Hotel ground automatic cleaning robot path planning method

Abstract

The invention relates to the technical field of robot path planning, in particular to a hotel ground automatic cleaning robot path planning method, which comprises the following steps of collecting multi-source spatio-temporal data in a hotel monitoring system; performing fusion processing based on the multi-source spatio-temporal data, and constructing a spatio-temporal thermodynamic diagram; based on the space-time thermodynamic diagram, the current channel state is judged according to the density level and the duration of the target channel region, real-time passenger flow characteristics reflected by the space-time thermodynamic diagram and historical stain distribution information are fused, and the cleaning priority of each region is dynamically calculated and output; and according to the channel state and the cleaning priority, generating a differentiated cleaning path matched with the channel state and the cleaning priority, and synchronously configuring a corresponding advancing speed, a cleaning mode and an interaction prompt parameter set for the robot executing the path. According to the method, the guest flow space-time thermodynamic diagram can be constructed and utilized, and the cleaning path is dynamically planned to actively adapt to the passenger flow rule, so that the cleaning efficiency and safety are improved.

Description

Technical Field

[0001] This application relates to the technical field of robot path planning, and in particular to a path planning method for an automatic hotel floor cleaning robot. Background Technology

[0002] With the accelerated intelligent transformation of the hotel industry, autonomous floor cleaning robots have become an important tool for improving operational efficiency and reducing labor costs. Hotel environments are characterized by long, narrow corridors, densely packed rooms, and frequent personnel flow, posing unique challenges to the autonomous navigation and path planning capabilities of cleaning robots. Robots not only need to move safely and efficiently to their target locations in complex static environments, but also must avoid temporary obstacles such as guests and luggage carts in dynamic environments in real time to ensure operational safety and guest experience. Therefore, developing a path planning method that adapts to the unique environmental structure of hotels while balancing efficiency and safety is crucial for the large-scale deployment of cleaning robots.

[0003] To address the aforementioned needs, existing technologies have proposed several solutions. For example, Chinese Patent Publication No. CN115129058A discloses a composite path planning method and system for hotel cleaning robots. This method establishes a grid map, marks target points, and employs a strategy combining global path planning and local dynamic obstacle avoidance. Furthermore, this solution switches between different obstacle avoidance strategies based on the robot's environment, such as stopping and waiting when encountering obstacles in a corridor, prioritizing backward movement in narrow spaces, and choosing to detour in spacious areas. This method, through composite planning and strategy switching, aims to improve the robot's environmental adaptability and operational efficiency, reduce unnecessary stationary rotations, and ensure safe movement within a known static map framework.

[0004] However, in actual operation, the aforementioned existing technologies still have a core limitation: their path planning and decision-making mainly rely on the robot's own sensors to perceive the immediate, local physical environment, failing to effectively integrate and utilize the spatiotemporal dynamics of guest flow in a hotel setting. Guest flow within a hotel is not randomly distributed, but exhibits highly regular density changes and directional characteristics at different times and in different areas. Specifically, during the morning rush hour, the main lobby corridor experiences dense unidirectional guest flow. If the robot enters this core corridor according to global path planning, its local obstacle avoidance strategy will frequently trigger a stop-and-wait mode. This not only causes the robot to remain stationary for extended periods, severely delaying cleaning tasks, but also creates a bottleneck due to the frequent interaction between the robot and the dense crowd, hindering guest passage and causing congestion and guest dissatisfaction. Conversely, to avoid crowds, the robot may be forced to detour around the edges, missing the stains that most urgently need cleaning due to high-frequency foot traffic in the core corridor during that time. During the afternoon off-peak hours, when passageways are empty, existing methods lack the ability to assess the cleaning value of specific areas. They fail to prioritize robots covering the areas most likely to generate stains, instead operating in a fixed order or simply following the nearest available space. This leads to suboptimal resource allocation and overall inefficiency. When sudden liquid stains occur, existing methods can only decide whether to proceed based on immediate obstacle perception. Proceeding during peak hours exacerbates conflict between humans and robots; failing to proceed may cause the stain to spread. Therefore, existing technologies fail to meet the core technical requirements of dynamically adjusting cleaning paths based on guest flow characteristics, including area priority, directional alignment, and safety buffer zones. This results in an inability to simultaneously balance cleaning efficiency, operational safety, and guest experience in real-world hotel scenarios. Summary of the Invention

[0005] In order to construct and utilize a spatiotemporal heat map of guest flow, dynamically plan cleaning paths to proactively adapt to guest flow patterns, thereby improving cleaning efficiency and safety, this application provides a path planning method for an automatic hotel floor cleaning robot.

[0006] This application provides a path planning method for an automatic hotel floor cleaning robot, which adopts the following technical solution: A path planning method for an automatic hotel floor cleaning robot includes the following steps: Collect multi-source spatiotemporal data from the hotel monitoring system. The multi-source spatiotemporal data includes at least the guest's real-time location, area density, and direction of movement. Based on the multi-source spatiotemporal data, a spatiotemporal heat map is constructed to quantitatively characterize the passenger flow density and flow patterns in different areas of the hotel at different time periods. Based on the spatiotemporal heat map, the current channel status is determined according to the density level and duration of the target channel area, and the real-time passenger flow characteristics and historical stain distribution information reflected by the spatiotemporal heat map are integrated to dynamically calculate and output the cleaning priority of each area. Based on the channel status and cleaning priority, a differentiated cleaning path is generated to match it, and the corresponding travel speed, cleaning mode and interactive prompt parameter set are synchronously configured for the robot executing the path.

