Robot social navigation system and method in man-machine coexistence environment

Abstract

The invention provides a robot social navigation system and method in a man-machine coexistence environment, and belongs to the technical field of robot autonomous navigation and man-machine coexistence, and the system comprises a human recognition and speed calculation module which detects the position, orientation and speed information of a human individual in the environment in real time through an RGB-D camera or a depth sensor; the dynamic comfort space modeling module is used for constructing a comfort space of a human individual based on a Gaussian mixture model; the social group comfort space modeling module is used for judging an interaction state and generating a group comfort space according to the orientation and relative position relation of human individuals; the cost map fusion module is used for embedding the constructed individual comfort space and group comfort space into a global cost map and a local cost map of the robot navigation system in real time in a cost function form; and the path planning and control module is used for generating a final navigation path based on the global cost graph and the local cost graph. The problems of single comfortable space modeling and insufficient social scene adaptability in the prior art are solved.

Description

Technical Field

[0001] This invention belongs to the field of robot autonomous navigation and human-robot coexistence technology, specifically relating to a robot social navigation system and method in a human-robot coexistence environment, applicable to safe, natural and socially compliant mobile robot path planning and behavior control in indoor human-robot coexistence environments. Background Technology

[0002] With the rapid development of robotics technology, mobile robots are increasingly widely used in service, healthcare, security, and home applications, and human-robot coexistence environments have become an important direction for robotics research. Traditional mobile robot navigation systems primarily employ geometry-based path planning methods, such as the A* algorithm, Dijkstra's algorithm, and Dynamic Window (DWA) method, with the core objective of achieving shortest path planning and obstacle avoidance. However, these geometry-first navigation methods have significant shortcomings in real-world human-robot coexistence scenarios, failing to meet human needs for socialized behavior and psychological comfort.

[0003] In human social activities, individuals typically maintain an invisible "psychological comfort space," which is influenced by various factors such as cultural background, environmental factors, individual personality, and interactive relationships. When robots move within this space, if their behavior lacks social cognition or appears abrupt, it often causes discomfort or even resistance in humans. Therefore, how to ensure both safety and efficiency for robots during movement while adhering to human behavioral norms and psychological boundaries has become a key issue in social navigation research.

[0004] Currently, human perception and social navigation methods are mainly divided into three categories: The first is the method using a fixed comfort space model. This type of method defines the human comfort space through a simple circular or elliptical region. However, due to the use of a static model, it fails to consider the impact of human motion state, orientation, and speed changes on the comfort space, resulting in unstable performance in dynamic scenarios. The second is the comfort space modeling method based on a single Gaussian distribution. This method uses a Gaussian function for probabilistic modeling. Although it can express the comfort distribution well in scenarios where a single person is stationary or walking at low speed, its simple shape and insufficient directionality make it difficult to accurately reflect the real spatial characteristics of humans in turning, avoiding obstacles, and complex social states. The third is the social navigation method based on reinforcement learning or neural networks. Although this type of method has a certain degree of adaptability, it requires a large amount of real interaction data for training, resulting in high model complexity and poor interpretability, making it difficult to directly apply to practical robot systems. In addition, existing research generally suffers from the limitation of lacking social interaction modeling. Most methods treat humans as independent individuals and ignore human-to-human interaction behaviors, causing robots to easily enter human communication areas and cause interference. Summary of the Invention

[0005] In view of this, to address the problems of simplistic comfort space modeling and insufficient adaptability to social scenarios in existing technologies, this invention proposes a robot social navigation system and method for human-robot coexistence environments. This system utilizes a six-directional Gaussian mixture model to asymmetrically model the dynamic comfort space of a single person, and combines this with the human's movement speed, orientation, and social status to construct an adaptively changing comfort space. By embedding the comfort space as a cost function into the ROS global and local navigation cost graphs, this invention achieves socialized navigation control that balances safety, naturalness, and real-time performance. This enables intelligent, safe, and comfortable navigation for robots in complex human-robot coexistence environments, providing a new technical approach for the application of mobile robots in such environments.

[0006] Therefore, the present invention provides the following technical solution: On one hand, the present invention provides a robot social navigation system in a human-machine coexistence environment, the system comprising: The human recognition and velocity calculation module is used to detect the position, orientation, and velocity information of individuals in the environment in real time using an RGB-D camera or depth sensor. The dynamic comfort space modeling module, connected to the human recognition and speed calculation module, is used to construct the comfort space of an individual human based on a Gaussian mixture model, wherein the space of the human body in the front, back, left, right, left front and right front directions is modeled by asymmetric two-dimensional Gaussian functions in six directions. The social group comfort space modeling module is connected to the human recognition and speed calculation module. When a social relationship is detected between two or more individuals, the module determines the interaction state and generates a group comfort space based on the orientation and relative position of the individuals. The cost map fusion module is connected to the dynamic comfort space modeling module and the social group comfort space modeling module respectively, and is used to embed the constructed individual comfort space and group comfort space into the global cost map and local cost map of the robot navigation system in real time in the form of cost function; The path planning and control module, connected to the cost map fusion module, is used to generate the final navigation path based on the global cost map and the local cost map, and to achieve socialized navigation control that balances safety and human comfort by minimizing the path cost function.

[0007] Furthermore, the Gaussian mixture model includes forward weighting coefficients and backward suppression coefficients. Furthermore, the human recognition and speed calculation module is specifically used for: Environmental information is collected using an RGB-D camera or depth sensor; Human individuals in the environment are detected in real time using the YOLO detection network, which is based on deep learning. The detected human keypoints or bounding box coordinates are converted into positions in the world coordinate system, and the velocity vector and orientation angle of the individual are calculated based on the position difference between consecutive frames to obtain the human's position, movement direction and dynamic state in the environment.

[0008] Furthermore, the comfort space of the individual human includes an asymmetric two-dimensional Gaussian distribution in six directions: front, back, left, right, left front, and right front, with the standard deviation of the comfort radius in each direction dynamically adjusted according to the individual's speed.

[0009] Furthermore, the interaction states include non-communication, face-to-face communication, and observing the same object from different angles and communicating; the social group comfort space modeling module is specifically used for: If two people have no orientation and no common intersection, they are defined as not communicating, and their individual human comfort spaces are simply merged. If two people are facing each other, it is defined as face-to-face communication, generating an elliptical shared comfortable space; If the two orientations are the same, it is defined as observing the same object from different angles and communicating with it, and the comfortable space is merged into a circular area; If there are more than two people, calculate the average position of each individual as the center of the group and construct a circular comfortable space.

[0010] Furthermore, the path planning and control module is specifically used for: In the global planning phase, the A* algorithm is used to calculate the optimal path based on the global cost graph to obtain the global planning result; During the local planning phase, the dynamic window DWA algorithm is used to adjust the robot's movement in real time, and the local cost map is dynamically updated based on the latest individual human position and speed. If the local cost map detects a sudden approach of a human or a change in the social area, an immediate replanning mechanism is triggered to pause or detour.

[0011] In another aspect, the present invention also provides a method for robot social navigation in a human-machine coexistence environment, the method comprising: Real-time detection of the position, orientation, and speed of individuals in the environment using RGB-D cameras or depth sensors; Human individual comfort space is constructed based on Gaussian mixture model, in which the space of the human body in the front, back, left, right, left front and right front directions is modeled by asymmetric two-dimensional Gaussian functions in six directions; When a social relationship is detected between two or more individuals, the interaction status is determined and a comfortable group space is generated based on the orientation and relative position of the individuals. The constructed individual comfort space and group comfort space are embedded in the global cost map and local cost map of the robot navigation system in real time in the form of cost functions; The final navigation path is generated based on the global cost graph and the local cost graph. Socialized navigation control that balances safety and human comfort is achieved by minimizing the path cost function.

[0012] Furthermore, the location, orientation, and velocity information of individuals in the environment are detected in real time using RGB-D cameras or depth sensors, including: Environmental information is collected using an RGB-D camera or depth sensor; Human individuals in the environment are detected in real time using the YOLO detection network, which is based on deep learning. The detected human keypoints or bounding box coordinates are converted into positions in the world coordinate system, and the velocity vector and orientation angle of the individual are calculated based on the position difference between consecutive frames to obtain the human's position, movement direction and dynamic state in the environment.

[0013] Furthermore, the interaction states include non-communication, face-to-face communication, and observing and communicating about the same object from different angles; when a social relationship is detected between two or more individuals, the interaction state is determined and a comfortable group space is generated based on the orientation and relative position of the individuals, including: If two people have no orientation and no common intersection, they are defined as not communicating, and their individual human comfort spaces are simply merged. If two people are facing each other, it is defined as face-to-face communication, generating an elliptical shared comfortable space; If the two orientations are the same, it is defined as observing the same object from different angles and communicating with it, and the comfortable space is merged into a circular area; If there are more than two people, calculate the average position of each individual as the center of the group and construct a circular comfortable space.

[0014] Further, generating the final navigation path based on the global cost graph and the local cost graph includes: In the global planning phase, the A* algorithm is used to calculate the optimal path based on the global cost graph to obtain the global planning result; During the local planning phase, the dynamic window DWA algorithm is used to adjust the robot's movement in real time, and the local cost map is dynamically updated based on the latest individual human position and speed. If the local cost map detects a sudden approach of a human or a change in the social area, an immediate replanning mechanism is triggered to pause or detour.

[0015] Advantages and positive effects of this invention: This invention achieves natural obstacle avoidance and social adaptation of robots in complex human-robot coexistence environments through dynamic comfort space modeling based on a six-directional Gaussian mixture model (SGMM). Its main technical advancements include: establishing a unified cost graph fusion framework, enabling seamless integration of comfort space information with traditional navigation modules; simultaneously incorporating human movement direction, speed, and interaction state into comfort space modeling to achieve high-fitness human behavior perception; and discussing comfort space construction for different interaction scenarios between two individuals and a group comfort space construction method independent of North Korea. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a structural diagram of a socialized robot navigation system with human comfort space modeling in an embodiment of the present invention; Figure 2 This is a flowchart of a robot social navigation method for human-robot coexistence environments, as described in this embodiment of the invention. Figure 3 This is a schematic diagram of the comfortable space for pedestrian movement in an embodiment of the present invention; Figure 4 This is a schematic diagram of the comfort space modeling for two people in different communication states in an embodiment of the present invention; Figure 5 This is a schematic diagram of the comfortable space modeling for two people observing the same object from different positions in an embodiment of the present invention; Figure 6 This is a schematic diagram of a comfortable space model for a group of people in an embodiment of the present invention; Figure 7 This is a diagram showing the navigation experiment results under a simulation environment in an embodiment of the present invention; Figure 8 This is a diagram illustrating the navigation experiment of an actual robot in an indoor human-robot coexistence environment, as described in this invention. Figure 9 To visualize the questionnaire survey results for various comfort space construction scenarios involving dynamic individuals, two static individuals, and multiple static individuals using the egg-shaped comfort space method and the method in the embodiments of the present invention. Detailed Implementation

[0018] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0019] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0020] like Figure 1 As shown, a robot social navigation system for human-robot coexistence environments includes: The human recognition and velocity calculation module is used to detect the position, orientation, and velocity information of individuals in the environment in real time using an RGB-D camera or depth sensor. In specific implementation, the human recognition and velocity calculation module is used to: collect environmental information using an RGB-D camera or depth sensor; detect human individuals in the environment in real time using the YOLO deep learning-based detection network; convert the coordinates of the detected human keypoints or bounding boxes into positions in the world coordinate system, and calculate the individual's velocity vector and orientation angle based on the position difference between consecutive frames, thereby obtaining the human's position, direction of movement, and dynamic state in the environment. This accurately obtains the human's position, direction of movement, and dynamic state in the environment, providing necessary input parameters for subsequent comfort space modeling. By introducing velocity and orientation information, this invention enables the robot not only to "see" the human's position but also to "understand" their movement trends, thus achieving more natural obstacle avoidance planning.

[0021] The dynamic comfort space modeling module, connected to the human recognition and speed calculation module, is used to construct the comfort space of an individual human based on a Gaussian mixture model. It models the space of the human body in the front, back, left, right, left front and right front directions using asymmetric two-dimensional Gaussian functions in six directions.

[0022] In practice, for each detected human individual, a dynamic comfort space is constructed based on a six-directional Gaussian mixture model. This space contains an asymmetric two-dimensional Gaussian distribution in six directions: front, back, left, right, left-front, and right-front. The standard deviation of the comfort radius in each direction is dynamically adjusted according to the individual's speed. The function of this module is to model the human comfort space, enabling the robot to automatically maintain a reasonable safe distance during navigation. By dynamically adjusting the standard deviation, a mixture model with six directional distributions is constructed, achieving an adaptive comfort space description for pedestrians at different speeds, significantly improving the naturalness of the interaction.

[0023] To more accurately describe the psychological differences between forward and backward directions, preferably, in this embodiment, a forward weighting coefficient and a backward inhibition coefficient are introduced into the six-direction Gaussian model to enhance the robot's awareness of avoiding obstacles in the space in front, which conforms to the rules of human interaction.

[0024] The social group comfort space modeling module, connected to the human recognition and speed calculation module, is used to determine the interaction state and generate a group comfort space when a social relationship is detected between two or more individuals, based on the orientation and relative position of the individuals. The cost map fusion module is connected to the dynamic comfort space modeling module and the social group comfort space modeling module respectively. It is used to embed the constructed individual comfort space and group comfort space into the global cost map and local cost map of the robot navigation system in real time in the form of cost function. The path planning and control module, connected to the cost map fusion module, is used to generate the final navigation path based on the global cost map and the local cost map. It achieves socialized navigation control that balances safety and human comfort by minimizing the path cost function. In specific implementation, the path planning and control module is used to: in the global planning stage, use the A* algorithm to calculate the optimal path based on the global cost map to obtain the global planning result; in the local planning stage, use the dynamic window DWA algorithm to adjust the robot's movement in real time, and dynamically update the local cost map according to the latest human individual position and speed; if the local cost map detects that a human suddenly approaches or the social area changes, it triggers an immediate replanning mechanism to pause or detour.

[0025] Global path planning and local obstacle avoidance are coordinated through a "target point interpolation" mechanism. The local planner reassesses the human state every 10ms. If a new human target is detected entering the planning range, the local path is corrected in real time, achieving a truly real-time social response.

[0026] like Figure 2 As shown in the figure, an embodiment of the present invention provides a robot social navigation method in a human-robot coexistence environment, the method comprising: S1. Real-time detection of the position, orientation, and speed of human individuals in the environment using an RGB-D camera or depth sensor; Human detection and localization uses an RGB-D depth camera and a YOLOv11 detection network to identify and locate human individuals in the scene, calculate their position and velocity in the world coordinate system, and determine their direction of movement.

[0027] In practice, S1 is executed according to the following steps: S11. Collect environmental information using an RGB-D camera or depth sensor; S12. Real-time detection of human individuals in the environment using the YOLO detection network based on deep learning; S13. The detected human key points or bounding box coordinates are converted into positions in the world coordinate system, and the velocity vector and orientation angle of the individual are calculated based on the position difference between consecutive frames to obtain the human's position, movement direction and dynamic state in the environment.

[0028] S2. Construct a human individual comfort space based on a Gaussian mixture model, in which the space of the human body in the front, back, left, right, left front and right front directions is modeled by asymmetric two-dimensional Gaussian functions in six directions; For a single moving individual, a six-directional asymmetric Gaussian model is used to define the comfort zone, and its cost function is expressed as: Among them, standard deviation The definition is as follows: f is the direction coefficient Basic standard deviation With parameter A being normalized to 1, β being the target point's orientation angle relative to the individual, and d being the spatial distance, the six-directional standard deviation is dynamically updated based on velocity v to achieve adaptive adjustment of the comfort space. A schematic diagram of the comfort space during pedestrian movement is shown below. Figure 3 As shown.

[0029] S3. When a social relationship is detected between two or more individuals, the interaction status is determined and a comfortable group space is generated based on the orientation and relative position of the individuals. The interactive states include non-communication, face-to-face communication, and observing and communicating about the same object from different angles. A schematic diagram of the comfortable space model for different communication states of two people is shown below. Figure 4 As shown in the diagram, this illustrates a comfortable space modeling diagram where two people observe the same object from different positions. Figure 5 As shown, when the robot detects two or more people approaching, it determines their social status based on individual orientation parameters and positional relationships: if they are interacting face-to-face, an elliptical joint space is generated, with a semi-minor axis of length 1 and a semi-major axis of length [missing value]. L represents the distance between two people. If the two people are not interacting, the two comfort spaces are simply merged, and the comfort space when a person is stationary is represented as a circle with a radius of 1. If the two people are observing the same object, a shared circular space is generated with the intersection point as the center, and the radius of the circle is the distance between the two people. If there are more than three people, the average coordinates of each member are calculated as the group center, and the radius is determined by the average mutual distance to plan the circular comfort space. A schematic diagram of the comfort space modeling for a multi-person group is shown below. Figure 6 As shown.

[0030] S4. The constructed individual comfort space and group comfort space are embedded in the global cost map and local cost map of the robot navigation system in real time in the form of cost functions; The comfort space model is represented as a cost function and integrated into the robot navigation system. The core of this function lies in receiving the pedestrian's position and velocity as input in real time. These calculated values ​​are dynamically merged into global and local cost maps, serving as key inputs for navigation decisions.

[0031] S5. Generate the final navigation path based on the global cost graph and the local cost graph, and achieve socialized navigation control that balances safety and human comfort by minimizing the path cost function; The robot continuously adjusts its trajectory based on a dynamic cost map, enabling natural movement that avoids pedestrians and maintains a comfortable distance from social groups. When the status of pedestrians or the interaction status changes, the comfortable space is automatically updated, and the navigation path is replanned accordingly.

[0032] In practice, during the global planning phase, the A* algorithm is used to calculate the optimal path based on the global cost map to obtain the global planning result. During the local planning phase, the dynamic window DWA algorithm is used to adjust the robot's movement in real time and dynamically update the local cost map based on the latest individual human position and speed. If the local cost map detects a sudden approach of a human or a change in the social area, an immediate replanning mechanism is triggered to pause or detour.

[0033] To facilitate understanding, the Fetch robot will be used as a simulation and real-world example to explain the navigation method in the above embodiments in detail below.

[0034] The Fetch robot runs in a ROS environment on Ubuntu 18.04, using the MOVE_BASE package as its navigation framework. Within this framework, a cost map update frequency of 30 Hz is implemented, ensuring that the computation time for a single planning cycle can be controlled within 10-20 ms. The robot's parameters are: maximum linear velocity v = 0.5 m / s, maximum linear acceleration... =1m / Maximum angular velocity =0.5 rad / s, maximum angular acceleration =0.5m / .

[0035] First, comparative experiments were conducted in a simulation environment to verify whether the proposed method meets individual comfort needs and ensures navigation efficiency in human-perceived social navigation. Compared to the egg-shaped comfort space navigation method: Robot navigation based on a social navigation layer utilizes a mixture of two Gaussian distributions of human velocity deformation to generate a circular comfort space when stationary and an egg-shaped comfort space when the person moves. It is a widely used open-source social navigation method in the ROS community. Compared to navigation methods without comfort space: Robot navigation is entirely based on obstacle information and does not require any comfort space; it is implemented through the standard obstacle layer of the ROS cost graph. Figure 7 As shown, this method improves navigation success rate while ensuring pedestrian comfort.

[0036] In addition, in a real-world environment, assess participants' experience with the constructed comfortable space, such as... Figure 8 As shown in the figure. Finally, the effectiveness of the comfort space construction methods was evaluated using questionnaires and statistical analysis. The questionnaire first explained the definitions of human-perceived social navigation and comfort space, and then proposed criteria for selecting comfort space strategies. Only responses from participants who clearly understood the definitions were considered valid. Participants were invited to choose their preferred comfort space construction method. To ensure fair evaluation, participants were divided into two groups: one consisting of robotics researchers or professionals (HE), and the other consisting of individuals without robotics-related technical experience (NE). In addition, the questionnaire included spatial shape diagrams to help participants better understand strategy selection. A total of 241 individuals participated in the questionnaire survey, and 169 valid questionnaires were obtained, including 73 robotics researchers or professionals and 96 individuals without robotics-related experience. The survey results are as follows: Figure 9 As shown, the navigation method in the above embodiments has gained widespread acceptance from participants, thus demonstrating its rationality and feasibility. Furthermore, because robotics researchers and professionals have a higher level of understanding of robot mechanics, their agreement on the proposed method is greater, further proving that the proposed method is more conducive to social navigation of robots with human perception.

[0037] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A robot social navigation system for human-machine coexistence environments, characterized in that, The system comprises: a human recognition and speed calculation module for detecting the position, orientation and speed information of human individuals in an environment in real time through an RGB-D camera or a depth sensor; a dynamic comfortable space modeling module connected with the human recognition and speed calculation module, for constructing the comfortable space of human individuals based on a Gaussian mixture model, wherein the space in front of, behind, left of, right of, left front of and right front of a human body is modeled by six directional asymmetric two-dimensional Gaussian functions; a social group comfortable space modeling module connected with the human recognition and speed calculation module, for judging the interaction state and generating a group comfortable space according to the orientation and relative position relationship of human individuals when detecting that there is a social relationship between two or more individuals; a cost map fusion module connected with the dynamic comfortable space modeling module and the social group comfortable space modeling module respectively, for embedding the constructed individual comfortable space and group comfortable space in the form of a cost function into a global cost map and a local cost map of a robot navigation system in real time; a path planning and control module connected with the cost map fusion module, for generating a final navigation path based on the global cost map and the local cost map, and realizing socialized navigation control with consideration of safety and human comfort by minimizing a path cost function.

2. The robot social navigation system in a human-robot coexistence environment of claim 1, wherein, The Gaussian mixture model comprises a forward weighting coefficient and a backward inhibition coefficient.

3. The robot social navigation system in a human-robot coexistence environment of claim 1, wherein, The human recognition and speed calculation module is specifically configured to: collect environment information by using an RGB-D camera or a depth sensor; detect human individuals in the environment information in real time by using a deep learning-based detection network YOLO; convert the detected human key points or bounding box coordinates into positions in a world coordinate system, and calculate the speed vector and orientation angle of an individual according to the position difference between consecutive frames to obtain the position, moving direction and dynamic state of the human in the environment.

4. The robot social navigation system in a human-robot coexistence environment of claim 1, wherein, The comfortable space of the human individual comprises asymmetric two-dimensional Gaussian distributions in six directions of front, back, left, right, left front and right front, and the comfortable radius standard deviation of each direction is dynamically adjusted according to the speed of the individual.

5. The robot social navigation system in a human-robot coexistence environment of claim 1, wherein, The interaction state comprises non-communication, face-to-face communication and communication while observing the same object at different angles; and the social group comfortable space modeling module is specifically configured to: if two people have no orientation and no common intersection point, define it as non-communication, and simply fuse the comfortable spaces of the two human individuals; if the two people face each other, define it as face-to-face communication, and generate an elliptical shared comfortable space; if the two orientations are consistent, define it as communication while observing the same object at different angles, and merge the comfortable spaces into a circular region; if there are more than two people, calculate the mean position of each individual as the center of the group and construct a circular comfortable space.

6. The robotic social navigation system in a human-robot coexistence environment of claim 1, wherein, The path planning and control module is specifically configured to: in the global planning stage, calculate an optimal path based on the global cost map by using an A* algorithm to obtain a global planning result; in the local planning stage, dynamically update the local cost map according to the latest position and speed of human individuals by using a dynamic window DWA algorithm to adjust the motion of the robot in real time; if the local cost map detects that a human suddenly approaches or the social area changes, trigger an immediate re-planning mechanism to pause or detour. 7.A method for social navigation of a robot in a human-robot coexisting environment, the method comprising: The method comprises: Real-time detection of the position, orientation and speed information of human individuals in the environment through an RGB-D camera or a depth sensor; Construction of a human individual comfort space based on a Gaussian mixture model, wherein the space in front of, behind, left of, right of, left front of and right front of the human body is modeled by six directional asymmetric two-dimensional Gaussian functions; When a social relationship is detected between two or more individuals, the interaction state is determined and a group comfort space is generated according to the orientation and relative position relationship of the human individuals; The constructed individual comfort space and group comfort space are embedded in the global cost map and local cost map of the robot navigation system in the form of a cost function in real time; Generation of a final navigation path based on the global cost map and the local cost map, and implementation of socialized navigation control that takes into account safety and human comfort by minimizing the path cost function. 8.The robot social navigation in a human-robot coexisting environment method of claim 7, wherein, Real-time detection of the position, orientation and speed information of human individuals in the environment through an RGB-D camera or a depth sensor, comprising: Collecting environmental information using an RGB-D camera or a depth sensor; Real-time detection of human individuals in the environmental information through a detection network YOLO based on deep learning; The human key points or bounding box coordinates detected are converted into positions in the world coordinate system, and the speed vector and orientation angle of the individual are calculated based on the position difference between consecutive frames to obtain the position, moving direction and dynamic state of the human in the environment. 9.The robot social navigation in a human-robot coexisting environment method of claim 7, wherein, The interaction state includes non-communication, face-to-face communication and observation of the same object at different angles and communication; When a social relationship is detected between two or more individuals, the interaction state is determined and a group comfort space is generated according to the orientation and relative position relationship of the human individuals, comprising: If two people have no orientation and no common intersection point, it is defined as non-communication, and the human individual comfort spaces of the two people are simply fused; If two people face each other, it is defined as face-to-face communication, and an elliptical shared comfort space is generated; If the two orientations are consistent, it is defined as observation of the same object at different angles and communication, and the comfort space is merged into a circular region; If there are more than two people, the mean position of each individual is calculated as the center of the group and a circular comfort space is constructed. 10.The robot social navigation in a human-robot coexisting environment method of claim 7, wherein, Generation of a final navigation path based on the global cost map and the local cost map, comprising: In the global planning stage, the A* algorithm is used to calculate the optimal path based on the global cost map to obtain the global planning result; In the local planning stage, the dynamic window DWA algorithm is used to adjust the robot motion in real time, and the local cost map is dynamically updated according to the latest human individual position and speed; if the local cost map detects that a human suddenly approaches or the social area changes, an immediate re-planning mechanism is triggered, and the robot is paused or detoured.

Images