CN112828883B

CN112828883B - Robot environment exploration method and system in unknown environment

Info

Publication number: CN112828883B
Application number: CN202011564261.8A
Authority: CN
Inventors: 孟庆虎; 王超群
Original assignee: Shenzhen Research Institute of CUHK
Current assignee: Shenzhen Research Institute of CUHK
Priority date: 2020-12-25
Filing date: 2020-12-25
Publication date: 2022-07-01
Anticipated expiration: 2040-12-25
Also published as: CN112828883A

Abstract

The invention discloses a robot environment exploration method and system in an unknown environment, wherein the method comprises the following steps: s1, carrying out environmental point cloud expression modeling; s2, constructing a topological map in a sensing visual range; s3, evaluating the search result gain under the expression of a topological map: namely, information gain evaluation under topological environment expression; and S4, motion decision based on information gain and the length of the planning path. The invention can enable the mobile robot to efficiently and accurately collect information and realize environment exploration and modeling. The method can be widely applied to environment exploration modeling and monitoring tasks based on the mobile robot, and has wide potential application prospects in a plurality of scenes such as disaster relief, home services and the like.

Description

Robot environment exploration method and system in unknown environment

Technical Field

The invention relates to the technical field of autonomous navigation of mobile robots, in particular to a method and a system for exploring and modeling an autonomous environment of a mobile robot.

Background

Mobile robots are increasingly used for tasks related to information collection, such as environmental monitoring and surveillance tasks. In the process, the robot can automatically collect the environmental information, so that the robot can be used for potential subsequent applications. At present, most monitoring robots collect information according to preset routes, and the method has low efficiency and cannot automatically and flexibly adapt to requirements according to environments.

In recent years, autonomous information collection algorithms based on mobile robots and related applications have been developed. However, the efficiency of path planning and the effectiveness of motion control decisions during exploration still remain to be improved. How to achieve environment modeling and monitoring with less time and less energy consumption to obtain more information is a long-sought goal of the research and related applications.

Disclosure of Invention

In order to make up for the defects of the prior art, the invention provides a robot environment exploration method in an unknown environment, and the exploration of the environment is completed quickly and accurately.

Therefore, the robot environment exploration method under the unknown environment provided by the invention comprises the following steps: s1, environmental point cloud expression modeling: constructing a point cloud topological environment expression under the incremental environment point cloud model for the established incremental environment point cloud model; s2, constructing a topological map in a sensing visual range: based on the point cloud topological environment expression, constructing a topological map expression, namely: establishing a topologically expressed path diagram along with the detection process in the visual range of a sensor carried by the robot; s3, evaluating the search result gain under the expression of a topological map: namely, information gain evaluation under topological environment expression; s4, motion decision based on information gain and planning path length: the method comprises the steps of exploring target point decision generation and motion planning, and after an information gain value of a planned exploring position is obtained, motion planning decision is made based on the information gain and the motion cost of reaching the target point.

In some embodiments, the following features are also included:

in step S2, a spatial position random point is generated within the sensor range. Connecting the random points based on a space barrier-free connection criterion, thereby constructing a topological graph structure of a complex environment to be explored; the graph structure is composed of generated spatially random nodes and inter-node connecting edges.

In step S3, the information gain of each generated node takes the degree of information entropy change as an evaluation criterion.

The entropy of the existing model of the environment is calculated according to a Shannon entropy formula, and the conditional entropy under the observation condition corresponding to the estimated exploration position and the position is calculated at the same time, so that the difference between the entropy of the existing model and the calculated conditional entropy is calculated, and the difference is used as an information gain value of the estimated exploration plan.

In step S4, the movement cost is measured by taking the distance between the current position and the target position of the robot as a metric, and the topological topography map constructed in step S2 prevents the robot from searching for a path in the whole space, thereby improving the efficiency of path search.

In planning decision, the maximum information gain and the minimum distance are used as optimization objective functions, the maximum information gain and the minimum distance are linearly superposed, coefficients of the linear superposition are determined through debugging in different environments, and positions to be explored are searched from topological nodes to optimize a cost function, so that the robot environment is independently explored.