[0007] Optionally, determining the current channel status includes: Based on the spatiotemporal heat map, the consistency characteristics of passenger flow direction in the target channel area are extracted; When the density of the target channel area on the spatiotemporal heat map is continuously greater than or equal to the first density threshold, and the duration of this state reaches the first duration threshold, and the passenger flow direction consistency characteristic is greater than or equal to the direction consistency threshold, it is determined to be a directional peak period state. When the density of the target channel area on the spatiotemporal heat map is continuously less than or equal to the second density threshold, and the duration of this state reaches the second duration threshold, while the passenger flow direction consistency characteristic is lower than the direction consistency threshold, it is determined to be a sparse off-peak period state.

[0008] Optionally, if the situation is determined to be the directional peak period, an edge forward path that is consistent with the mainstream flow direction of guests corresponding to the passenger flow direction consistency feature is generated, and a first parameter set that matches the directional peak period is configured for the robot. If the situation is determined to be a sparse off-peak period, a spiral coverage path is generated with priority given to covering high-cleaning-priority areas along the main flow direction of guests, and a second set of parameters is configured for the robot to match the sparse off-peak period.

[0009] Optionally, when generating the edge forward path or spiral covering path, if the main flow direction of the guests is a single direction and its direction consistency is higher than the direction consistency threshold, then the robot's travel direction is controlled to be consistent with the main flow direction of the guests. If the main flow of guests is bidirectional, the robot's direction of travel is controlled to be consistent with the direction with the higher flow rate based on the real-time bidirectional flow rate extracted from the spatiotemporal heat map, and the robot's speed and lateral avoidance range are dynamically adjusted according to the bidirectional flow rate.

[0010] Optional features include dynamic obstacle avoidance steps: The robot is matched with a corresponding safe buffer distance threshold based on the channel status. When the real-time distance between the robot and the guest is less than the safe buffer distance threshold, an avoidance action corresponding to the channel state is executed; Wherein, if the channel status is a directional peak period status, the avoidance action is a deceleration warning and tangential steering; if the channel status is a sparse off-peak period status, the avoidance action is a pause prompt.

[0011] Optionally, when an emergency cleaning task is received for a specified location, a tiered response procedure is executed: Query the spatiotemporal heat map to obtain the real-time density level at the specified location; If the real-time density level is lower than the preset intervention threshold, then based on the mainstream flow direction of the guests corresponding to the specified location extracted from the spatiotemporal heat map, a temporary path is planned for the robot to approach the specified location from the current path along a direction that merges into the mainstream flow direction of the guests at a preset angle. If the real-time density level is higher than or equal to the intervention threshold, the emergency cleaning task and the designated location are marked, and a robot is scheduled to handle the task during subsequent sparse or low-peak periods.

[0012] Optionally, it may also include a closed-loop iterative optimization step: After the robot performs a cleaning task, it collects performance data including path tracking deviation, area cleaning coverage, and user feedback data. Based on the execution effect data, the feature parameters for generating the spatiotemporal heatmap, the threshold for determining the channel state, or the decision rules for planning cleaning paths are adaptively adjusted.

[0013] Optionally, the adaptive adjustment includes: When the cleaning coverage rate of the area is lower than a preset coverage threshold for more than a preset duration threshold, the cleaning priority of the corresponding area in the spatiotemporal heat map is increased. When the frequency of conflict reflected by the user feedback data exceeds the preset conflict threshold for an extended period beyond the preset duration threshold, the threshold for determining the directional peak period state or the sparse off-peak period state is adjusted accordingly.

[0014] Optionally, the spiral coverage path is generated through the following steps: Starting from the center of the high-cleaning-priority area, the starting radius and direction of spiral expansion are determined based on the density characteristics and cleaning priority distribution of the surrounding area extracted in real time from the spatiotemporal heat map. The robot is controlled to follow the expansion direction, executing a spiral path from the inside out and consistent with the main flow direction of the guests, to complete continuous coverage of the surrounding area.

[0015] Optionally, the multi-source spatiotemporal data is acquired through the following methods: The hotel monitoring system acquires real-time passenger flow data, including timestamps, real-time locations, area density, and direction of movement. Historical passenger flow pattern data containing predefined operating period information is synchronously obtained from the hotel management system. The passenger flow data and the historical passenger flow pattern data are aligned and integrated according to the time and space dimensions to form multi-source spatiotemporal data for constructing the spatiotemporal heat map.

