Catering service robot control method and system based on multi-modal perception
By dynamically allocating task priorities for catering service robots using multimodal perception technology, the problem of lack of dynamic task priority scheduling in existing technologies is solved. Dynamic priority adjustment based on dish type and user level is achieved, improving the timeliness of catering services and user experience.
Patent Information
- Application Number
- CN202510759929.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-09
- Publication Date
- 2025-09-19
AI Technical Summary
The task priority scheduling in the existing catering service robot control method lacks dynamics, resulting in long waiting times for hot drinks and reduced taste, and the priority fails to be dynamically adjusted according to user level.
Multimodal perception technology is used to obtain information on dish types and user levels, combined with environmental perception data, to dynamically assign task priorities, optimize driving routes through path planning algorithms, integrate global and local obstacle avoidance strategies, and adjust priority weights in real time.
It has achieved dynamic adjustment of task priority according to dish type and user level, improved the timeliness and pertinence of catering services, ensured the rapid delivery of hot drinks, and improved user experience.
Smart Images
Figure CN120669580A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a control method and system for a catering service robot based on multimodal perception. Background Art
[0002] In catering scenarios, especially during peak meal times, catering service robots need to handle multiple tasks simultaneously. However, existing control methods have the following technical problems:
[0003] Task priority scheduling lacks dynamism. Existing solutions use a fixed "first-come, first-served" order to execute tasks, failing to fully consider the impact of task attributes on priority. For example, hot drinks, which require high delivery times, can easily degrade in taste if they are queued for delivery in a fixed order due to long wait times. Summary of the Invention
[0004] In order to solve the technical problem of lack of dynamics in task priority scheduling in the prior art, the present invention provides a catering service robot control method based on multimodal perception.
[0005] The technical solution adopted in the present invention is:
[0006] A first aspect of the present application provides a method for controlling a catering service robot based on multimodal perception, comprising the following steps:
[0007] Step 1: Obtain task trigger information, identify the dish type and obtain user level information, and generate a task data packet containing the target location, task type, dish type and user level.
[0008] Step 2: Fusion processing is performed on the task data packet and the environmental perception data to generate task execution data containing real-time environmental information and task attributes, wherein the task attributes include dish type and user level.
[0009] Step 3: assign a priority level to the task according to the task attributes, wherein the priority level is associated with the temperature sensitivity of the dish type and the user level; and generate a path planning parameter including a priority weight based on the task execution data and the priority level.
[0010] Step 4: Combine the path planning parameters with a preset path planning algorithm to generate a driving route, and execute the task based on the driving route; wherein the preset path planning algorithm integrates global path planning and local obstacle avoidance strategy, and the global path planning adjusts the path calculation rules according to the priority weight.
[0011] Preferably, the step 1 includes the following:
[0012] Acquire dish type information based on RFID tag recognition technology, where the dish types include at least hot drinks, room temperature drinks, and frozen drinks;
[0013] Extract and classify the user's wearing identification or table identification through the visual recognition module to determine the user level information, where the user level includes at least VIP level and ordinary level;
[0014] The target location information is obtained based on the order system interface, and the target location, dish type and user level information are encapsulated to generate a standardized task data packet.
[0015] Preferably, step 2 includes the following:
[0016] The environment perception module collects obstacle distribution information, channel traffic status information and other robot location information in real time;
[0017] The Kalman filter algorithm is used to fuse the environmental perception data to generate an environmental grid map containing the probability distribution of obstacle positions;
[0018] The task data packet is aligned with the environmental grid map in time and space to generate task execution data including task priority attributes and environmental constraints.
[0019] Preferably, step 3 includes the following:
[0020] Based on the preset mapping relationship table between dish type and temperature sensitivity, corresponding temperature sensitivity weights are assigned to different dish types;
[0021] Determine a corresponding service priority coefficient according to the user level, wherein the service priority coefficient is used to adjust the timing priority of task execution;
[0022] Performing weighted fusion of the temperature sensitivity weight and the service priority coefficient to generate a priority level of the task;
[0023] The heuristic function parameters in the path planning algorithm are adjusted according to the priority level to generate path planning parameters including priority weights.