Step S2 specifically includes: generating space position random points in a sensor range, randomly acquiring some points in the space to obtain space random position points, and connecting the random points based on a space barrier-free connection criterion to construct a topological graph structure for exploring a complex environment, wherein the topological graph structure is composed of generated space random nodes and connecting edges among the nodes.

Step S3 specifically includes: firstly, calculating the entropy Encopy (i) of the existing model of the environment according to a Shannon entropy formula; the position i of each point cloud in the environment has a certain occupation probability p, the Entropy of the map can obtain Encopy (i) through a Shannon formula, the conditional Entropy Encopy (i, x) under the condition of observation obtained under the estimated exploration position and the corresponding position is calculated, x is the position expected to be explored, the difference between the Entropy of the existing model and the calculated conditional Entropy is calculated, and the difference is used as an information gain value of the estimated exploration plan

I＝Entropy(i)-Entropy(i,x)。

Step S4 specifically includes: obtaining the motion cost by inquiring the topological road map, wherein the motion cost takes the distance between the current position and the target position of the robot as a measurement standard; inquiring on the topological road map to find a route from the current position of the robot to a target point, and obtaining a line segment consisting of a plurality of nodes and line segments, wherein the length of the line segment is the movement cost from the current position of the robot to the target position and is marked as tau; in planning decision, the position to be explored is searched from the topological nodes by taking the maximum information gain and the minimum distance as optimization objective functions to optimize a cost function cost ═ Ie ^ (lambda tau), wherein lambda is an adjustable coefficient, so that the robot environment autonomous exploration is realized.

The invention also provides a robot environment exploration system in an unknown environment, and the robot environment exploration method is adopted.

Compared with the prior art, the invention has the advantages that:

according to the technical scheme, the robot is helped to quickly judge the target point by combining the information gain of the map constructed in the exploration process and the path length from the robot to each decision point, so that the exploration on the environment is quickly and accurately completed.

The topological road map provided by the invention can help the robot to rapidly plan a path in space. By effectively evaluating the length of the planned path and the obtained information, the efficient implementation of exploration motion planning is realized. And further, by utilizing the motion planning result, the environment exploration efficiency and the high-precision real-time operation of environment modeling are realized. Meanwhile, the invention has been tested in different environments, and the result shows that the method has better performance compared with the existing method.

The invention can assist the robot to quickly complete the construction of the unknown environment, thereby assisting the robot in navigating in the unknown environment, and simultaneously, quickly and accurately providing effective information in the unknown environment for the human, so that the invention can be widely applied to environment exploration modeling and monitoring tasks based on the mobile robot, and has wide potential application prospect in a plurality of scenes such as disaster relief, home service and the like.

Drawings

FIG. 1 is a schematic flow diagram of a system according to an embodiment of the present invention;

FIG. 2a is an environmental diagram of an embodiment of the present invention in practice;

FIG. 2b is an environmental point cloud MAP constructed by RTAB-MAP according to an embodiment of the present invention;

fig. 3a and fig. 3b are schematic diagrams illustrating a topology node environment expression construction provided in an embodiment of the present application;

FIG. 4a is experimental verification test experimental hardware, including a quad-rotor drone platform, of an embodiment of the present invention;

FIG. 4b is an experimental test scenario of an embodiment of the present invention;

FIG. 4c is an exploration process and environmental modeling results implemented with the method of an embodiment of the present invention.

Detailed Description

The following further describes the embodiments of the present invention with reference to the drawings and preferred embodiments. It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict.

It should be noted that the terms of orientation such as left, right, up, down, top and bottom in the present embodiment are only relative concepts to each other or are referred to the normal use state of the product, and should not be considered as limiting.

The embodiment of the invention aims to develop an autonomous exploration strategy based on a road map, so that a mobile robot can efficiently and accurately collect information and realize environment exploration and modeling.

The technical problems to be solved include:

(1) the problems of information collection and environment exploration in a complex environment are solved;

(2) solving the problems of path planning and information gain evaluation and providing information gain evaluation quantitative indexes;

(3) the problem of exploration visual angle planning in the autonomous exploration process of the mobile robot is solved;

the technical problem of the embodiment is solved by the following technical scheme:

in the present embodiment, the planned path length and the collected environment information are considered in the environment exploration algorithm of the present embodiment.

Firstly, the algorithm is used for establishing a topological route map of the incremental environment point cloud model in the exploration process. The incremental point cloud model is generated by an existing depth camera based composition modeling algorithm (RTAB-MAP). And constructing a topological environment expression under the model based on the established environment point cloud model.

And secondly, constructing a topological map expression based on the existing point cloud environment expression. And establishing a path diagram of topological expression along with the detection process in the visual range of the sensor carried by the robot. In this process, spatially positioned random points are first generated within the sensor range. And connecting the random points based on a space barrier-free connection criterion, thereby constructing a topological graph structure of complex environment to be explored. The graph structure is composed of generated spatially random nodes and inter-node connecting edges. In the subsequent information gain evaluation, the node is used as a unit for development, so that the calculation cost of the system is reduced, and the real-time performance of the system is improved.

And thirdly, evaluating information gain under the topological environment expression. In this embodiment, the information gain of each generated node takes the degree of information entropy change as an evaluation criterion. Specifically, the present embodiment first calculates the entropy of the existing model of the environment according to the shannon entropy formula, and simultaneously calculates the conditional entropy under the estimated exploration position and the corresponding obtained observation condition under the position. Thereby calculating the difference between the existing model entropy and the calculated conditional entropy. And the difference is used as the information gain value of the evaluated exploration plan.

And fourthly, exploring target point decision generation and motion planning. And after the information gain value of the planned exploration position is obtained, carrying out motion planning decision based on the information gain and the motion cost of reaching the target point. The motion cost is measured by the distance between the current position of the robot and the target position. The topological topographic map constructed in the second step prevents the robot from searching paths in the whole space, and the efficiency of path searching is improved. In planning decision, the maximum information gain and the minimum distance are used as optimization objective functions, the maximum information gain and the minimum distance are linearly superposed, coefficients of the linear superposition are determined through debugging in different environments, and positions to be explored are searched from topological nodes to optimize a cost function, so that the robot environment is independently explored.

Fig. 1 is a flow chart of the system of this embodiment, and in this block diagram, the following steps are included:

in step S1, an environment model is first constructed based on the depth vision camera by the existing RTAB-MAP algorithm. The model has large storage data volume but rich information. And utilizing the initial point cloud expression of the environment obtained by the exploration to provide input for a subsequent algorithm.

As shown in fig. 2a and 2b, fig. 2a shows an actual environment MAP, and fig. 2b shows an environment point cloud MAP constructed by RTAB-MAP. The point cloud map is a discretized representation of a complex spatial environment.

And step S2, constructing a topological node map corresponding to the model by using the established environment point cloud model. And establishing a path diagram of topological expression along with the detection process in the visual range of the sensor carried by the robot. In this process, first, a random point of spatial position is generated within the sensor range, which is the range that can be sensed by the sensor, as shown in fig. 3 a. And randomly taking some points in the space to obtain spatial random position points. And connecting the random points based on a space barrier-free connection criterion (the space barrier-free criterion is that for a connecting line between any two points in the space, if the position of each point on the connecting line is not coincident with the position of an obstacle in the space, the connecting line is considered to be barrier-free), thereby constructing a topological graph structure of the complex environment to be explored, as shown in fig. 3b, black points represent points obtained by random sampling, black connecting lines are barrier-free, and the points and lines form a topological graph in the space. The graph structure is composed of generated spatially random nodes and inter-node connecting edges. In the subsequent information gain evaluation, the node is used as a unit for development, so that the calculation cost of the system is reduced, and the real-time performance of the system is improved.