[0016] In summary, this application includes the following beneficial technical effects: 1. This method constructs and utilizes a spatiotemporal heat map reflecting the density and flow patterns of passenger flow in various areas of a hotel to dynamically determine the status of passageways and calculate cleaning priorities. This generates differentiated cleaning paths that proactively adapt to passenger flow patterns, overcoming the core shortcomings of existing technologies that rely solely on local, real-time robot perception and cannot integrate the spatiotemporal dynamic patterns of passenger flow. This enables robots to proactively avoid core areas and operate along the edges during peak periods of high passenger flow, and prioritize covering high-demand areas during off-peak periods of low passenger flow. As a result, the overall efficiency of cleaning operations is significantly improved in complex dynamic environments, and operational interruptions and safety risks caused by conflicts between robots and dense crowds are greatly reduced.

[0017] 2. This method designs a graded response mechanism for emergency cleaning tasks based on real-time passenger flow density. When a sudden cleaning task is received, the system first queries the spatiotemporal heat map to obtain the real-time density level of the target area, and then makes an intelligent decision based on a preset threshold: whether to immediately plan a temporary path to intervene and handle the task, or to mark the task and delay its execution to an off-peak period. This effectively solves the dilemma faced by existing technologies in handling sudden stains and achieves a dynamic balance between emergency cleaning needs and ensuring the passenger passage experience.

[0018] 3. This method introduces a closed-loop iterative optimization step. After each task is performed by the robot, the system automatically collects multi-dimensional effect data such as path tracking deviation, area cleaning coverage, and user feedback. Based on this data, the system adaptively adjusts the feature parameters, state judgment thresholds, or path planning rules of the generated heat map. This enables the entire system to have continuous learning and optimization capabilities, allowing it to adapt to changes in hotel operation scenarios and the personalized characteristics of different areas, thereby steadily improving the long-term stability of cleaning services and user satisfaction. Attached Figure Description

[0019] Figure 1 This is a flowchart of the overall path planning method; Figure 2 This is the overall logic diagram of the path planning method. Detailed Implementation

[0020] The following is in conjunction with the appendix Figure 1-2 This application will be described in further detail.

[0021] This application discloses a path planning method for an automated hotel floor cleaning robot. For example... Figure 1 and Figure 2As shown, a path planning method for an automated hotel floor cleaning robot aims to solve the problems of low cleaning efficiency, frequent human-robot conflicts, and poor guest experience caused by existing technologies that rely on local real-time robot perception and fail to integrate the spatiotemporal patterns of hotel guest flow. The following steps are described in detail: S1 Multi-Source Spatiotemporal Data Acquisition This step integrates real-time passenger flow dynamics and historical passenger flow patterns in the hotel scenario to construct a comprehensive and accurate multi-source spatiotemporal dataset. This breaks through the limitations of existing technologies that rely solely on local perception data from robots, providing reliable support for the subsequent construction of spatiotemporal heat maps. The integrity and timeliness of the data directly ensure the rationality of subsequent decisions.

[0022] S11 Real-time Passenger Flow Data Acquisition The robot is integrated into the hotel's surveillance system. The human detection algorithm on the surveillance cameras is optimized for different scenarios such as hotel corridors and lobbies. The algorithm takes video frames captured in real time by the surveillance cameras as input and outputs human contour recognition results. The system counts the human contours in 1m increments. 2 The system calculates area density for statistical units. This statistical unit objectively reflects the level of crowding in an area, meeting the needs of hotel crowd flow statistics. Simultaneously, a trajectory tracking algorithm takes continuous video frames as input and outputs a sequence of movement trajectory coordinates for each guest. The system extracts the guest's movement direction information based on this coordinate sequence. Cameras are pre-calibrated in physical space, and the system converts the image pixel coordinates into hotel physical space coordinates using the calibration relationship to determine the guest's real-time location. The system synchronously records the timestamp for each data point, accurate to the second, ensuring data temporal correlation. The final acquired real-time crowd flow data includes timestamps, each guest's real-time location, and movement direction. Based on this, the system calculates density in 1-meter increments. 2 The area density is calculated in units (which can be adjusted according to environmental characteristics such as the actual width of the corridor).

[0023] Historical Passenger Flow Pattern Data Acquisition for S12 The robot establishes a data synchronization connection with the hotel management system, extracting historical passenger flow pattern data from the system. This data comes from over a year of operational passenger flow statistics records. After statistical analysis of this data, the system divides the system into predefined operating periods, including peak check-out times, peak check-in times, and sparse off-peak times. The division of these periods references hotel industry operating patterns to accurately reflect the passenger flow distribution characteristics at different times. This historical passenger flow pattern data provides a reference for interpreting real-time passenger flow data, helping the system more accurately determine whether the current passenger flow status conforms to normal patterns and reducing the impact of misjudgments caused by short-term passenger flow fluctuations.

[0024] S13 Data Alignment and Integration Processing The system integrates real-time passenger flow data with historical passenger flow pattern data according to time and spatial dimensions. In the time dimension, the system matches real-time data with historical data from the same period, on a minute-by-minute basis, ensuring that data at the same point in time can be cross-referenced. In the spatial dimension, the system associates data from different sources to the same area based on physical coordinates, achieving spatial data unification. During the integration process, the system employs a preprocessing workflow of data cleaning, deduplication, and completion. Data cleaning algorithms remove false detection data caused by camera obstruction and changes in lighting; deduplication algorithms delete duplicate records; and completion algorithms use interpolation to fill data gaps caused by short-term signal interruptions. These preprocessing algorithms are adapted to the characteristics of hotel passenger flow data, effectively improving data quality. The resulting multi-source spatiotemporal data combines real-time dynamics with historical pattern characteristics, providing high-quality data support for subsequent stages.