[0024] Preferably, step 4 includes the following sub-steps:
[0025] Mapping the priority weights in the path planning parameters to node cost evaluation parameters of a preset path planning algorithm, wherein the node cost evaluation parameters are used to adjust the output values of the evaluation function of the path nodes to generate a first node sequence with priority weights;
[0026] Based on the first node sequence with priority weights, a dynamic window algorithm is used to perform local obstacle avoidance planning to generate a continuous trajectory with spatiotemporal constraints;
[0027] Acquire the robot's posture information in real time through multi-sensor fusion positioning technology, and generate a motor control instruction sequence for adjusting the robot's motion state based on the temporal and spatial constraint deviation between the posture information and the continuous trajectory;
[0028] The path replanning process is triggered when any of the following conditions are detected:
[0029] The cumulative execution error of the motor control instruction sequence exceeds a preset threshold;
[0030] The target node in the continuous trajectory does not enter the robot's reachable workspace within the preset time window;
[0031] Environmental perception data shows that the target node is occupied by a dynamic obstacle and the predicted occupancy time exceeds the task time threshold.
[0032] The path replanning process adjusts the node expansion order and evaluation function parameters according to the current task priority level to generate a second node sequence with priority weights.
[0033] Preferably, during the task execution process of step 4, a dynamic priority adjustment step is also included:
[0034] Real-time monitoring of task status parameters, including the temperature change rate of the dish, the remaining aging threshold, and the distance between the current location and the target location;
[0035] Based on the task status parameters and the environmental perception data, determining whether a preset priority upgrade condition is met;
[0036] When the priority upgrade condition is met, the priority level of the task is recalculated and the priority weight in the path planning parameters is updated;
[0037] The updated priority weights are synchronized to the path replanning process to generate a third node sequence with priority weights.
[0038] Preferably, the determining whether a preset priority upgrade condition is met includes:
[0039] If the temperature change rate of the hot drink exceeds the preset threshold, and the difference between the current temperature and the target temperature is less than the first preset range, it is determined that the temperature-related upgrade condition is met;
[0040] If the ratio of the remaining time threshold to the time required to complete the current path is less than the second preset range, and the distance between the current location and the target location exceeds the third preset range, it is determined that the time-related upgrade conditions are met;
[0041] If the environmental perception data indicates that there is a dynamic obstacle in the path ahead and the predicted time to pass through the obstacle area exceeds a fourth preset range, it is determined that the environmental-related upgrade conditions are met;
[0042] When any of the above upgrade conditions is met, the priority level of the task is recalculated and the priority weight in the path planning parameters is updated.
[0043] The second aspect of the present application provides a catering service robot control system based on multimodal perception, which applies the above-mentioned catering service robot control method based on multimodal perception, including:
[0044] The task perception module is used to obtain task trigger information, identify dish type and obtain user level information, and generate a task data packet containing the target location, task type, dish type and user level.
[0045] A data fusion module is used to fuse the task data packet with the environmental perception data to generate task execution data containing real-time environmental information and task attributes, wherein the task attributes include dish type and user level.
[0046] A priority scheduling module is used to assign priority levels to tasks based on the task attributes, where the priority levels are associated with the temperature sensitivity of the dish type and the user level; and based on the task execution data and the priority levels, a path planning parameter including a priority weight is generated.
[0047] A path planning and execution module is used to combine the path planning parameters with a preset path planning algorithm to generate a driving route and perform tasks based on the driving route; wherein the preset path planning algorithm integrates global path planning and local obstacle avoidance strategies, and the global path planning adjusts the path calculation rules according to priority weights.
[0048] The beneficial effects of the present invention are at least one of the following: by obtaining task attribute information such as dish type and user level, combined with a mechanism for dynamically allocating priority levels based on task attributes, the traditional "first come, first served" fixed mode is changed, and the priority can be increased according to the user level to ensure a quick response, significantly improving the timeliness and pertinence of the service.
