Active scene mapping method based on constraint guidance and space optimization strategies

Patent metadata

Field	Value
Publication number	US-12228941-B2
Application number	US-202318464917-A
Country	US
Kind code	B2
Filing date	Sep 11, 2023
Priority date	Mar 3, 2023
Publication date	Feb 18, 2025
Grant date	Feb 18, 2025

Abstract

The present invention proposes an active scene mapping method based on constraint guidance and space optimization strategies, comprising a global planning stage and a local planning stage; in the global planning stage, the next exploration goal of a robot is calculated to guide the robot to explore a scene; and after the next exploration goal is determined, specific actions are generated according to the next exploration goal, the position of the robot and the constructed occupancy map in the local planning stage to drive the robot to go to a next exploration goal, and observation data is collected to update the information of the occupancy map. The present invention can effectively avoid long-distance round trips in the exploration process so that the robot can take account of information gain and movement loss in the exploration process, find a balance of exploration efficiency, and realize the improvement of active mapping efficiency.

First claim

The invention claimed is:

1. An active scene mapping method based on constraint guidance and space optimization strategies, comprising a global planning stage and a local planning stage; in the global planning stage, the next exploration goal of a robot is calculated to guide the robot to explore a scene; and when the next exploration goal is determined, specific actions are generated according to the next exploration goal, the position of the robot and the constructed occupancy map in the local planning stage to drive the robot to go to the next exploration goal, and observation data is collected to update the information of the occupancy map, specifically comprising the following steps:
step 1: generating a state according to the scanning data of the robot;
the state comprises three parts: an occupancy map M t , a geodesic distance map D(M t , ω t ) and a frontier-based entropy I(M t ); and ω t indicates the position of the robot;
s (ω t )=( M t ,D ( M t ,ω t ), I ( M t )

1.1) Occupancy Map;
a 3D scene model is obtained by back projection of the depth with the pose of the robot according to the observation C(ω t ) of the robot in position ω t ; a 2D global map is constructed from the top-down view of the 3D scene model as an occupancy map M t ; at time t, the occupancy map is expressed as M t ∈[0,1] X×Y×2 , and X, Y represent the length of the occupancy map and the width of the occupancy map, respectively; the occupancy map comprises two channels indicating the explored and occupied regions, respectively; the grids in the occupancy map M t are classified as follows: free (explored but not occupied), occupy (occupied) and unknown (not explored); and the frontier grid F t ⊂M t is a free grid adjacent to an unknown grid;
1.2) geodesic distance map;
given the current position ω t and the currently constructed occupancy map M t , a geodesic distance map D(M t , ω t )∈ X×Y is constructed, wherein D x,y (M t ,ω t ) represents the geodesic distance from the position (x, y) to the position ω t of the robot:
D x,y ( M t ,ω t )=dist M t (( x,y ),ω t )

the geodesic distance dist M t is the shortest distance for traversal between two points in the occupancy map M t , and the geodesic distance dist M t is calculated by the fast marching method;
1.3) frontier-based entropy;
the frontier-based entropy is introduced as constraints to reduce the searching space when the height of the occupancy map M t is incomplete; and the frontier point f∈F t represents a next potential optimal exploration goal, a frontier-based entropy I is introduced for encoding the spatial distribution information of the frontier points based on the characteristics of small-range aggregation and large-range dispersion of frontier points, the encoded spatial distribution information is used as one of the inputs of the actor network in the global planning strategy Π to constrain action search, and the frontier-based entropy I is defined as follows:

I

x
,
y

(

M
t

)

=



{

f
∈

F
t

|

dist

M
t

(

f
,

(

x
,
y

)

)

<
γ

}



wherein I x,y (M t ) represents the number of frontier points within the γ×γ neighborhood centered on the position (x,y) of the frontier point in the occupancy map M t ; and the spatial distribution information contained in each frontier point includes the (x,y) coordinate of the position of the point and the statistical information of the spatial distribution of the frontier points within the neighborhood;
step 2: calculating the probability distribution of the action space of the robot according to the state input;
off-policy learning approach proximal policy optimization (PPO) is used as a policy optimizer for training optimization and decision execution of the global planning strategy; and the policy optimizer comprises an actor network and a critic network;
the actor network uses a multi-layer perceptron (MLP) as an encoder for feature extraction and uses a graph neural network for feature fusion; and a graph is constructed according to the frontier point given by the state s(ω t ), and feature extraction and feature fusion are carried out on the constructed graph to obtain the score of the frontier point;
the critic network comprises five convolutional layers, a flatten operation and three linear layers, a ReLu activation function is attached behind each convolutional layer and each linear layer, and the flatten operation is used for flattening multi-dimensional data into one-dimensional data; and the critic network is used for predicting the state value V(s(ω t )) of the occupancy map to indicate the critic value obtained by the current state of the frontier point, and the critic value, as part of the loss function, is used for training the actor network;
the process of calculating the probability distribution of the action space according to the state input is specifically as follow:
a graph G(F t ,Ω t ) is constructed based on the frontier grid F t and the exploration path Ω t ={ω 0 , . . . , ω t } to represent the context information of the current scene, a corresponding relation between the robot and the frontier point extracted from the occupancy map M t is established in the graph G(F t ,Ω t ), and the information given by the state s(ω t ) is assigned to the node and edge of G(F t ,Ω t ); for each node n i , the node input feature feat(n i )∈ 5 includes: (x,y) coordinate information in the occupancy map M t , semantic label information indicating n i ∈F t or n i ∈Ω t , historical label information indicating that n i is the current node at or historical exploration node n i ∈{ω 0 , . . . , ω t−1 }, and a frontier-based entropy I n i (M t ); the node edge feature feat(l ij )∈ 32 is extracted by the multi-layer perceptron (MLP), wherein l ij ∈ 1 represents the geodesic distance from node n j to node n i ; the node input feature and the node edge feature are input into the actor network for feature extraction and feature fusion, and a set of scores of frontier points are output; and the sampling probability Π mask (f|s(ω t )) of each action of the robot is calculated based on the set of scores of frontier points;
step 3: carrying out action mask guided space alignment and selecting the next exploration goal;
the robot selects the frontie

Assignees

• Univ Dalian Tech

Inventors

• Yang Xin
• Yin Xuefeng
• Yin Baocai
• Wei Xiaopeng

Classifications

• G05D1/246

  using environment maps, e.g. simultaneous localisation and mapping [SLAM] · CPC title

• G05D1/644

  Optimisation of travel parameters, e.g. of energy consumption, journey time or distance · CPC title

• G05D1/227

  Handing over between remote control and on-board control; Handing over between remote control arrangements · CPC title

• G05D1/0217

  in accordance with energy consumption, time reduction or distance reduction criteria · CPC title

• G05D1/0088

  characterized by the autonomous decision making process, e.g. artificial intelligence, predefined behaviours (using knowledge based models G06N5/00) · CPC title