[0025] S2 Spatiotemporal Heatmap Construction This step takes over the multi-source spatiotemporal data output from S1 and transforms the scattered passenger flow information into a spatiotemporal heat map that can intuitively reflect the passenger flow density and flow patterns in different areas of the hotel at different times through targeted data fusion and quantitative representation. The generated heat map can provide global and dynamic support for subsequent decision-making and improve the accuracy of decision-making.

[0026] S21 Multi-source Data Fusion Processing The system employs a weighted fusion algorithm to process the multi-source spatiotemporal data output by S1. This algorithm is specifically adapted to the temporal correlation characteristics of hotel passenger flow. The input consists of real-time passenger flow data and historical passenger flow pattern data after alignment and integration by S1. The system dynamically adjusts the weights of the two types of data based on the timeliness of the data, with more recent historical data having a higher weight than older historical data. For example, historical data within the last 7 days is weighted at 0.4, historical data older than 7 days is weighted at 0.1, and real-time passenger flow data is weighted at 0.5. This weight allocation highlights the dynamics of real-time passenger flow while also taking into account the reference value of historical patterns, avoiding bias from a single data source, and ultimately outputting fused structured passenger flow data.

[0027] S22 Thermal Map Characterization Quantification The system constructs a minute-level precision time axis with time as the horizontal axis and a spatial axis with the hotel's physical coordinates as the vertical axis, forming a two-dimensional coordinate system. The system employs a kernel density estimation algorithm to process the fused passenger flow density data. This algorithm is adapted to the density distribution characteristics of different spatial scenarios, such as narrow hotel corridors and open lobbies. The input is the fused regional density data, and the output is the quantified density value at each location in the coordinate system. The system maps the quantified density values ​​to different color depths, with color depth positively correlated with passenger flow density. Based on statistical analysis of hotel passenger flow monitoring data, red corresponds to a passenger flow density ≥ 4 people / ㎡ (indicating congestion), yellow corresponds to a passenger flow density ≥ 2 people / ㎡ and < 4 people / ㎡ (indicating busyness), and blue corresponds to a passenger flow density < 2 people / ㎡ (indicating smooth flow). This classification references the pedestrian service level of public spaces and is based on the passage experience under typical hotel corridor widths (e.g., 1.5 meters), accurately and continuously representing different levels of congestion. Meanwhile, based on the fused passenger flow direction data, the system marks the main flow direction of guests in each area with arrows. The length of the arrows is positively correlated with the consistency of passenger flow direction. When the consistency of direction reaches more than 70%, the length of the arrows increases significantly. This threshold is determined by analyzing a large amount of hotel passenger flow trajectory data. It can effectively distinguish between single-direction and bidirectional passenger flow characteristics and finally generate a spatiotemporal heat map to realize the structured and visualized transformation of unstructured passenger flow information.

[0028] S3 Channel Status Determination and Cleaning Priority Calculation This step follows the spatiotemporal heat map generated in S2. By extracting core characteristics of passenger flow, accurately determining the status of passageways, and dynamically allocating cleaning priorities, it provides core decision support for subsequent differentiated path planning. The determination results and priority allocation directly affect the rational allocation of cleaning resources, helping to balance cleaning efficiency and guest experience.

[0029] S31 Passenger Flow Direction Consistency Feature Extraction The system uses the spatiotemporal heatmap generated by S2 as its data foundation, focusing on feature analysis of target passage areas. The system statistically analyzes the movement directions of all guests within the area, identifying the direction with the highest percentage as the mainstream flow direction. It then calculates the ratio of the number of guests moving in the mainstream flow direction to the total number of guests in the area; this ratio represents the consistency feature of the passenger flow direction. The feature value ranges from 0 to 1; a higher feature value indicates a more concentrated passenger flow within the area, while a lower feature value indicates a more dispersed passenger flow. This extraction process is optimized for different scenarios within hotel passageways and can be implemented using conventional statistical algorithms. The input is guest movement direction data from the heatmap, and the output is the consistency feature value.

[0030] S32 Current Channel Status Determination The system has a series of preset judgment thresholds, each of which has been tested in five hotels with different room sizes and verified against industry standards to adapt to the passenger flow characteristics of hotel scenarios: The first density threshold is 4 people / ㎡, which is derived from the passenger flow density statistics of the main corridor in the lobby during the morning peak hours of hotels. Exceeding this value will significantly increase the pressure on the corridor. The second density threshold is 1 person / ㎡, which corresponds to the normal passenger flow density of the corridor during the off-peak hours of hotels, when the corridor is more spacious. The first duration threshold is 10 minutes. Through simulation testing, a high density state that lasts for more than 10 minutes can be stably judged as a peak period, avoiding misjudgments caused by short-term fluctuations in passenger flow. The second duration threshold is 15 minutes. Passenger flow stability is relatively weak during off-peak hours, and a duration of 15 minutes can improve the accuracy of state judgment. The directional consistency threshold is 70%. It is derived from the analysis of a large amount of hotel passenger flow trajectory data. Reaching or exceeding this value can be considered as having a clear mainstream direction of passenger flow in the area.

