic spot cloud map: controlling a mobile robot to execute a rotation action of rotating in situ for 360 degrees, recording the posture information of the IMU at a preset frequency, estimating the posture according to the time stamps of each point of the laser point cloud and the posture information of the IMU under each time stamp by linear interpolation, sequentially reading the laser point cloud information under a laser radar coordinate system, and splicing the laser point cloud information to obtain a full-scenic point cloud map;

(2) Global initialization pose search: performing global initial pose search in a mode of scanning and matching based on the multi-resolution grid map and the high-confidence region, or performing global initial pose search in a mode of searching based on the full-confidence region;

and L3, selecting a search space with the largest scanning matching score from the four small areas, performing a next round of branching operation on the selected search space, and repeatedly executing the step L2 with a lower-layer resolution grid map until the optimal pose is found.

and M3, selecting a search space with the largest scanning matching score from the four small areas, performing a next round of branching operation on the selected search space, and repeatedly executing the step L2 with a lower-layer resolution grid map until the optimal pose is found.

If the high-low confidence region division is not executed in the pre-processing, the global initial pose search is carried out based on the mode of scanning and matching the multi-resolution grid map, and the method comprises the following steps:

Step S300 of following the positioning includes the following operations:

(1) Motion update: based on the attitude information acquired by the IMU and the position information acquired by the odometer, carrying out motion update on the pose of the mobile robot based on an EKF algorithm to obtain a predicted value of each pose;

(2) And (5) observing and updating: the method comprises the steps of taking a predicted value of a pose, laser point cloud information acquired through a laser radar and a minimum resolution grid map in a multi-resolution grid map as inputs, calculating and outputting an updated value of the pose in a mode of calculating particle weights through a likelihood domain observation algorithm, observing and updating the updated value of the pose based on a high confidence region and a particle genetic mutation algorithm to obtain an observed value of each pose, and constructing particle sets based on the observed value of each pose;

(3) Resampling: and selecting a pose with the largest observation value from the particle set, outputting the pose if the observation value of the pose is larger than a threshold THR, if the observation value of the pose is smaller than or equal to the threshold THR, resampling based on a high confidence region, selecting the pose in the high confidence region, adding the pose into the particle set to obtain a new particle set, selecting the pose with the largest observation value from the new particle set, outputting the pose if the observation value of the pose is larger than the threshold THR, performing global random sampling based on the high confidence region and the low confidence region if the observation value of the pose is smaller than or equal to the threshold THR, randomly selecting the pose, adding the pose into the particle set to obtain the new particle set, selecting the pose with the largest observation value from the new particle set, if the observation value of the pose is larger than the threshold THR, selecting the particle with the highest weight in the particle set, and outputting the particle set to ensure the positioning real-time performance.

Step (3) performs the following operations before resampling: according to the weight of the particles, the particles are divided into two groups, one third of the particles with small weight are selected for particle crossing and mutation to add new unknown particles, and the calculation formula is as follows:

While the invention has been illustrated and described in detail in the drawings and in the preferred embodiments, the invention is not limited to the disclosed embodiments, and it will be appreciated by those skilled in the art that the code audits of the various embodiments described above may be combined to produce further embodiments of the invention, which are also within the scope of the invention.

Claims

1. The mobile robot positioning system based on laser radar scanning information is characterized by being applied to a mobile robot provided with a perception sensor, wherein the perception sensor comprises an odometer, an IMU and a laser radar and is used for acquiring pose information and environment information, the positioning system comprises a pre-processing module, a global initialization positioning module and a following positioning module,

the following positioning module is used for carrying out motion updating based on gesture information acquired by the IMU and position information acquired by the odometer to obtain a predicted value of the gesture, carrying out laser radar observation updating based on the predicted value of the gesture, the likelihood domain observation model, the high confidence region and the particle genetic mutation to obtain an observed value of the gesture, and carrying out resampling based on the observed value of the gesture to obtain an updated gesture.