[0049] By integrating task data packets with environmental perception data, the preset path planning algorithm can obtain comprehensive task execution data. Global path planning can adjust path calculation rules according to priority weights. When sudden congestion occurs in the restaurant, high-priority tasks can automatically plan the shortest path. BRIEF DESCRIPTION OF THE DRAWINGS
[0050] Figure 1This is a schematic diagram of the method flow of embodiment 1 of the present invention;
[0051] Figure 2 This is a system structure block diagram of embodiment 2 of the present invention. DETAILED DESCRIPTION
[0052] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.
[0053] Example 1
[0054] This embodiment provides a control method for a catering service robot based on multimodal perception, such as Figure 1 As shown, the following steps are included:
[0055] Step 1: Obtain task trigger information, identify the dish type and obtain user level information, and generate a task data packet containing the target location, task type, dish type and user level.
[0056] In a possible implementation, step 1 includes the following:
[0057] Based on RFID tag recognition technology, dish type information is obtained, and the dish types include at least hot drinks, room temperature dishes, and frozen dishes; the visual recognition module is used to extract and classify features of the user's wearing identification or table identification to determine the user level information, and the user level includes at least VIP level and ordinary level; based on the order system interface, the target location information is obtained, and the target location, dish type and user level information are data-encapsulated to generate a standardized task data packet.
[0058] In a specific embodiment, for example, an RFID tag with a type identification is affixed to the dish when it is plated. The tag has built-in temperature sensitivity level data (such as a hot drink label marked "must be delivered within 10 minutes"). The robot reads the tag information through the built-in RFID reader to analyze the dish type (hot drink, room temperature, frozen) and basic time-sensitive attributes.
[0059] For example, the robot's visual recognition module (such as a camera) collects visual features such as the VIP badge worn by the user and the VIP sign placed on the dining table, and uses a convolutional neural network (CNN) model to classify the logo pattern and color and output the user level (VIP level / ordinary level).
[0060] For example, the API interface is used to connect with the restaurant order system in real time to obtain the target location coordinates corresponding to the order (such as the three-dimensional coordinates mapped by the table number), and the target location, dish type, and user level are encapsulated in JSON format as a standardized task data packet, including metadata such as timestamp and task ID.
[0061] Step 2: Fusion processing is performed on the task data packet and the environmental perception data to generate task execution data containing real-time environmental information and task attributes, wherein the task attributes include dish type and user level.
[0062] In a possible implementation, step 2 includes the following:
[0063] Obstacle distribution information, channel traffic status information and other robot position information are collected in real time through the environmental perception module; the environmental perception data are fused and processed using the Kalman filter algorithm to generate an environmental grid map containing the probability distribution of obstacle positions; the task data packet is aligned with the environmental grid map in time and space to generate task execution data containing task priority attributes and environmental constraints.
[0064] In a specific embodiment, for example, an environmental perception module composed of a lidar, an ultrasonic sensor, and a visual camera is used to collect in real time the coordinates of obstacles (such as tables and pillars) in the restaurant, the width of the channel, the dynamic density of people (through visual recognition statistics), and the positions of surrounding robots (IDs and coordinates are obtained through wireless communication).
[0065] For example, a Kalman filter algorithm is used to denoise multi-sensor data and generate an environmental grid map with a grid resolution of 0.5m. Each grid is labeled with a pass probability (0-1), where the obstacle area is labeled as 0, the congested area (people density > threshold) is labeled as 0.3-0.7, and the unobstructed area is labeled as 1.
[0066] For example, the target location coordinates in the task data packet are aligned with the environmental grid map coordinate system (e.g., the UTM coordinate system), and the current timestamp is added to the task data to match the real-time update timestamp of the environmental grid map, thereby generating task execution data including task attributes (e.g., hot drink category, VIP level) and environmental constraints (e.g., target path congestion probability 0.6).
[0067] Step 3: assign a priority level to the task according to the task attributes, wherein the priority level is associated with the temperature sensitivity of the dish type and the user level; and generate a path planning parameter including a priority weight based on the task execution data and the priority level.
[0068] In a possible implementation, step 3 includes the following:
[0069] Based on the preset mapping relationship table between dish types and temperature sensitivity, corresponding temperature sensitivity weights are assigned to different dish types; the corresponding service priority coefficient is determined according to the user level, and the service priority coefficient is used to adjust the timing priority of task execution; the temperature sensitivity weight and the service priority coefficient are weighted and fused to generate the priority level of the task; the heuristic function parameters in the path planning algorithm are adjusted according to the priority level to generate path planning parameters including priority weights.
[0070] In a specific implementation, for example, a "Dish Type-Temperature Sensitive Weight Mapping Table" is preset, for example, hot drinks have a corresponding weight of 0.6, room temperature drinks have a weight of 0.3, and frozen drinks have a weight of 0.2. The corresponding weight value is retrieved according to the dish type identified in step 1, and the corresponding coefficient of VIP level is defined as 1.5 and that of ordinary level as 1.0. The value is directly assigned through user level information to improve the timing priority of VIP tasks.
[0071] For example, a linear weighted formula (priority = temperature sensitivity weight × 0.7 + service priority coefficient × 0.3) is used to generate a priority value in the range of 0-2. The higher the value, the higher the priority (for example, the priority of the hot drink VIP task = 0.6 × 0.7 + 1.5 × 0.3 = 0.87).
[0072] The priority value is mapped to the heuristic function parameter of the path planning algorithm. For example, in the A* algorithm, the heuristic cost weight of the target node of the high-priority task is increased by 30%, and the movement cost weight of the non-target node is reduced by 20%, generating the path planning parameters including the priority weight.
[0073] Step 4: Combine the path planning parameters with a preset path planning algorithm to generate a driving route, and execute the task based on the driving route; wherein the preset path planning algorithm integrates global path planning and local obstacle avoidance strategy, and the global path planning adjusts the path calculation rules according to the priority weight.
[0074] In one possible implementation, step 4 includes the following sub-steps:
[0075] Mapping the priority weights in the path planning parameters to node cost evaluation parameters of a preset path planning algorithm, wherein the node cost evaluation parameters are used to adjust the output values of the evaluation function of the path nodes to generate a first node sequence with priority weights; based on the first node sequence with priority weights, using a dynamic window algorithm to perform local obstacle avoidance planning to generate a continuous trajectory containing spatiotemporal constraints; acquiring the robot's posture information in real time through multi-sensor fusion positioning technology, and generating a motor control instruction sequence for adjusting the robot's motion state based on the spatiotemporal constraint deviation between the posture information and the continuous trajectory;
[0076] The path replanning process is triggered when any of the following conditions are detected:
[0077] The cumulative execution error of the motor control instruction sequence exceeds a preset threshold;
[0078] The target node in the continuous trajectory does not enter the robot's reachable workspace within the preset time window;
[0079] Environmental perception data shows that the target node is occupied by a dynamic obstacle and the predicted occupancy time exceeds the task time threshold.
[0080] The path replanning process adjusts the node expansion order and evaluation function parameters according to the current task priority level to generate a second node sequence with priority weights.
[0081] In a specific embodiment, for example, the priority weight in the path planning parameters is converted into a node cost evaluation parameter. For example, for a task with a priority ≥ 1.5, the G value (actual cost) of the path node is calculated by subtracting 20% of the congestion cost, and the H value (heuristic cost) is increased by 15% of the target attractiveness.
[0082] For example, a first node sequence with priority weights is generated (such as the node sequence from the food pickup point to the VIP table: A→B→C, where node B is a congested area and the high-priority task forces the selection of the alternative path A→D→C with a higher cost but shorter distance).
[0083] For example, a dynamic window algorithm is used. Based on the global node sequence, 10 possible velocity vectors (linear velocity 0.2-0.5 m / s, angular velocity ±15° / s) are sampled with the current position of the robot as the center, and the predicted trajectory for the next 2 seconds at each speed is calculated.
[0084] For example, the predicted trajectories are evaluated for cost (collision risk, deviation from the global path distance), and the trajectory with the lowest cost is selected as the continuous trajectory to ensure real-time avoidance of dynamic obstacles (such as walking customers).
[0085] For example, the robot's position and posture are obtained through the fusion of a UWB positioning module, an inertial measurement unit (IMU), and a wheel speedometer, and are updated every 50ms. Based on the lateral deviation (>10cm) and angular deviation (>5°) from the target trajectory, a PID controller generates motor speed commands, driving the differential wheels to adjust the motion state.
[0086] For example, when the accumulated posture deviation exceeds 50 cm (derived from sensor fusion positioning error statistics), re-planning is triggered.
[0087] For example, the preset time window is 80% of the remaining time. If the Euclidean distance between the target node and the current position is greater than the maximum speed of the robot (0.6 m / s) × the time window, it is determined to be unreachable.
[0088] For example, if the target node in the environment grid map is marked as congested (passing probability < 0.4) three times in a row (1.5 seconds), and the remaining time is less than the predicted passing time, re-planning is triggered.
[0089] In a possible implementation, during the task execution process of step 4, a dynamic priority adjustment step is further included:
[0090] Real-time monitoring of task status parameters, including the temperature change rate of the dish, the remaining aging threshold, and the distance between the current location and the target location.
[0091] For example, for hot drinks, a built-in temperature sensor collects real-time temperature data and calculates the temperature change rate (°C / min). The remaining aging threshold is calculated as the initial aging time minus the elapsed time. The distance between the current location and the target location is calculated by accumulating the Euclidean distance of the path node sequence.
[0092] Based on the task status parameters and the environmental perception data, it is determined whether a preset priority upgrade condition is met.
[0093] When the priority upgrade condition is met, the priority level of the task is recalculated and the priority weight in the path planning parameters is updated.
[0094] The updated priority weights are synchronized to the path replanning process to generate a third node sequence with priority weights.
[0095] In a possible implementation, determining whether a preset priority upgrade condition is met includes:
[0096] If the temperature change rate of the hot drink exceeds the preset threshold, and the difference between the current temperature and the target temperature is less than the first preset range, it is determined that the temperature-related upgrade condition is met;
[0097] If the ratio of the remaining time threshold to the time required to complete the current path is less than the second preset range, and the distance between the current location and the target location exceeds the third preset range, it is determined that the time-related upgrade conditions are met;
[0098] If the environmental perception data indicates that there is a dynamic obstacle in the path ahead and the predicted time to pass through the obstacle area exceeds a fourth preset range, it is determined that the environmental-related upgrade conditions are met;
[0099] When any of the above upgrade conditions is met, the priority level of the task is recalculated and the priority weight in the path planning parameters is updated.
[0100] For example, if the temperature change rate is greater than 5°C / min and the current temperature is less than the target temperature -10°C (the preset hot drink taste threshold), it is determined that the priority needs to be increased.
[0101] If the remaining time threshold / the time required for the remaining distance of the current path is less than 0.8 (the preset risk ratio), and the remaining distance is greater than 5 meters (the preset long distance threshold), the time risk is judged to be high.
[0102] If there is a dynamic obstacle within 3 meters ahead (moving humans are detected through visual recognition) and the predicted passing time is greater than 50% of the remaining time, the environmental risk is judged to be high.
[0103] When temperature-related conditions are triggered, the temperature sensitivity weight increases by 20%; when time-limit conditions are triggered, the service priority coefficient increases by 0.3; when environmental conditions are triggered, the emergency priority rule (highest priority) is directly called.
[0104] After recalculating the priority, the node cost evaluation parameters in the path planning parameters are updated (e.g., the attractiveness of the target node is increased by 50%) and synchronized to the path replanning process.
[0105] For example, a preset "Priority Adjustment-Attractiveness Mapping Table" specifies that for every 0.1 increase in priority, the target node's heuristic cost weight (H-value weight) increases by 10%. For example, if a hot drink task triggers a priority upgrade due to temperature, assuming the original H-value weight is 0.8, it increases by 50% to 1.2 (i.e., 0.8 x 1.5).
[0106] For non-target nodes (such as detour nodes), their movement cost (G value) increases by a factor of 1.3 after the VIP mission priority is upgraded. This forces the path planning algorithm to prioritize paths with a higher concentration of target nodes.
[0107] Exemplarily, the specific steps of updating the node cost evaluation parameters include:
[0108] When a priority upgrade condition is determined to be met (such as an excessive temperature change rate or high aging risk), a parameter update command is sent. This command contains the trigger condition type (temperature / aging / environment), and the corresponding adjustment strategy is invoked based on the type. For example, a temperature-related upgrade focuses on adjusting the temperature sensitivity weight of the target node (such as adding a temperature attenuation factor to the H value); a aging-related upgrade shortens the path search window to improve real-time performance.
[0109] Based on the priority calculation logic of claim 3, assuming that the original priority level is P and the upgraded priority level is P', the node cost evaluation parameter adjustment formula is: H new =H old×{1+k×(P'-P)}; where k is the preset attraction adjustment coefficient (e.g., 0.5). The adjusted H-value and G-value weights are encapsulated into a new path planning parameter package, which includes metadata such as the task ID, adjustment timestamp, and parameter version number to ensure parameter uniqueness and timeliness.
[0110] For example, after receiving a new path planning parameter package, the robot immediately suspends the current local obstacle avoidance planning thread and triggers the global path replanning process. Based on the original global node sequence, only the path segments affected by the priority adjustment (such as nodes within 3 meters from the current position) are re-searched instead of the entire path reconstruction, thereby reducing computing resource consumption.
[0111] Example 2
[0112] This embodiment provides a catering service robot control system based on multimodal perception, and applies the catering service robot control method based on multimodal perception described in Example 1. Figure 2 Shown, including:
[0113] The task perception module is used to obtain task trigger information, identify dish type and obtain user level information, and generate a task data packet containing the target location, task type, dish type and user level.
[0114] A data fusion module is used to fuse the task data packet with the environmental perception data to generate task execution data containing real-time environmental information and task attributes, wherein the task attributes include dish type and user level.
[0115] A priority scheduling module is used to assign priority levels to tasks based on the task attributes, where the priority levels are associated with the temperature sensitivity of the dish type and the user level; and based on the task execution data and the priority levels, a path planning parameter including a priority weight is generated.
[0116] A path planning and execution module is used to combine the path planning parameters with a preset path planning algorithm to generate a driving route and perform tasks based on the driving route; wherein the preset path planning algorithm integrates global path planning and local obstacle avoidance strategies, and the global path planning adjusts the path calculation rules according to priority weights.
[0117] The above-described embodiments merely represent specific implementations of the present invention. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, and all such variations and improvements fall within the scope of protection of the present invention.
Claims
1. A control method for a catering service robot based on multimodal perception, characterized in that: The following steps are involved: Step 1: Obtain task trigger information, identify the dish type and obtain user level information, and generate a task data packet containing the target location, task type, dish type and user level; Step 2: fusing the task data packet with the environmental perception data to generate task execution data containing real-time environmental information and task attributes, wherein the task attributes include dish type and user level; Step 3: assigning a priority level to the task based on the task attributes, wherein the priority level is associated with the temperature sensitivity of the dish type and the user level; and generating a path planning parameter including a priority weight based on the task execution data and the priority level; Step 4: generating a driving route based on the path planning parameters and a preset path planning algorithm, and executing the task based on the driving route; in; The preset path planning algorithm integrates global path planning and local obstacle avoidance strategy, and the global path planning adjusts the path calculation rules according to the priority weight.
2. The method for controlling a catering service robot based on multimodal perception according to claim 1, characterized in that: The step 1 includes the following contents: Acquire dish type information based on RFID tag recognition technology, where the dish types include at least hot drinks, room temperature drinks, and frozen drinks; Extract and classify the user's wearing identification or table identification through the visual recognition module to determine the user level information, where the user level includes at least VIP level and ordinary level; The target location information is obtained based on the order system interface, and the target location, dish type and user level information are encapsulated to generate a standardized task data packet.
3. The method for controlling a catering service robot based on multimodal perception according to claim 1, characterized in that: The step 2 includes the following: The environment perception module collects obstacle distribution information, channel traffic status information and other robot location information in real time; The Kalman filter algorithm is used to fuse the environmental perception data to generate an environmental grid map containing the probability distribution of obstacle positions; The task data packet is aligned with the environmental grid map in time and space to generate task execution data including task priority attributes and environmental constraints.
4. The method for controlling a catering service robot based on multimodal perception according to claim 1, characterized in that: The step 3 includes the following contents: Based on the preset mapping relationship table between dish type and temperature sensitivity, corresponding temperature sensitivity weights are assigned to different dish types; Determine a corresponding service priority coefficient according to the user level, wherein the service priority coefficient is used to adjust the timing priority of task execution; Performing weighted fusion of the temperature sensitivity weight and the service priority coefficient to generate a priority level of the task; The heuristic function parameters in the path planning algorithm are adjusted according to the priority level to generate path planning parameters including priority weights.
5. The method for controlling a catering service robot based on multimodal perception according to claim 4, characterized in that: The step 4 includes the following sub-steps: Mapping the priority weights in the path planning parameters to node cost evaluation parameters of a preset path planning algorithm, wherein the node cost evaluation parameters are used to adjust the output values of the evaluation function of the path nodes to generate a first node sequence with priority weights; Based on the first node sequence with priority weights, a dynamic window algorithm is used to perform local obstacle avoidance planning to generate a continuous trajectory with spatiotemporal constraints; Acquire the robot's posture information in real time through multi-sensor fusion positioning technology, and generate a motor control instruction sequence for adjusting the robot's motion state based on the temporal and spatial constraint deviation between the posture information and the continuous trajectory; The path replanning process is triggered when any of the following conditions are detected: The cumulative execution error of the motor control instruction sequence exceeds a preset threshold; The target node in the continuous trajectory does not enter the robot's reachable workspace within the preset time window; Environmental perception data shows that the target node is occupied by a dynamic obstacle and the predicted occupancy time exceeds the task time threshold. The path replanning process adjusts the node expansion order and evaluation function parameters according to the current task priority level to generate a second node sequence with priority weights.
6. The method for controlling a catering service robot based on multimodal perception according to claim 1, characterized in that: During the task execution process of step 4, a dynamic priority adjustment step is also included: Real-time monitoring of task status parameters, including the temperature change rate of the dish, the remaining aging threshold, and the distance between the current location and the target location; Based on the task status parameters and the environmental perception data, determining whether a preset priority upgrade condition is met; When the priority upgrade condition is met, the priority level of the task is recalculated and the priority weight in the path planning parameters is updated; The updated priority weights are synchronized to the path replanning process to generate a third node sequence with priority weights.
7. The method for controlling a catering service robot based on multimodal perception according to claim 6, characterized in that: The determination of whether the preset priority upgrade condition is met includes: If the temperature change rate of the hot drink exceeds the preset threshold, and the difference between the current temperature and the target temperature is less than the first preset range, it is determined that the temperature-related upgrade condition is met; If the ratio of the remaining time threshold to the time required to complete the current path is less than the second preset range, and the distance between the current location and the target location exceeds the third preset range, it is determined that the time-related upgrade conditions are met; If the environmental perception data indicates that there is a dynamic obstacle in the path ahead and the predicted time to pass through the obstacle area exceeds a fourth preset range, it is determined that the environmental-related upgrade conditions are met; When any of the above upgrade conditions is met, the priority level of the task is recalculated and the priority weight in the path planning parameters is updated.
8. A catering service robot control system based on multimodal perception, characterized by The method for controlling a catering service robot based on multimodal perception according to any one of claims 1 to 7 comprises: A task perception module is used to obtain task trigger information, identify dish type and obtain user level information, and generate a task data packet containing the target location, task type, dish type and user level; a data fusion module, configured to fuse the task data packet with the environmental perception data to generate task execution data containing real-time environmental information and task attributes, wherein the task attributes include dish type and user level; a priority scheduling module configured to assign a priority level to a task based on the task attributes, the priority level being associated with the temperature sensitivity of the dish type and the user level; and to generate a path planning parameter including a priority weight based on the task execution data and the priority level; A path planning and execution module is used to combine the path planning parameters with a preset path planning algorithm to generate a driving route and perform tasks based on the driving route; wherein the preset path planning algorithm integrates global path planning and local obstacle avoidance strategies, and the global path planning adjusts the path calculation rules according to priority weights.