[0031] When the density of the target channel area is continuously greater than or equal to the first density threshold, and the duration of this state reaches the first duration threshold, and the passenger flow direction consistency characteristic is greater than or equal to the direction consistency threshold, the system determines that the area is in a directional peak period state.

[0032] When the density of the target channel area is consistently less than or equal to the second density threshold, and the duration of this state reaches the second duration threshold, while the passenger flow direction consistency characteristic is less than the direction consistency threshold, the system determines that the area is in a sparse off-peak state.

[0033] If the target passage area does not meet the above-mentioned criteria for determining peak or sparse off-peak periods, it is classified as a normal passage state. In this state, the system will perform operations based on cleaning priority, using a default path (such as traveling along the central axis of the passage) and parameter set that balances efficiency and safety.

[0034] This multi-condition combined judgment method effectively avoids the one-sidedness of single indicator judgment and improves the accuracy of channel status recognition.

[0035] S33 Cleaning Priority Dynamic Calculation The system integrates real-time passenger flow characteristics reflected by spatiotemporal heat maps with historical stain distribution information to calculate cleaning priorities. The historical stain distribution information comes from the robot's past cleaning records and feedback from hotel manual cleaning. After statistical analysis of this information, the system clarifies the frequency of stain generation and common stain types in each area.

[0036] Priority calculation uses the probability of stain formation and cleaning urgency as core indicators, employing a weighted summation method to quantify and rank priorities, with level one being the highest priority. The weight of the stain formation probability is set based on the historical stain frequency of the area, with high-frequency stain areas having a higher weight than low-frequency stain areas. The cleaning urgency value is calculated inversely based on real-time passenger flow density (the lower the density, the higher the urgency value), and this value carries corresponding weight in the priority calculation. Simultaneously, the system will adjust the cleaning urgency based on the current passage status. For example, during peak periods, the system will appropriately lower the final cleaning priority of high-density core areas to align with the overall edge-oriented operational strategy, prioritizing passage safety and efficiency. This weight allocation ensures that cleaning resources are tilted towards high-demand, low-interference areas. For instance, the core area of ​​the main lobby passage has a high historical stain frequency; if the real-time passenger flow density is low, its cleaning priority will be significantly increased. Conversely, if the edge areas have high real-time passenger flow, even with a small number of stains, the cleaning priority will be appropriately reduced, thus achieving a dynamic balance between cleaning needs and guest experience.

[0037] S4 Differentiated Cleaning Path Generation and Parameter Configuration This step follows the channel status and cleaning priority determined by S3, generates cleaning paths adapted to different customer flow scenarios, and simultaneously configures operating parameters. Through the coordinated optimization of path shape and operating parameters, it minimizes interference with customers while ensuring cleaning effectiveness, thus solving the problem of efficiency and experience imbalance caused by fixed paths and single parameters in existing technologies.

[0038] S41 Peak Hour Status When S3 determines that the channel status is a directional peak period, the system generates an edge forward path. The path planning focuses on the edge area of ​​the channel, avoiding the core high-density area, and ensures that the path direction is completely consistent with the mainstream flow direction of customers, reducing cross contact with customers. This path can effectively cover high-frequency stains in the edge area while avoiding dense customer flow.

[0039] The system configures the robot with the following first set of parameters: the travel speed is set to 0.3m / s, which reduces the frequency of sudden stops and turns, avoiding interference with fast-moving customer traffic; the cleaning mode adopts a light sweeping mode with a roller brush speed of 1500rpm and a suction power of 70%, which effectively controls the operating noise while meeting the cleaning needs of light stains such as dust in edge areas; the interactive prompt is set to a 2Hz, 50dB warning sound, which alerts guests 3m in advance, serving as a warning without being too loud and affecting the guest experience.

[0040] S42 Sparse Low-Peak Status When S3 determines that the channel status is in a sparse, low-peak period, the system generates a spiral coverage path. The path generation starts from the center of the high-cleaning-priority area. The system extracts the density characteristics and cleaning priority distribution of the surrounding area in real time from the spatiotemporal heat map generated by S2 to determine the starting radius and direction of spiral expansion. The starting radius is set to 0.3m, which matches the robot's cleaning width to ensure that no area is missed in the initial cleaning range. The expansion direction prioritizes areas with higher cleaning priority, ensuring that cleaning resources are tilted towards areas with high demand.

[0041] The robot follows a spiral path from the inside out, consistent with the main flow of guests, to continuously cover the surrounding area. The system is simultaneously configured with a second set of parameters: a travel speed of 0.6 m / s to improve cleaning efficiency in low-interference environments; a deep cleaning mode with a brush speed of 2000 rpm and 100% suction power to target stubborn stains such as coffee stains and footprints in key areas, ensuring effective cleaning; and interactive warning sounds are disabled to avoid disturbing guests' rest or activities.

[0042] S43's direction and speed are dynamically adjusted. When generating edge-forward paths or spiral-covered paths, the system dynamically adjusts the robot's direction and speed based on the distribution characteristics of the main customer flow direction. If the main customer flow direction is unidirectional and the direction consistency characteristic is higher than the 70% direction consistency threshold set by S3, the robot's direction of travel will be consistent with the main customer flow direction, reducing the probability of reverse collisions.

[0043] If the main flow of customers is bidirectional, the system extracts the real-time bidirectional flow ratio from the spatiotemporal heatmap. The bidirectional flow ratio = number of customers in one direction / total number of customers in both directions. The system controls the robot's movement direction to align with the direction with the higher flow ratio. Simultaneously, the system dynamically adjusts the movement speed and lateral avoidance range based on the bidirectional flow ratio: the greater the difference in flow ratio, the closer the movement speed is to the default speed of the corresponding path, and the smaller the lateral avoidance range, with a minimum of 0.2m; conversely, the smaller the difference in flow ratio, the lower the movement speed, dropping to a minimum of 0.2m / s, and the larger the lateral avoidance range, with a maximum of 0.5m. These values ​​are theoretical calculations; actual avoidance will be subject to real-time constraints from the actual corridor width, walls, and fixed facilities, allowing for adjustments based on actual conditions. This adjustment logic effectively adapts to different flow distributions, balancing cleaning efficiency and traffic safety. For the general passage conditions determined by S3, the system uses a preset default cleaning path (such as traveling along the central axis of the channel) and a corresponding set of balanced parameters for operation, such as a travel speed of 0.45m / s and a standard cleaning mode (such as a roller brush speed of 1750rpm and a suction power of 85%), in order to achieve a balance between efficiency and safety.

[0044] S5 Dynamic Obstacle Avoidance This step follows the channel status determined by S3. By matching differentiated safety buffer distances and avoidance actions, it enables safe passage when the robot comes into close contact with customers. Its obstacle avoidance logic, which is dynamically adapted according to the customer flow scenario, not only ensures operational safety but also reduces the impact on cleaning efficiency, avoiding the problem of insufficient adaptation of a single obstacle avoidance mode in different customer flow scenarios.

[0045] S51 safety buffer distance threshold matching Based on the channel status determined by S3, the system matches the corresponding safe buffer distance threshold for the robot. During peak hours, when customer flow is dense and fast, the safe buffer distance threshold is set to 1.0m. This value allows the robot sufficient time to complete deceleration and turning maneuvers, effectively avoiding collisions with dense customer flow.

[0046] During sparse, off-peak periods, with fewer customers and a slower flow, a safety buffer distance threshold is set at 0.5m. Simulation results for off-peak scenarios show that this value, while ensuring safety, can reduce unnecessary robot pauses and improve the continuity of the cleaning process.

[0047] S52 Avoidance maneuver execution The robot is equipped with LiDAR and visual sensors, and uses laser SLAM combined with visual feature matching and preset environmental markers for localization and navigation. The LiDAR sensor detects the distance to the customer in real time; its measurement range is 0-10m with an accuracy of ±2cm, accurately acquiring real-time distance data. When the detected real-time distance is less than the safe buffer distance threshold for the corresponding passageway state, the robot performs an adaptive avoidance maneuver.

[0048] If the passageway is in a peak-hour state, the avoidance action is to slow down, issue a warning, and tangentially turn. The robot first reduces its speed to 0.1 m / s and issues a warning sound to alert nearby customers. Then, based on real-time perception of the gaps in the passenger flow, it uses a local path replanning algorithm (such as the dynamic window method) to generate an avoidance trajectory that tends to deviate in the forward direction, merge with the passageway with minimal interference, or approach the edge of the passageway to avoid direct conflict with dense passenger flow. After traveling a safe distance along this trajectory, it resumes its original path.

[0049] If the passageway is in a sparse, off-peak state, the avoidance action is a pause prompt. The robot immediately stops moving and issues a prompt through the voice module, such as "Please make way, cleaning in progress." It resumes moving after the customer leaves and the real-time distance is greater than the safe buffer distance threshold. This differentiated avoidance strategy can balance safety and efficiency in different scenarios and reduce interference with customer passage.

[0050] S6 Emergency Cleaning Level Response This step is designed for emergency cleaning tasks. It takes over the spatiotemporal heat map generated in S2 and the density level classification in S3, and realizes differentiated response through real-time density determination. It dynamically selects the processing time and path according to the passenger flow density, which not only ensures the effective treatment of emergency stains, but also avoids aggravating human-machine conflict, thus balancing emergency needs and customer experience.

[0051] S61 Real-time Density Level Query Emergency cleaning tasks are triggered by hotel staff reporting via a mobile app and by monitoring systems detecting liquid stains. When the system receives an emergency cleaning task, it immediately queries the spatiotemporal heat map generated by S2 to obtain real-time density data for the specified location, and classifies the real-time density into high, medium, and low levels based on the density threshold set by S3.

[0052] The system has a preset intervention threshold of 3 people / ㎡. This value has been verified by emergency scenario tests in eight hotels of different sizes. When the value is lower than this, the robot's entry for cleaning has less impact on guest passage. When the value is higher or equal to this, it is more likely to cause congestion or conflict.

[0053] S62 Temporary route planning or task marking appointment If the real-time density level is lower than the intervention threshold, the system extracts the main flow direction of guests corresponding to the specified location from the spatiotemporal heat map and plans a temporary path for the robot. The temporary path starts from the robot's current path and approaches the specified location along a direction that merges with the main flow direction of guests at a small angle (usually less than 30°), avoiding lateral crossing of the main flow of people. After cleaning, the robot returns to the original path along a similar path to ensure path continuity.

[0054] If the real-time density level is higher than or equal to the intervention threshold, the system marks the emergency cleaning task and the designated location. The marking information includes the coordinates of the stain location, the stain type, and the reporting time. Simultaneously, the system sends a reminder to the hotel's cleaning supervisor, dispatching a robot to handle the task during subsequent off-peak hours to ensure no cleaning task is missed. This differentiated response method has been validated in actual emergency scenarios at multiple hotels, balancing emergency cleaning needs with guest comfort under varying guest flow conditions.

[0055] S7 Closed-Loop Iterative Optimization This step builds upon the cleaning task execution results of the preceding steps. By collecting multi-dimensional effect data and optimizing system parameters and rules in reverse, it endows the system with self-learning capabilities based on business experience. It dynamically adjusts decision-making criteria and strategies based on actual operational data, allowing the system to continuously adapt to changes in hotel operation scenarios and steadily improve cleaning efficiency and user experience.

[0056] S71 Execution Effect Data Collection After the robot completes each cleaning task, the system automatically collects three types of core performance data to comprehensively reflect the quality of task execution and user feedback.

[0057] Path tracking deviation data represents the coordinate deviation between the actual path and the planned path. The system sets a reasonable deviation range of ≤5cm, which effectively measures the accuracy of path planning. Area cleaning coverage data is the ratio of the cleaned area to the total area, directly reflecting the coverage effect of the cleaning task. User feedback data includes the number of guest complaints and employee satisfaction ratings. The number of guest complaints is counted through the hotel service system, and employee satisfaction ratings are collected through employee mobile feedback. These two types of data reflect the degree of user satisfaction with the cleaning service from different perspectives.

[0058] S72 parameters and rules adaptive adjustment Based on the collected execution performance data, the system makes targeted adaptive adjustments to key parameters and decision rules to ensure the rationality and effectiveness of the adjustments.

[0059] The system has a preset coverage threshold of 95% and a preset duration threshold of 7 days. When the area's cleaning coverage rate is below the coverage threshold for an extended period exceeding the preset duration threshold, the system will initiate a diagnostic process: First, it will attempt to increase the cleaning priority of the corresponding area in the spatiotemporal heat map to increase the cleaning frequency. Simultaneously, the system will check the historical path accessibility data and obstacle reports for that area during the planned time period. If there are persistent physical obstructions or path planning failure records, it will send an anomaly alert to the management platform, prompting manual verification, rather than simply increasing robot scheduling.

[0060] The system has a preset conflict threshold of 1 time per week. When the frequency of conflicts reflected by user feedback data exceeds the conflict threshold for a continuous period exceeding the preset duration threshold, it indicates that the channel status judgment threshold is not well adapted to the actual passenger flow characteristics. The system will adjust the threshold for judging the directional peak period status or the sparse off-peak period status accordingly. For example, the first density threshold during the peak period will be lowered from 4 people / ㎡ to 3.5 people / ㎡ to trigger the peak period path and parameter configuration in advance.

[0061] When the area with a path tracking deviation consistently ≥8cm accounts for more than 10% of the total cleaning area of ​​the hotel, it indicates insufficient positioning accuracy in that area. The system marks a positioning correction zone in the underlying environmental map of that area and instructs the robot to activate its enhanced correction functions when traversing that area in subsequent tasks. This includes increasing the frequency of visual feature matching or utilizing pre-set environmental markers for pose correction to improve path tracking accuracy. These adjustments effectively address adaptation issues encountered in real-world scenarios, allowing for continuous system performance optimization.

[0062] The implementation principle of the path planning method for an automatic hotel floor cleaning robot in this application embodiment is as follows: By globally perceiving and predicting passenger flow patterns through spatiotemporal heat maps, the robot can actively avoid high-density core areas and operate along the edges during peak periods. This reduces cleaning interruptions and congestion caused by frequent obstacle avoidance, while also covering easily contaminated edge areas, significantly improving cleaning efficiency and operational safety. During sparse off-peak periods, a spiral coverage path is implemented according to cleaning priorities, prioritizing the cleaning of areas with a high probability of stains, thus optimizing the dynamic allocation of cleaning resources. At the same time, the robot dynamically adjusts its direction of travel, speed, and obstacle avoidance strategy based on real-time passenger flow characteristics, and responds according to real-time density levels during emergency cleaning tasks. This effectively reduces the frequency of human-robot conflicts and avoids interference with guest passage, thereby achieving a synergistic improvement in cleaning efficiency, operational safety, and guest experience overall.

[0063] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A path planning method for an automatic hotel floor cleaning robot, characterized in that, Includes the following steps: Collect multi-source spatiotemporal data from the hotel monitoring system. The multi-source spatiotemporal data includes at least the guest's real-time location, area density, and direction of movement. Based on the multi-source spatiotemporal data, a spatiotemporal heat map is constructed to quantitatively characterize the passenger flow density and flow patterns in different areas of the hotel at different time periods. Based on the spatiotemporal heat map, the current channel status is determined according to the density level and duration of the target channel area, and the real-time passenger flow characteristics and historical stain distribution information reflected by the spatiotemporal heat map are integrated to dynamically calculate and output the cleaning priority of each area. Based on the channel status and cleaning priority, a differentiated cleaning path is generated to match it, and the corresponding travel speed, cleaning mode and interactive prompt parameter set are synchronously configured for the robot executing the path.

2. The method according to claim 1, characterized in that, The determination of the current channel status includes: Based on the spatiotemporal heat map, the consistency characteristics of passenger flow direction in the target channel area are extracted; When the density of the target channel area on the spatiotemporal heat map is continuously greater than or equal to the first density threshold, and the duration of this state reaches the first duration threshold, and the passenger flow direction consistency characteristic is greater than or equal to the direction consistency threshold, it is determined to be a directional peak period state. When the density of the target channel area on the spatiotemporal heat map is continuously less than or equal to the second density threshold, and the duration of this state reaches the second duration threshold, while the passenger flow direction consistency characteristic is lower than the direction consistency threshold, it is determined to be a sparse off-peak period state.

3. The method according to claim 2, characterized in that: If the situation is determined to be the directional peak period, an edge forward path that is consistent with the mainstream flow direction of guests corresponding to the passenger flow direction consistency feature is generated, and a first parameter set that matches the directional peak period is configured for the robot. If the situation is determined to be a sparse off-peak period, a spiral coverage path is generated with priority given to covering high-cleaning-priority areas along the main flow direction of guests, and a second set of parameters is configured for the robot to match the sparse off-peak period.

4. The method according to claim 3, characterized in that, When generating the edge forward path or spiral covering path, if the main flow direction of the guests is a single direction and its direction consistency is higher than the direction consistency threshold, then control the robot's travel direction to be consistent with the main flow direction of the guests. If the main flow of guests is bidirectional, the robot's direction of travel is controlled to be consistent with the direction with the higher flow rate based on the real-time bidirectional flow rate extracted from the spatiotemporal heat map, and the robot's speed and lateral avoidance range are dynamically adjusted according to the bidirectional flow rate.

5. The method according to claim 2, characterized in that, Including dynamic obstacle avoidance steps: The robot is matched with a corresponding safe buffer distance threshold based on the channel status. When the real-time distance between the robot and the guest is less than the safe buffer distance threshold, an avoidance action corresponding to the channel state is executed; Wherein, if the channel status is a directional peak period status, the avoidance action is a deceleration warning and tangential steering; if the channel status is a sparse off-peak period status, the avoidance action is a pause prompt.

6. The method according to claim 3, characterized in that, When an emergency cleaning task is received for a specified location, a tiered response procedure is executed: Query the spatiotemporal heat map to obtain the real-time density level at the specified location; If the real-time density level is lower than the preset intervention threshold, then based on the mainstream flow direction of the guests corresponding to the specified location extracted from the spatiotemporal heat map, a temporary path is planned for the robot to approach the specified location from the current path along a direction that merges into the mainstream flow direction of the guests at a preset angle. If the real-time density level is higher than or equal to the intervention threshold, the emergency cleaning task and the designated location are marked, and a robot is scheduled to handle the task during subsequent sparse or low-peak periods.

7. The method according to claim 2, characterized in that, It also includes closed-loop iterative optimization steps: After the robot performs a cleaning task, it collects performance data including path tracking deviation, area cleaning coverage, and user feedback data. Based on the execution effect data, the feature parameters for generating the spatiotemporal heatmap, the threshold for determining the channel state, or the decision rules for planning the cleaning path are adaptively adjusted.

8. The method according to claim 7, characterized in that, The adaptive adjustment includes: When the cleaning coverage rate of the area is lower than the preset coverage threshold for a continuous period of time exceeding the preset duration threshold, the cleaning priority of the corresponding area in the spatiotemporal heat map is increased. When the frequency of conflict reflected by the user feedback data exceeds the preset conflict threshold for an extended period beyond the preset duration threshold, the threshold for determining the directional peak period state or the sparse off-peak period state is adjusted accordingly.

9. The method according to claim 3, characterized in that, The spiral coverage path is generated through the following steps: Starting from the center of the high-cleaning-priority area, the starting radius and direction of spiral expansion are determined based on the density characteristics and cleaning priority distribution of the surrounding area extracted in real time from the spatiotemporal heat map. The robot is controlled to follow the expansion direction, executing a spiral path from the inside out and consistent with the main flow direction of the guests, to complete continuous coverage of the surrounding area.

10. The method according to claim 1, characterized in that, The multi-source spatiotemporal data is obtained through the following methods: The hotel monitoring system acquires real-time passenger flow data, including timestamps, real-time locations, area density, and direction of movement. Historical passenger flow pattern data containing predefined operating period information is synchronously obtained from the hotel management system. The passenger flow data and the historical passenger flow pattern data are aligned and integrated according to the time and space dimensions to form multi-source spatiotemporal data for constructing the spatiotemporal heat map.

Images