2. The mobile robot positioning system based on lidar scanning information of claim 1, wherein the pre-processing module interacts with the user through a pre-processing interface for performing the following:

constructing an a priori grid map: the remote control mobile robot scans the working environment, and according to the environmental information acquired by the sensing sensor, a priori grid map with specified resolution is constructed based on graph optimization;

dividing the high-low confidence regions: under the condition of configuring high-low confidence region division, writing task information set by a user into a priori grid map in a grid form, and generating a likelihood domain space based on a track corresponding to the task information, wherein a likelihood domain space calculation formula is as follows:

wherein, (x) _free ,y _free ) For free space points in a priori grid map, (x) _line ,y _line ) For points in the track corresponding to the task information, dist_arg is used as a likelihood domain setting width parameter, score is the distance from a blank grid to a set route in the multi-resolution grid map, a likelihood domain space in the dist_arg range is used as a high confidence interval, and scores exceeding the set width of the likelihood domain are all-1, and the score is used as a low confidence area.

3. The mobile robotic positioning system based on lidar scanning information of claim 1, wherein the global initialization positioning module is configured to perform the following:

if the pre-processing module does not perform the high-low confidence region division, the global initialization module is used for performing global initial pose search based on a mode of scanning and matching a multi-resolution grid map, and the method comprises the following steps:

4. A mobile robot positioning system based on lidar scanning information according to any of claims 1-3, characterized in that the following positioning module is adapted to perform the following:

resampling: and selecting a pose with the largest observation value from the particle set, outputting the pose if the observation value of the pose is larger than a threshold THR, resampling the pose based on a high confidence region if the observation value of the pose is smaller than or equal to the threshold THR, selecting the pose with the largest observation value from the new particle set, outputting the pose if the observation value of the pose is larger than the threshold THR, performing global random sampling based on the high confidence region and the low confidence region if the observation value of the pose is smaller than or equal to the threshold THR, randomly selecting the pose, adding the pose into the particle set, obtaining the new particle set, selecting the pose with the largest observation value from the new particle set, outputting the pose if the observation value of the pose is larger than the threshold THR, and selecting the particle with the highest weight in the particle set to output if the observation value of the pose is smaller than or equal to the threshold THR, so as to ensure the positioning in real time.

5. The mobile robotic positioning system based on lidar scanning information of claim 4, wherein the follow-up positioning module is configured to perform, prior to resampling, the following: according to the weight of the particles, the particles are divided into two groups, one third of the particles with small weight are selected for particle crossing and mutation to add new unknown particles, and the calculation formula is as follows:

6. A method for mobile robot positioning based on lidar scanning information, applied to a mobile robot equipped with a perception sensor comprising an odometer, an IMU and a lidar for acquiring pose information and environmental information, the method for positioning by a mobile robot positioning system based on lidar scanning information according to any of claims 1-5, the method comprising the steps of:

7. The method for mobile robot positioning based on lidar scanning information of claim 6, wherein the pre-processing comprises the operations of:

8. The method for mobile robot positioning based on lidar scanning information of claim 6, wherein globally initializing the positioning comprises:

wherein M is _smoot h represents a bicubic interpolation function for smoothing gray information of a target grid map, T

Representing attitude adjustment variable, z _i Measuring information for an ith laser spot;

wherein M is _smoot h represents a bicubic interpolation function forSmoothing gray information of a target grid map, T representing an attitude adjustment variable, z _i Measuring information for an ith laser spot;

9. The method of mobile robot positioning based on lidar scanning information according to any of claims 6 to 8, wherein following the positioning comprises the following operations:

10. The method for mobile robot positioning based on lidar scanning information of claim 9, wherein the following operations are performed before resampling: according to the weight of the particles, the particles are divided into two groups, one third of the particles with small weight are selected for particle crossing and mutation to add new unknown particles, and the calculation formula is as follows: