Body-equipped robot control system for hotel bed-making service

Abstract

The invention discloses a control system of a robot with a body for hotel bed-making service. The control system comprises a robot body, a behavior execution module, a sensor module and a control module, wherein the robot body comprises a movable base, a mechanical arm and an end effector; the behavior execution module is deployed on the robot body and comprises a movement control unit, a mechanical arm control unit and a tail end execution unit; the sensor module is integrated on the robot body and comprises a visual system unit, a force sense feedback unit and an environment sensing unit. The control module is deployed on a control server and comprises an operation specification library, a task planning engine and a mode switching unit; and the control module dynamically plans a bed-making action sequence through a task planning engine, and generates a function instruction for controlling the behavior execution module through the mode switching unit. According to the technical scheme, the bed-making process of the robot can be refined based on the preset bed-making quality standard, and full-flow automatic control over the bed-making process of the bed-making service robot is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of service robot control, and more specifically, to an embodied robot control system for hotel bed-making services. Background Art

[0002] Currently, the bed-making work in hotel guest rooms is mainly completed manually by housekeepers. The housekeepers need to go to each room to replace bedding such as sheets, duvet covers, pillowcases, etc. according to the hotel bed-making standards and tidy up the bed surface. The whole process is mainly completed manually, highly dependent on personnel, with a large amount of repetitive work, and it is difficult to grasp the completion quality.

[0003] With the rapid development of technologies such as artificial intelligence and the Internet of Things, the hotel industry is developing towards intelligence. Some studies have tried to introduce robot technology into the field of hotel guest room services. For example: The bed-making device disclosed in Patent CN201800179U can help people quickly make the bed, but the structure of this bed-making device is too simple, with insufficient practicality during use, and low intelligence of the device, still requiring manual operation by the user. Patent CN109397304A designs a bed-making robot including a main body shell, a control panel, and a storage bin, and the replaced sheets can be collected and stored through the storage bin, but the mechanical structure is complex, the reliability needs to be improved, there is a lack of necessary perception, and there is also a lack of detailed control logic for the robot in bed-making. In the prior art, the control system for hotel bed-making robots still has the following deficiencies: First, there is a lack of refined control ability for the entire bed-making process, and it is difficult to effectively control complex bed-making processes and detailed actions. For example, during the process of making the bed sheet, it is difficult to precisely control the force of the robotic arm waving the bed sheet and the timing of the end effector releasing the bed sheet, resulting in unstable bed-making effects; Second, there is a lack of bed-making quality detection and closed-loop optimization mechanisms, unable to effectively evaluate whether the bed-making quality meets the standards, and even less able to iteratively optimize the bed-making actions of the robot according to the detection results, resulting in difficult guarantee and continuous improvement of the bed-making quality.

[0004] Therefore, there is an urgent need for an embodied robot control system with high-quality bed-making service capabilities to solve the above problems existing in the prior art and achieve efficient, high-quality, and autonomous hotel bed-making services by the embodied robot. Summary of the Invention

[0005] To overcome the deficiencies of the prior art and achieve high-quality and autonomous hotel bed-making services, the present application provides an embodied robot control system for hotel bed-making services, including: a robot body, a behavior execution module, a sensor module, and a control module;

[0006] Wherein, the robot body includes: a mobile base, a robotic arm, and an end effector;

[0007] The behavior execution module is deployed on the robot body and includes a movement control unit, a robotic arm control unit, and an end effector unit. The behavior execution module is used to respond to the control instructions issued by the control module and drive the robot body to perform bed-making operations. Among them, the movement control unit is used to control the autonomous navigation and movement of the mobile base in the guest room environment, the robotic arm control unit is used to control the dual-arm robotic arm to perform various bed-making operations, and schedule the end effector unit to coordinate the collaborative movement between the robotic arm and the end effector to complete complex bed-making actions.

[0008] Among them, the end effector integrates a pneumatic suction cup and a flexible gripper. The pneumatic suction cup is installed at the front center of the end effector and uses the negative pressure adsorption principle to give priority to grasping flat and smooth textiles such as sheets and duvet covers. The flexible grippers are symmetrically distributed on both sides of the end effector and are suitable for grasping plane-shaped, fluffy or irregular items through multi-finger collaborative clamping.

[0009] The sensor module is integrated into the robot body and includes a vision system unit, a force feedback unit, and an environment perception unit. The sensor module collects guest room environment information, and the environment information includes the three-dimensional point cloud data of the mattress and the position and attitude information of the bedding.

[0010] The control module is deployed on the control server and includes an operation specification library, a task planning engine, and a mode switching unit. The control module, based on the environment information and the preset bed-making quality standards in the operation specification library, dynamically plans the bed-making action sequence through the task planning engine and generates function instructions for controlling the behavior execution module through the mode switching unit.

[0011] Among them, the vision system unit includes an image acquisition device and an image processing module. The image acquisition device is used to acquire the bed surface image, and the image processing module, based on the image processing algorithm, identifies the type, position, and attitude of the bedding and extracts quality defect information.

[0012] The force feedback unit is installed on the end of the robotic arm and the end effector and is used to real-time monitor the force feedback information during the bed-making operation of the robotic arm, including the swinging force of the robotic arm, the grasping force of the end effector, and the contact force, to judge whether the bedding is stably grasped, whether the folding angle force is appropriate, and the contact state between the end effector and the bed surface.

[0013] The environment perception unit is used to obtain the depth information, distance information, and obstacle information of the guest room environment and construct a dynamic map of the guest room environment.

[0014] The task planning engine adopts a reinforcement learning algorithm and a human-machine collaborative strategy, with the optimization goals of maximizing the bed-making quality score and minimizing the bed-making time, and dynamically plans to generate the bed-making action sequence, thus realizing intelligent decision-making and optimization of the bed-making task.

[0015] The bed-making action sequence is planned and generated based on the environmental information modeling results. The task planning engine performs task decoupling and functional modularization packaging on the robot bed-making process, decomposing the entire bed-making process into several basic operation behaviors, and assigning the basic operation behaviors to the behavior execution module to realize hierarchical instruction packaging;

[0016] The mode switching unit is used to convert the bed-making action sequence into functional instructions executable by the robot. The functional instructions are an ordered combination of multiple control instructions, which are used to control the coordinated actions of the various components of the robot body to complete the predetermined bed-making operation.

[0017] Furthermore, the hierarchical instruction encapsulation assigned to the behavior execution module includes: the mobile control unit is responsible for the instruction encapsulation of the robot's global displacement behavior; the robotic arm control unit is mainly responsible for the basic operation encapsulation of actions executed by the robotic arm as the main body, which does not require fine adjustment of the end; the end execution unit is used for the instruction encapsulation of high-precision operation behaviors, including refined actions of end effector tactile feedback and force-position mixed control.

[0018] Furthermore, the structure of the bed-making action sequence includes: sequence number, operation target, implementation unit, operation behavior, quality inspection standard, and function instruction; wherein the operation behavior includes: robot navigation movement, bedding grabbing, bedding unfolding and laying, bed-making quality inspection, and quality defect correction;

[0019] The content of the control instruction includes: instruction identification, execution module ID, operation component ID, target point position, operation type, control state, and quality detection parameters; wherein, the operation component ID is the component identification of each component controlled by the behavior execution module, the target point position is the coordinate value in the dynamic map of the room, and the operation type is the action that the robot component can respond to, including: unfolding the bed sheet, folding the bed sheet corners, flattening the bedspread, and placing the pillow; the control state is a supplementary parameter of the operation type, including: grasping force, moving speed, vibration amplitude frequency, bed sheet unfolding flatness threshold, bed sheet folding angle range, and time parameters; the time parameters are used to accurately control the coordinated operation of multiple components, and each component executes synchronously or asynchronously according to the specified time difference.

[0020] Among them, before generating the control instruction, the control module constructs a dynamic map of the room containing environmental obstacle information, segments and extracts the mattress point cloud; determines the key points for making the bed; according to the key points for making the bed, the pose information of the base of the embodied robot, and the working space information of the robotic arm, calculates the target point positions of each operating component in the dynamic map of the room, and constitutes the target point position parameters in the control instruction; the key points for making the bed include: the robot standing point, the geometric center of the mattress, the corner points of the mattress and the quilt, and the temporarily identified focus points; among them, the corner points of the mattress and the quilt are used to determine the operation path points of the end effector when the robot collects the bedding and makes the bed; the geometric center of the mattress is used to assist in planning the initial deployment posture of the robotic arm when the robot makes the bed.

[0021] Further, after the behavior execution module controls the robot body to execute the function instruction, the environmental perception unit of the sensor module instantaneously acquires the environmental information containing the bed surface image information within the specified time, and returns the environmental information to the control module for the task planning engine to call the quality standard for bed-making quality detection and determine whether the preset service quality requirements are met.

[0022] Among them, the operation specification library stores multiple bed-making quality standards, and the bed-making quality standards include the bed-making skill point rules and the quality detection area rules; the bed-making skill point rules include the standard actions and parameter thresholds of the bed-making operation, and the quality detection area rules include the detection area, shooting angle, feature extraction method, and determination standard of the quality detection items;

[0023] The bed-making quality detection is used to determine whether the bed-making effect meets the preset service quality requirements; if the quality detection result does not meet the service quality requirements, a temporary quality focus point is generated, triggering a bed-making quality correction operation; the temporary quality focus point contains the unqualified position information; the bed-making quality correction operation means that the task planning engine re-plans the subsequent bed-making action sequence according to the temporary quality focus point to iteratively optimize the bed-making quality.

[0024] Further, the behavior execution module is pre-integrated with a protection mechanism to cope with various potential emergencies; the protection mechanism includes: collision protection, emergency braking, video monitoring and obstacle avoidance, and anti-pinch protection; after triggering the protection mechanism, the behavior execution module stops executing the current control instruction and preferentially executes the protection mechanism action sequence corresponding to the emergency event without waiting for the control instruction from the control center.

[0025] The embodied robot control system for hotel bed-making service provided by the present invention has a quality detection and correction function. It can refine the robot to implement the bed-making process based on the preset bed-making quality standard, and control the action elements and task planning through multi-sensor information fusion and reinforcement learning algorithm, so as to realize the full-process autonomous control and high-quality operation of the bed-making process of the bed-making service embodied robot. In the present invention, multi-sensor fusion (RGB-D, ToF, force sense) is used to realize real-time modeling of different beddings and millimeter-level operation accuracy, enabling the robot to adapt to the bed-making requirements of different room types; a special end effector is adopted, integrating a flexible gripper and a suction cup, and dynamically adjusting the pressure through an adaptive grasping algorithm to realize fine operation of different linen materials; and a vision-based bed-making quality detection and evaluation algorithm is used to quantify the bed-making effect by extracting image features and rework the unqualified areas in time; at the same time, to ensure the safety of the robot itself, multiple safety protection measures are adopted to maximize the safety of the service process, effectively solving the problems of low intelligence of the bed-making robot control system and difficult to guarantee the bed-making quality in the prior art, realizing the high-quality, automated and intelligent operation of the hotel bed-making service robot, and having broad application prospects and important application values. Brief Description of the Drawings

[0026] Figure 1 It is a schematic diagram of the system architecture of the embodied robot control system for hotel bed-making service provided by an embodiment of the present invention. Detailed Embodiments

[0027] The following will describe in detail the specific embodiments of the embodied robot control system for hotel bed-making service provided by the present invention with reference to the accompanying drawings of the specification.

[0028] As Figure 1 shown, the embodied robot control system for hotel bed-making service provided by the present invention mainly consists of four parts: a robot body P100, a sensor module P110, a behavior execution module P120, and a control module P130. Among them, the robot body P100 is the execution carrier of the technical solution of the present invention and is an embodied robot for realizing hotel guest room bed-making operations. Its main components include: a mobile base P101, a robotic arm P102, and an end effector P103.

[0029] Mobile base P101: As the mobile platform of the robot body, it adopts an omnidirectional mobile chassis to achieve the all-round, flexible and autonomous movement of the robot in the hotel guest room environment. The omnidirectional mobile chassis can adopt structures such as Mecanum wheels or omnidirectional wheels to ensure that the robot can flexibly turn, translate and rotate in narrow spaces. To achieve autonomous navigation and positioning, the mobile base P101 integrates various sensors such as lidar, depth camera, and inertial measurement unit (IMU), and is configured with SLAM (Simultaneous Localization and Mapping) algorithm to construct a map of the guest room environment and achieve autonomous positioning and navigation of the robot. In addition, considering the high diversity of the heights of hotel guest room beds, the mobile base P101 is also equipped with an electric lifting mechanism with a lifting stroke range of 0 - 60 cm and a load capacity of not less than 150 kg to meet the needs of making beds for beds of different heights.

[0030] Manipulator P102: As the main operation execution mechanism of the robot body, it adopts a dual-arm collaborative robot. Each manipulator has 7 degrees of freedom and a repeat positioning accuracy of up to ±0.02 mm to ensure that the robot can flexibly and precisely complete complex actions such as grasping, unfolding, and laying bedding. The manipulator P102 adopts a lightweight design with a material of high-strength aluminum alloy to reduce the overall weight of the robot and improve the movement flexibility and response speed. To achieve safe and reliable human-robot collaboration, the manipulator P102 integrates a collision detection sensor and a torque sensor, which can real-time sense external collisions and force feedback, and stop moving in time when a collision or overload occurs to protect the safety of personnel and equipment.

[0031] End effector P103: Installed at the end of the manipulator P102, it is the execution component for the robot to directly interact with operation objects such as bedding. To meet the needs of different beddings (such as sheets, duvet covers, pillowcases, etc.) and different operations (such as grasping, unfolding, folding, tidying, etc.) during hotel bed making, the end effector P103 integrates two execution mechanisms: a pneumatic suction cup and a flexible gripper. Among them, the pneumatic suction cup adopts the vacuum negative pressure adsorption principle with a negative pressure value of -80 kPa, and is preferably used to grasp flat and large-area textiles such as sheets and duvet covers; the flexible gripper adopts a multi-finger structure design imitating the human hand, with an opening and closing range of 50 - 300 mm, and can dexterously grasp pillows, crumpled bedding, obstacles and other flat-shaped, fluffy or irregular items. Through the integration of the pneumatic suction cup and the flexible gripper, as well as the intelligent switching and coordination between the two, the versatility and operation flexibility of the robot end effector can be significantly improved, enabling it to be competent for various complex operation tasks of hotel guest room bed making services.

[0032] Sensor module P110: As the perception core of the robot control system, it is integrated on the robot body P100 and consists of a vision system unit P111, a force feedback unit P112, and an environment perception unit P113. It is used to collect the guest room environment information in real time and comprehensively, providing the robot with environment perception capabilities. The environment information collected by the sensor module P110 is an important basis for the control module P130 to perform task planning, path planning, quality inspection, and safety protection.

[0033] Vision system unit P111: It includes an image acquisition device and an image processing module; the image acquisition device is used to collect the bed surface image, and the image processing module is configured to identify the types, positions, and postures of beddings such as sheets, duvet covers, and pillowcases based on image processing algorithms, and extract quality defect information such as bed surface wrinkles and uneven areas. The image acquisition device consists of devices of a high-precision machine vision system, including an RGB camera (e.g., resolution 3840×2160), a depth camera; the image processing module includes an image acquisition card and an image processing unit, etc. The vision system unit P111 is mainly used to obtain the image information of the guest room bed surface and beddings, and realize functions such as bedding type identification, position and posture detection, and quality defect detection. The vision system unit P111 controls the camera to collect images of the bed surface at preset shooting points and angles according to the instructions issued by the control module P130. The image processing unit integrates advanced image processing and pattern recognition algorithms (e.g., object detection, image segmentation, and feature extraction algorithms based on deep learning), and can preprocess, extract features, and analyze the collected images, identify the materials, patterns, and shape features of different fabrics, accurately identify the types of beddings such as sheets, duvet covers, and pillowcases; detect quality defects such as wrinkles, stains, and damages of the beddings, and perform positioning; cooperate with the control module P130 to perform 3D reconstruction of the bed surface and midline recognition, determine important positions such as key points for making the bed, identify the target types, and provide data support for subsequent bed-making action planning and quality inspection.

[0034] Force feedback unit P112: It consists of a high-precision six-axis force sensor (e.g., accuracy reaching ±0.1N) and a force feedback control system. The force feedback unit P112 is installed on the end of the robotic arm P102 and the end effector P103, and is used to monitor the force feedback information of the robotic arm during the bed-making operation in real time, including the swinging force of the robotic arm, the grasping force of the end effector, and the contact force, etc. The force feedback information collected by the force feedback unit P112 can be used to judge whether the beddings are stably grasped, whether the corner-folding force is appropriate, and the contact state between the end effector and the bed surface, etc. The task planning engine P132 of the control module P130 can dynamically adjust the motion trajectory of the robotic arm and the operation parameters of the end effector according to the force feedback information, realize force-position hybrid control, improve the stability and reliability of the bed-making operation, and prevent damage to the beddings due to excessive or insufficient force.

[0035] Environmental perception unit P113: It is composed of the fusion of multiple sensors such as lidar (for example, the scanning frequency is 20Hz), depth camera, ultrasonic sensor, and infrared sensor. The environmental perception unit P113 is mainly used to obtain depth information, distance information, obstacle information, etc. of the guest room environment, and construct a dynamic map of the guest room environment. The lidar and depth camera are used to obtain the three-dimensional point cloud data of the guest room environment, and construct a dynamic map of the room and a point cloud model of the mattress. The ultrasonic sensor and infrared sensor are used to detect obstacles around the robot, such as furniture, walls, pedestrians, etc., and provide the ability to perceive close-range obstacles. The environmental information collected by the environmental perception unit P113 provides a basis for navigation and obstacle avoidance for the mobile control unit P121, enabling the robot to autonomously navigate and move safely in a complex and dynamic guest room environment.

[0036] Behavior execution module P120: As the execution mechanism of the robot body P100, it is deployed on the robot body and consists of a mobile control unit P121, a robotic arm control unit P122, and an end effector unit P123. It is used to respond to the control instructions issued by the control module P130 and drive the robot body to perform the bed-making operation. The behavior execution module adopts a hierarchical control architecture, and each unit works together to complete complex bed-making tasks.

[0037] Mobile control unit P121: Responsible for controlling the autonomous navigation and movement of the mobile base P101 in the guest room environment. It integrates motion planning algorithms and navigation control algorithms inside, and can plan the optimal movement path according to the navigation instructions issued by the control module P130, and control the mobile base to move smoothly and precisely to the preset position beside the bed. The mobile control unit P121 also has the functions of environmental perception and obstacle avoidance, and can adjust the movement path in real time according to the environmental information provided by the sensor module P110, and flexibly avoid obstacles in the guest room to ensure the safety and reliability of the robot movement.

[0038] Robotic arm control unit P122: Responsible for controlling the dual-arm robotic arm P102 to perform various bed-making operations, including bed linen grasping, unfolding, laying, etc. It integrates kinematics and dynamics control algorithms, trajectory planning algorithms, and force control algorithms inside, and can accurately control the joint movement, end posture, and operating force of the robotic arm according to the robotic arm control instructions issued by the control module P130. The robotic arm control unit P122 also has the ability to schedule the end effector unit P123, and can coordinate the collaborative movement between the robotic arm and the end effector to complete complex bed-making actions.

[0039] End effector unit P123: Responsible for controlling the end effector P103 to perform delicate bed-making operations, such as sheet corner folding, edge finishing, and bed linen smoothing. It integrates a force / position hybrid control algorithm and a fine operation control algorithm internally, and can accurately control the position, posture, and clamping force of the end effector according to the end effector control instructions issued by the control module P130, so as to achieve delicate operations on the bed linen. The end effector unit P123 is also responsible for the switching and coordinated control of the pneumatic suction cup and the flexible gripper to meet the requirements of different bed linen and operations.

[0040] Control module P130: It is the core component of the control system of the present invention, deployed on the control server, and serves as the control center of the bed-making service embodied robot. The control module P130 is composed of an operation specification library P131, a task planning engine P132, and a mode switching unit P133. It is used to dynamically plan the bed-making action sequence through the task planning engine P132 according to the environmental information provided by the sensor module P110 and based on the preset bed-making quality standards in the operation specification library P131, and generate function instructions for controlling the behavior execution module P120 through the mode switching unit P133, so as to achieve autonomous decision-making, intelligent control, and quality assurance in the robot bed-making process. The control module P130 adopts a modular and distributed software architecture, and is developed and deployed based on the ROS (Robot Operating System) operating system, with good scalability and maintainability.

[0041] Operation specification library P131: As the core component of the control module P130, it is used to store a number of preset bed-making quality standards. The bed-making quality standards in the operation specification library P131 mainly include two types of rules: bed-making skill point rules and quality inspection area rules. The bed-making skill point rules define the standard actions, operation sequences, and parameter thresholds (such as the flatness threshold for sheet unfolding, the corner folding angle range, etc.) of the bed-making operation, and are the action guidance and evaluation basis for the robot to perform the bed-making operation. The quality inspection area rules define the inspection areas, shooting angles, feature extraction methods, and judgment criteria for different quality inspection items (such as sheet flatness, bed linen position, corner folding effect, etc.), and are used to guide the robot to perform bed-making quality inspection and evaluation. The operation specification library P131 adopts a modular design and supports flexible configuration and expansion according to different hotel bed-making standards and guest room types.

[0042] For example: The content regarding the sheet flatness defined in the quality inspection area rules includes:

[0043] Shooting point 1: 45 degrees outside the bed tail, at the same height as the bed surface, 30 cm away from the bedside;

[0044] Shooting point 2: At the 2 / 3 position of the bed tail, directly above, 50 cm away from the bed surface;

[0045] Extract features: edge segment curvature, length, and inclination angle;

[0046] Judgment rules: curvature radius>1m, length deviation<2cm, inclination angle<5°;

[0047] For example, the regional rule for the symmetry of bedding placement defined in the quality inspection regional rule can be defined as:

[0048] Shooting point: directly in front of the bed, at the same height as the pillow, 50cm away from the bed;

[0049] Extract features: center coordinates and angles of the pillow rectangle;

[0050] Judgment rules: The angle between the center line and the bedside is 90°±2°, and the two centers are equidistant from the bed midline by ±5cm.

[0051] The expression of regional rules can be agreed upon through protocol, such as:

[0052] {

[0053] "Test Item":"Bed Sheet Centerline Alignment",

[0054] "Detection target":"bed",

[0055] "Range Type":"Bed Surface",

[0056] "Detection Dimension":"Oddness",

[0057] "Shooting point coordinates": (x, y, z), / / Based on the mattress coordinate system

[0058] "Camera orientation": (α, β, γ), / / Euler angles define pitch / yaw / rotation

[0059] "shooting parameters":{

[0060] "Resolution":"3840×2160",

[0061] "Light compensation": "Automatic",

[0062] "Depth of field range":"0.5-1.2m"

[0063] },

[0064] "Feature point definition":[

[0065] {"Name":"Starting point of center line","Coordinate range":(x1±Δx,y1±Δy,z1±Δz)},

[0066] {"Name":"Midline end point","Coordinate range":(x2±Δx,y2±Δy,z2±Δz)}

[0068] }

[0069] Task planning engine P132: As the decision-making center of the control module P130, it is responsible for dynamically planning and generating a bed-making action sequence according to the environmental information provided by the sensor module P110 and the bed-making standards defined in the operation specification library P131. The task planning engine P132 adopts a reinforcement learning algorithm (e.g., PPO algorithm) and a human-machine collaboration strategy, with the optimization goal of maximizing the bed-making quality score and minimizing the bed-making time, dynamically planning and generating the bed-making action sequence to achieve intelligent decision-making and optimization of the bed-making task.

[0070] The bed-making action sequence is planned and generated based on the modeling results of the guest room environment. Its essence is to decouple the tasks and modularize the functions of the robot bed-making process. The whole bed-making process is decomposed into several basic operation behaviors (e.g., robot movement, sheet grasping, sheet unfolding, sheet corner folding, etc.), and according to the functional characteristics and actuator differences of each operation behavior, they are assigned to different units of the behavior execution module P120 (mobile control unit P121, robotic arm control unit P122, end effector unit P123) to achieve hierarchical instruction encapsulation.

[0071] Among them: The mobile control unit is mainly responsible for the instruction encapsulation of the robot's global displacement behavior, such as the path planning and motion control for the robot to navigate from the initial position to the periphery of the bed; the robotic arm control unit is mainly responsible for the encapsulation of basic operations performed by the robotic arm as the main body, including actions such as sheet grasping and planar unfolding that do not require fine adjustment at the end, and its function encapsulation ignores the internal control details of the end effector; while the end effector unit is dedicated to the instruction encapsulation of high-precision operation behaviors, such as fine actions that rely on the tactile feedback and force-position hybrid control of the end effector, such as sheet corner forming and duvet edge alignment. In short, through the above hierarchical encapsulation mechanism, the composite operation behaviors are decoupled into standardized functional modules at different levels, effectively reducing the complexity of multi-mechanism collaborative control, while ensuring the reusability of the action sequence and the scalability of the system.

[0072] The structure of the bed-making action sequence includes: serial number, operation target, implementation unit, operation behavior, quality inspection standard, function instruction; the operation behavior at least includes: robot navigation and movement, bed product grasping, bed product unfolding and laying, bed-making quality inspection, quality defect correction.

[0073] When the robot completes a bed-making task, it needs to perform service contents such as changing the bedsheet, changing the duvet cover, and changing the pillowcase. Different bed-making action sequences are preset for each service content corresponding to different operation behaviors. The operation behaviors specifically include: robot movement, grasping target, transferring target, controlling target, and detailed control target, etc. For example, when the service content is changing the bedsheet, a partial action sequence is as shown in the following table:​

[0074]

[0075]

[0076] Among them, the function instructions corresponding to each specific operation behavior are composed of the operation instructions of multiple control operation components, and the operation instructions are generated by the mode switching unit P133.

[0077] Mode switching unit P133: As the instruction generation center of the control module P130, it is used to convert the bed-making action sequence generated by the task planning engine P132 into function instructions executable by the robot. The function instructions are an ordered combination composed of multiple control instructions, which are used to control the coordinated actions of the various components of the robot body P100 (mobile base P101, robotic arm P102, end effector P103) to complete the predetermined bed-making operation. The mode switching unit P133 converts each operation behavior in the bed-making action sequence into a corresponding function instruction according to the preset instruction template and protocol, and optimizes and encapsulates the function instructions. For example, multiple control instructions are combined into a macro instruction or an instruction set, and then sent to the behavior execution module P120 or a specified unit of the behavior execution module in a group to improve the efficiency and real-time performance of instruction execution.

[0078] The content of the control instruction includes: instruction identifier, execution module ID, operation component ID, target point position, operation type, control status, quality inspection parameter; among them, the operation component ID is the component identifier of each component controlled by the behavior execution module, the target point position is the coordinate value in the room dynamic map, the operation type is the action that the robot component can respond to, and at least includes: sheet unfolding, sheet corner folding, bedspread flattening, pillow placement; the control status is a supplementary parameter of the operation type, including: grasping force, moving speed, vibration amplitude frequency, sheet unfolding flatness threshold, sheet corner folding angle range, time parameter; the time parameter is used to precisely control the synchronous or asynchronous execution of each component at a specified time difference when multiple components cooperate.

[0079] Before generating control instructions, the control module constructs a dynamic map of the room containing environmental obstacle information based on the RGB-D sensor data fusion algorithm, and segments and extracts the mattress point cloud from the dynamic map of the room based on the point cloud segmentation algorithm of deep learning; and based on the mattress point cloud, determines the key bed-making points including at least the corners of the mattress core and the geometric center of the mattress through the point cloud edge detection and corner extraction algorithms; further, according to the positions of the key bed-making points, as well as the pose information of the base of the bed-making service embodied robot and the working space information of the robotic arm, before performing the bed-making operation behavior, calculates the target point positions of each operating component in the dynamic map of the room, and constitutes the target point position parameters in the control instructions; the key bed-making points at least include: the robot standing point, the geometric center of the mattress, the corners of the mattress core, and the temporarily identified focus points; wherein, the corners of the mattress core are used to determine the operation path points of the end effector when the robot collects bedding and makes the bed; the geometric center of the mattress is used to assist in planning the initial deployment posture of the robotic arm when the robot makes the bed.

[0080] After generating the function instructions, the control module controls the sensor module to collect the environmental information of the current bed-making state, and calls the quality detection area rules in the operation specification library to perform image acquisition and feature extraction on the preset area, and compares the extracted feature parameters with the parameter thresholds defined in the bed-making skill point rules to perform bed-making quality detection and judge whether the bed-making effect meets the preset service quality requirements; if the quality detection result is that the service quality requirements are not met, generates a temporary quality focus point containing the unqualified position information, and triggers a bed-making quality correction operation; the bed-making quality correction operation refers to that the task planning engine calls the corresponding quality correction actions from the operation specification library according to the temporary quality focus point, and re-plans the subsequent bed-making action sequence to iteratively optimize the bed-making quality.

[0081] The complete working process of the hotel bed-making service embodied robot control system mainly includes the following steps:

[0082] 1) Environmental perception and modeling: The robot moves into the guest room, and the environmental perception unit P113 of the sensor module P110 (for example, lidar, depth camera) scans the guest room environment, constructs a dynamic map of the room, and identifies the position of the mattress to generate a mattress point cloud model;

[0083] 2) Detection of key bed-making points: The task planning engine P132 of the control module P130 determines the key bed-making points, such as the geometric center of the mattress, the corners of the mattress core, etc., based on the mattress point cloud model through point cloud processing algorithms (for example, point cloud edge detection, corner extraction);

[0084] 3) Bed-making action sequence planning: The task planning engine P132 dynamically plans and generates a bed-making action sequence according to the guest room environment map, the positions of the key points for bed-making, and the bed-making standards defined in the operation specification library P131; the bed-making task is decomposed into a series of ordered operation behaviors (such as robot movement, sheet grasping, sheet unfolding, sheet corner folding, etc.) through the bed-making action sequence;

[0085] 4) Function instruction generation and distribution: The mode switching unit P133 converts the bed-making action sequence into corresponding function instructions and distributes the function instructions to the corresponding units (movement control unit P121, robotic arm control unit P122, end effector unit P123) of the behavior execution module P120;

[0086] 5) Robot action execution: Each unit of the behavior execution module P120 parses and executes the function instructions, driving the components of the robot body P100 to cooperate and complete the bed-making operation;

[0087] After the behavior execution module controls the operation components of the robot to execute the function instructions, the environmental perception unit of the sensor module instantaneously obtains environmental information including the bed surface image information within 200 milliseconds and returns the environmental information to the task planning engine, which is used by the task planning engine to call quality standards such as the preset sheet flatness standard, sheet corner folding angle standard, and bedspread edge alignment standard in the operation specification library for bed-making quality detection, and determines whether the execution of the function instructions meets the preset service quality requirements according to the quality detection results.

[0088] 6) Bed-making quality detection and evaluation: After the robot completes each bed-making action, the vision system unit P111 of the sensor module P110 collects the bed surface image, and the task planning engine P132 calls the quality detection area rules and bed-making skill point rules in the operation specification library P131 to perform quality detection and evaluation on the bed-making effect;

[0089] 7) Quality correction and iterative optimization: If the quality detection result does not meet the preset service quality requirements, the task planning engine P132 will generate temporary quality focus points and trigger a bed-making quality correction operation, re-planning the subsequent bed-making action sequence to iteratively optimize the bed-making quality until the quality requirements are met;

[0090] 8) Completion of the bed-making task: When all bed-making actions are executed and the quality detection result meets the service quality requirements, the robot completes the current bed-making task and returns to the standby state or executes the next task.

[0091] The present invention provides specific embodiments to detail how the control module P130 generates function instructions to control the robot to achieve bed-making actions: In this embodiment, the robot is controlled to perform the "center line alignment" operation during the sheet unfolding process.

[0092] The goal of the "midline alignment" operation is to align the center line of the flat and unfolded bedsheet with the center line of the mattress, with the deviation controlled within 1 cm to meet the requirements of high-quality bed making. The specific steps and the process of generating control instructions for this operation are as follows:

[0093] 1) Obtain input information: The task planning engine P132 first obtains the mattress point cloud data from the environmental perception module P110, and calculates the geometric center coordinates of the mattress based on the point cloud data (assumed to be

[0094] [X_bed_center, Y_bed_center, Z_bed_center]); at the same time, obtain the pose information of the current end effector of the dual-arm manipulator P102 (for example, the center coordinates of the end effector of manipulator 1 [X_arm1, Y_arm1, Z_arm1], and the center coordinates of the end effector of manipulator 2 [X_arm2, Y_arm2, Z_arm2]);

[0095] 2) Calculate the target point position: The task planning engine P132 calculates the target point positions of the dual-arm manipulator P102 during the "midline alignment" operation according to the mattress geometric center coordinates and the current pose information of the manipulator, combined with the preset bed-making operation parameters (such as the unfolding height of the bedsheet, vibration amplitude, vibration frequency, etc.). The calculation process includes coordinate system transformation, geometric transformation, inverse kinematics solution, etc. Assume that the calculated target point position of manipulator 1 is [X_target1, Y_target1, Z_target1], and the target point position of manipulator 2 is [X_target2, Y_target2, Z_target2];

[0096] 3) Plan and generate the bed-making action sequence; The task planning engine P132 plans and generates the bed-making action sequence according to the target point positions and the current pose information of the manipulator. The bed-making action sequence includes action sequences such as robot movement, standard process - start bed making, standard process - adsorb target, standard process - midline alignment, etc.

[0097] 4) Generate function instructions: The mode switching unit P133 decomposes the operations of the bed-making action sequence into a series of control instructions, and generates corresponding function instructions according to the preset instruction templates and protocols. The function instructions contain multiple control instructions for controlling the coordinated actions of the manipulator P102 and the end effector P103. For example, the control instructions of "standard process - midline alignment" complete the operations of bedsheet unfolding and midline alignment. An example of the generated function instructions is shown in the following table:

[0098]

[0099]

[0100] After each component of the robot executes the functional instruction of the standard process - center line alignment in sequence, the vision system unit acquires the image data of the bed and the bedsheet and returns it to the task planning engine to judge the completion of the unfolding of the bedsheet.

[0101] 5) Instruction issuance and execution: The mode switching unit P133 issues the generated functional instruction to the robotic arm control unit P122 and the end effector unit P123 of the behavior execution module P120. Each unit of the behavior execution module parses and executes the control instruction, driving the robotic arm P102 and the end effector P103 to act in coordination to complete the operations of unfolding the bedsheet and aligning the center line.

[0102] 6) Quality inspection and feedback: After the operation is completed, the vision system unit P111 acquires the image data of the bed surface and returns the image data to the task planning engine P132. The task planning engine P132 calls the quality inspection area rules in the operation specification library P131 to inspect the alignment degree of the center line of the bedsheet and judge whether it meets the preset quality standard (for example, the center line deviation ≤ 1 cm). If the inspection result does not meet the standard, the task planning engine P132 will generate a temporary quality concern point and trigger the operation of correcting the bed-making quality, such as readjusting the posture of the robotic arm and the vibration parameters and executing the "center line alignment" operation again until the quality meets the standard.

[0103] To ensure the safe and reliable operation of the bed-making service embodied robot in complex environments such as hotel guest rooms, the present invention provides a protection mechanism: A multiple safety protection mechanism is integrated in the behavior execution module P120, including safety measures at the hardware and software levels, to cope with various potential emergency events and maximize the protection of the safety of the robot itself, the operator, and the guest room environment. When the sensor module detects emergency events such as collision, motor overload, abnormal visual monitoring screen, or the clamping force of the end effector exceeding the safety threshold, the behavior execution module immediately triggers the protection mechanism: stops executing the current control instruction and preferentially executes the action sequence corresponding to the emergency event without waiting for the control instruction from the control center to ensure the safety of the robot and the environment and improve the safety and reliability of the robot operation.

[0104] The protection mechanism mainly includes the following aspects:

[0105] 1) Collision protection: Highly sensitive contact stop sensors are assembled at key parts of the robotic arm P102 (such as joints, connecting rods, end effectors, etc.). When the robotic arm is in motion and the sensor detects that the external collision force exceeds a preset threshold (e.g., 20 N), the behavior execution module P120 immediately stops the motion of the robotic arm and sends an alarm signal to the control module P130. After receiving the alarm signal, the control module P130 will enter the safety protection mode and prompt the operator to check and handle, thus effectively avoiding damage to the robot or environmental damage caused by collisions and ensuring the safety of the robot and the environment.

[0106] 2) Emergency braking: An emergency braking switch (such as a red button or an emergency stop pull cord) is set at a conspicuous position on the robot body P100. When an emergency occurs or it is necessary to immediately stop the operation of the robot, the operator can press the emergency braking switch to cut off the power supply of the robot with one key, causing the robot to immediately stop all actions. As the last line of safety defense, the emergency braking switch can effectively respond to emergencies, prevent accidents from occurring, and ensure the safety of personnel and equipment.

[0107] 3) Visual monitoring and obstacle avoidance: Through a high-definition camera (such as an RGB-D camera) installed on the robot body P100, real-time visual monitoring of the robot's working area is achieved. The visual detection algorithm of the control module P130 analyzes the image data collected by the camera in real time to identify obstacles (such as pedestrians, furniture, sundries, etc.) in the working area. When an obstacle is detected on the moving path of the robot, the movement control unit P121 will automatically trigger the obstacle avoidance mechanism, re-plan a safe and collision-free moving path, and guide the robot to bypass the obstacle to ensure that the robot safely and smoothly completes the bed-making task. During the obstacle avoidance process, if the obstacle cannot be bypassed, the robot will stop moving and alarm, waiting for manual intervention.

[0108] 4) Anti-pinch protection: Non-contact safety protection devices such as safety light curtains or infrared sensors are set around the moving area of the robotic arm P102. When the safety light curtain or infrared sensor detects a human body (such as the operator's arm or body) entering the dangerous moving area of the robotic arm, the behavior execution module P120 will immediately trigger the anti-pinch protection mechanism to control the robotic arm to decelerate to a safe speed or immediately stop moving. The anti-pinch protection mechanism can effectively prevent the robotic arm from pinching people during movement and ensure personal safety. In addition, the flexible gripper of the end effector P103 is also designed with a force sensor and force control function, which can monitor the clamping force in real time to avoid damaging the bedding due to excessive clamping force.

[0109] The embodied robot control system for hotel bed-making service provided by the present invention realizes the full-process automation and intelligence of hotel guest room bed-making service through innovative technologies such as multi-sensor fusion perception, intelligent task planning, refined motion control, quality detection and correction, etc., and has the following remarkable advantages and expected effects:

[0110] First, significantly improve the automation level and greatly save labor costs: The control system provided by the present invention enables the embodied robot for bed-making service to autonomously and efficiently complete the bed-making tasks in hotel guest rooms without manual intervention, significantly reducing the hotel's dependence on labor and greatly saving labor costs. As expected, one robot can complete the bed-making tasks for 30-50 guest rooms per day, and the bed-making efficiency is 3-5 times that of manual labor. It can work 24 hours a day, effectively alleviating the hotel's employment pressure and reducing operating costs.

[0111] Second, improve the bed-making efficiency and productivity and shorten the guest room cleaning cycle: The robot has a fast bed-making speed and high efficiency, which can significantly shorten the guest room cleaning cycle, increase the guest room turnover rate, and improve the hotel's operating efficiency and productivity. The guest room attendants can invest more time in other more valuable service work, such as in-depth guest room cleaning, personalized services, etc., to improve the overall service quality.

[0112] Third, ensure the bed-making quality and service consistency and improve customer satisfaction: The control system provided by the present invention, based on the preset bed-making quality standards and quality detection mechanism, can ensure that each bed-making operation of the robot reaches a unified high-quality standard, avoiding the randomness and uncertainty of manual operations, and improving the stability and consistency of the bed-making quality. High-quality and standardized bed-making services can enhance the customer's check-in experience and improve customer satisfaction and loyalty.

[0113] Fourth, flexibly adapt to different room types and beddings, with strong robustness: The present invention adopts multi-sensor real-time modeling technology and compliant control technology, enabling the robot to flexibly adapt to different room types and beddings of different sizes and heights without the need to precisely model and calibrate the guest room environment in advance, and having strong environmental adaptability and robustness. Even in an unstructured and dynamically changing guest room environment, the robot can stably and reliably complete the bed-making tasks.

[0114] Fifth, realize the closed-loop control and continuous optimization of the bed-making quality: The control system provided by the present invention integrates vision detection and quality assessment algorithms, can quantitatively evaluate the bed-making effect, timely detect non-compliant areas, and perform closed-loop improvement through quality correction operations to achieve the continuous optimization of the bed-making quality. By continuously accumulating bed-making operation data and quality detection results, the robot control strategy can be continuously trained and optimized to continuously improve the bed-making efficiency and quality.

[0115] Sixth, multiple safety protection mechanisms to ensure human-robot safety: At the level of the robot body and control system, the present invention integrates multiple safety protection mechanisms such as collision protection, emergency braking, video monitoring, and anti-pinch protection, minimizing usage risks and ensuring the safety of the robot itself, operators, and guests. Combined with functions such as ultraviolet disinfection, it creates a safer and more hygienic living environment for guests.

[0116] Seventh, human-robot collaboration to improve service quality: The present invention advocates a service model of human-robot collaboration, liberating the waiter from heavy physical labor and transforming them into a role of assisting supervision and quality control, focusing on handling complex tasks and personalized services that robots are unable to handle, giving full play to the advantages of humans in dealing with complex tasks and making up for the deficiencies of robots, jointly improving the quality and efficiency of hotel room service.

[0117] In summary, the embodied robot control system for hotel bed-making service provided by the present invention fundamentally revolutionizes the traditional hotel bed-making mode. With robot technology as the core, it integrates cutting-edge technologies such as multi-sensor perception, artificial intelligence, and human-robot collaboration, forming a complete, efficient, and intelligent solution. This will greatly improve the automation level, service quality, and operation efficiency of hotel room service, bringing a brand-new check-in experience to guests. More importantly, the application of the hotel bed-making service robot provides a successful example for human-robot collaboration to complete complex living labor tasks, having important demonstration significance and promotion value for promoting the intelligent transformation of the service industry.

[0118] The above discloses only several specific embodiments of the present invention. However, the present invention is not limited thereto, and any changes that can be conceived by those skilled in the art shall fall within the protection scope of the present invention.

Claims

1. An embodied robot control system for hotel bed-making service, characterized in that, Including: A robot body, a behavior execution module, a sensor module, and a control module; Among them, the robot body includes: a mobile base, a robotic arm, and an end effector; The behavior execution module is deployed on the robot body and includes a movement control unit, a robotic arm control unit, and an end execution unit; the behavior execution module is used to respond to the control instructions issued by the control module and drive the robot body to perform the bed-making operation; among them, the movement control unit is used to control the autonomous navigation and movement of the mobile base in the guest room environment, the robotic arm control unit is used to control the dual-arm robotic arm to perform various bed-making operations, and schedule the end execution unit to coordinate the cooperative movement between the robotic arm and the end effector to complete complex bed-making actions; The sensor module is integrated into the robot body and includes a vision system unit, a force feedback unit, and an environment perception unit. The sensor module collects guest room environment information, and the environment information includes the three-dimensional point cloud data of the mattress and the position and attitude information of the bedding; The control module is deployed on a control server and includes an operation specification library, a task planning engine, and a mode switching unit. The control module, based on the environment information and the preset bed-making quality standard in the operation specification library, dynamically plans the bed-making action sequence through the task planning engine, and generates a function instruction for controlling the behavior execution module through the mode switching unit.

2. The embodied robot control system according to claim 1, wherein The vision system unit includes an image acquisition device and an image processing module. The image acquisition device is used to acquire the bed surface image, and the image processing module, based on the image processing algorithm, identifies the type, position, and attitude of the bedding and extracts quality defect information; The force feedback unit is installed on the end of the robotic arm and the end effector, and is used to monitor the force feedback information of the robotic arm during the bed-making operation in real time, including the swinging force of the robotic arm, the grasping force of the end effector, and the contact force, and judge whether the bedding is stably grasped, whether the corner folding force is appropriate, and the contact state between the end effector and the bed surface; The environment perception unit is used to obtain the depth information, distance information, and obstacle information of the guest room environment and construct a dynamic map of the guest room environment.

3. The embodied robot control system according to claim 1, characterized in that, The task planning engine adopts a reinforcement learning algorithm and a human-machine collaboration strategy, and takes maximizing the bed-making quality score and minimizing the bed-making time as the optimization goal, dynamically plans and generates the bed-making action sequence, and realizes the intelligent decision-making and optimization of the bed-making task; The bed-making action sequence is planned and generated based on the modeling result of the environment information. The task planning engine decouples the tasks and modularizes the functions of the robot bed-making process, decomposes the whole bed-making process into several basic operation behaviors, and assigns the basic operation behaviors to the behavior execution module to achieve hierarchical instruction encapsulation; The mode switching unit is used to convert the bed-making action sequence into a function instruction executable by the robot. The function instruction is an ordered combination composed of multiple control instructions, and is used to control the coordinated actions of each component of the robot body to complete the predetermined bed-making operation.

4. The embodied robot control system according to claim 3, wherein The implementation of hierarchical instruction encapsulation by the assigned behavior execution module includes: The movement control unit is responsible for the instruction encapsulation of the robot's global displacement behavior; the robotic arm control unit is mainly responsible for the basic operation encapsulation of actions that are mainly executed by the robotic arm and do not require fine adjustment at the end; the end effector unit is used for the instruction encapsulation of high-precision operation behaviors, including the refined actions of tactile feedback and force-position hybrid control of the end effector.

5. The embodied robot control system according to claim 1, characterized in that, The structure of the bed-making action sequence includes: serial number, operation target, implementation unit, operation behavior, quality inspection standard, functional instruction; among them, the operation behaviors include: robot navigation and movement, bed product grasping, bed product unfolding and laying, bed-making quality inspection, and quality defect correction; The content of the control instruction includes: instruction identifier, execution module ID, operation component ID, target point position, operation type, control status, quality inspection parameters; among them, the operation component ID is the component identifier of each component controlled by the behavior execution module, the target point position is the coordinate value in the dynamic room map, and the operation type is the action that the robot components can respond to, including: sheet unfolding, sheet corner folding, bedspread flattening, pillow placement; the control status is a supplementary parameter of the operation type, including: grasping force, moving speed, vibration amplitude and frequency, sheet unfolding flatness threshold, sheet corner folding angle range, time parameter; the time parameter is used to precisely control that when multiple components cooperate in operation, each component executes synchronously or asynchronously according to the specified time difference.

6. The embodied robot control system according to claim 1, wherein, Before generating the control instruction, the control module constructs a dynamic room map containing environmental obstacle information, segments and extracts the mattress point cloud; determines the key bed-making points; according to the key bed-making points, the pose information of the embodied robot base, and the workspace information of the robotic arm, calculates the target point positions of each operation component in the dynamic room map, and constitutes the target point position parameter in the control instruction; The key bed-making points include: robot standing point, mattress geometric center, mattress and quilt corner points, and temporarily identified focus points; among them, the mattress and quilt corner points are used to determine the operation path points of the end effector when the robot collects the bed products and makes the bed; the mattress geometric center is used to assist in planning the initial unfolding posture of the robotic arm when the robot makes the bed.

7. The embodied robot control system according to claim 1, wherein After the behavior execution module controls the robot body to execute the functional instruction, the environmental perception unit of the sensor module immediately obtains the environmental information containing the bed surface image information within the specified time, and returns the environmental information to the control module for the task planning engine to call the quality standard for bed-making quality inspection and judge whether the preset service quality requirements are met.

8. The embodied robot control system according to claim 7, wherein The operation specification library stores multiple bed-making quality standards, and the bed-making quality standards include bed-making skill point rules and quality inspection area rules; the bed-making skill point rules include the standard actions and parameter thresholds of bed-making operations, and the quality inspection area rules include the inspection areas, shooting angles, feature extraction methods, and judgment criteria of quality inspection items; The bed-making quality detection is used to determine whether the bed-making effect meets the preset service quality requirements; if the quality detection result does not meet the service quality requirements, a temporary quality concern point is generated to trigger the bed-making quality correction operation; The temporary quality concern point includes the information of the non-compliant position; The bed-making quality correction operation means that the task planning engine re-plans the subsequent bed-making action sequence according to the temporary quality concern point to iteratively optimize the bed-making quality.

9. The embodied robot control system according to claim 1, characterized in that The end effector integrates a pneumatic suction cup and a flexible gripper; the pneumatic suction cup is installed at the center of the front end of the end effector and uses the negative pressure adsorption principle to be preferentially used for grasping flat and smooth textiles such as sheets and duvet covers; the flexible grippers are symmetrically distributed on both sides of the end effector and are suitable for grasping plane-shaped, fluffy or irregular items through multi-finger cooperative clamping.

10. The embodied robot control system according to claim 1, wherein, The behavior execution module is pre-integrated with a protection mechanism to cope with various potential emergencies; the protection mechanism includes: collision protection, emergency braking, visual monitoring and obstacle avoidance, and anti-pinch protection; after the protection mechanism is triggered, the behavior execution module stops executing the current control instruction and preferentially executes the action sequence of the protection mechanism corresponding to the emergency event without waiting for the control instruction of the control center.

Images