And step S3, evaluating the information gain under the topological environment expression. In this embodiment, the information gain of each generated node takes the degree of information entropy change as an evaluation criterion. Specifically, the present embodiment first calculates the entropy of the existing model of the environment according to the shannon entropy formula. The position i of each point cloud in the environment has a certain occupation probability p, the Entropy of the map can obtain control (i) through shannon formula, and the conditional Entropy of the estimated exploration position and the corresponding obtained observation condition of the position is simultaneously calculated, wherein x is the position expected to be explored. Thereby calculating the difference between the entropy of the existing model and the calculated conditional entropy. And the difference is used as the information gain value I ═ control (I) -control (I, x) for the evaluated exploration plan.

Step S4, exploration target point decision generation and movement planning. And after the information gain value of the planned exploration position is obtained, performing motion planning decision based on the information gain and the motion cost of reaching the target point. The movement cost takes the distance between the current position and the target position of the robot as a measurement standard, and the movement cost can be obtained by inquiring a topological road map. As can be seen from fig. 3b, in order to find a route from the current position of the robot to the target point, a query may be directly performed on the topological road map to obtain a line segment composed of a plurality of nodes and line segments, and the length of the line segment is the movement cost from the current position of the robot to the target position, and is denoted as τ. In planning decision, the position to be explored is searched from the topological nodes by taking the maximum information gain and the minimum distance as optimization objective functions to optimize a cost function cost ═ Ie ^ (lambda tau), wherein lambda is an adjustable coefficient, so that the robot environment autonomous exploration is realized.

Fig. 3a and 3b are schematic diagrams of environment expression and construction of the topology node according to this embodiment.

In fig. 3a and 3b, r1 and r2 are the visible range parameters of the sensor mounted on the mobile robot, and these two parameters determine the range of the environment that the mobile robot can sense. The cylinder is illustrated as an obstacle (obsacle) that expresses the obstacle encountered during planning in this space. In fig. 2a, the identified light purple cone space is the perception space (sensor scope) of the robot, and the dotted line is the sensor visual line schematic (sensor beam).

In fig. 3b, nodes (nodes) are random nodes generated in the sensing space, and an edge (edge) is established between the nodes by using accessible and barrier-free judgment. The reachability is judged whether the straight line connection between the two nodes passes through the barrier: if not, judging that the connection is feasible; if so, the connection is judged to be not feasible. Therefore, a topological map is established in the perception space, and subsequent information gain calculation and motion planning decision based on the nodes are realized.

4a, 4b, 4c, example of the results of the test experiments of the present algorithm

Fig. 4a is the present algorithm test experimental hardware, including a quad-rotor drone platform. A drone onboard computer (Raspberry Pi), a drone motion controller (Pixhawk), and a drone onboard depth vision sensor (RealSense).

Fig. 4b is an experimental test scenario, which includes a limited exploration space formed by the ceiling and surrounding walls, and exploration process obstacles formed by pillars.

Fig. 4c shows the exploration process and the environment modeling result of the algorithm. Wherein the exploration path starting point (Start) and ending point (End) are identified in yellow. The red line is the explored planned path obtained by the algorithm of the present embodiment. And the model formed by the small blocks is an environment model obtained by exploring and constructing the robot.

Compared with the traditional method, the modeling method can obviously improve the mapping efficiency and the mapping precision. Importantly, in the process of drawing, the method does not need to traverse the whole environment to obtain an exploration path, and only needs to query a topological road map to obtain a path from the position of the robot to the target position, so that the target nodes are quickly evaluated, the decision effectiveness is improved, more information can be obtained by using shorter energy, and the environment is quickly and effectively monitored.

The above description is a detailed description of the present embodiment with reference to specific preferred embodiments, and it should not be construed that the specific implementations of the present embodiment are limited to these descriptions. For those skilled in the art to which the embodiment belongs, several equivalent substitutions or obvious modifications can be made without departing from the concept of the embodiment, and the same performance or purpose should be considered as belonging to the protection scope of the embodiment.

Claims

1. A robot environment exploration method under an unknown environment is characterized by comprising the following steps:

s1, environmental point cloud expression modeling: constructing a point cloud topological environment expression under the incremental environment point cloud model;

s2, constructing a topological map in a sensing visual range: based on the point cloud topological environment expression, constructing a topological map expression, namely: establishing a topologically expressed path diagram along with the detection process in the visual range of a sensor carried by the robot;

s3, evaluating the search result gain under the expression of a topological map: namely, information gain evaluation under topological environment expression;

s4, motion decision based on information gain and planning path length: the method comprises the steps of exploring target point decision generation and motion planning, wherein after an information gain value of a planned exploring position is obtained, motion planning decision is made based on the information gain value and the motion cost of reaching the target point;

in step S2, generating spatial position random points in the range of the sensor, and connecting the random points based on the space barrier-free connection criterion, thereby constructing a topological graph structure of complex environment to be explored; the graph structure is formed by generated space random nodes and connecting edges among the nodes;

step S3 specifically includes: firstly, according to a Shannon entropy formula, calculating entropy Encopy (i) of an existing model of an environment; the position i of each point cloud in the environment has a certain occupation probability p, the Entropy of the map can obtain Encopy (i) through a Shannon formula, and meanwhile, the conditional Entropy Encopy (i, x) under the estimated exploration position and the corresponding obtained observation condition under the position is calculated, wherein x is the position expected to be explored, so that the difference value between the existing model Entropy and the calculated conditional Entropy is calculated, and the difference value is used as an information gain value of the estimated exploration plan:

I= Entropy(i)- Entropy(i,x)；

step S4 specifically includes:

obtaining the movement cost by inquiring a topological map, wherein the movement cost takes the distance between the current position and the target position of the robot as a measurement standard;

inquiring on a topological map to find a route from the current position of the robot to a target point and obtain a line segment consisting of a plurality of nodes and line segments, wherein the length of the line segment is the movement cost from the current position of the robot to the target position and is marked as tau;

in planning decision, with maximized information gain and minimized distance as optimization objective functions, searching a position to be explored from a topological node to optimize a cost function cost = Ie ^ (lambda tau), wherein lambda is an adjustable coefficient, and therefore autonomous exploration of the robot environment is achieved.

2. The robot environment exploring method according to claim 1, wherein: in step S3, the information gain of each generated node takes the degree of information entropy change as an evaluation criterion.

3. The robot environment exploring method according to claim 2, wherein: the entropy of the existing model of the environment is calculated according to a Shannon entropy formula, and the conditional entropy under the observation condition corresponding to the estimated exploration position and the position is calculated at the same time, so that the difference value between the entropy of the existing model and the calculated conditional entropy is calculated, and the difference value is used as an information gain value of the estimated exploration plan.

4. The robot environment exploring method according to claim 1, wherein: in the step S4, the movement cost is measured by using the distance between the current position and the target position of the robot as a metric, and the topological map constructed in the step S2 prevents the robot from searching a path in the whole space, thereby improving the efficiency of path search.

5. The robot environment exploring method according to claim 4, wherein: in planning decision, the maximum information gain and the minimum distance are used as optimization objective functions, the maximum information gain and the minimum distance are linearly superposed, coefficients of the linear superposition are determined through debugging in different environments, and positions to be explored are searched from topological nodes to optimize a cost function, so that the robot environment is independently explored.

6. The robotic environment exploration method of claim 1, wherein: step S2 specifically includes: generating space position random points in the range of a sensor, randomly acquiring some points in the space to obtain space random position points, and connecting the random points based on a space barrier-free connection criterion to construct a topological graph structure of complex environment to be explored, wherein the topological graph structure is composed of generated space random nodes and connecting edges among the nodes.

7. A robot environment exploration system in an unknown environment, characterized in that: a method for exploring the environment of a robot according to any one of claims 1 to 6.