A hierarchical temporal-based multi-task cooperative scheduling method and system for rice robots
By using a hierarchical time-series scheduling method and real-time environmental perception, the timing of multi-task collaborative operations of rice robots is dynamically adjusted, solving the problems of misalignment and pollution under complex working conditions, and realizing efficient and safe multi-module collaborative operations.
Patent Information
- Application Number
- CN202611217666.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-12
- Publication Date
- 2026-11-10
AI Technical Summary
Existing multi-task collaborative scheduling methods for rice robots are difficult to adapt to changes in mud depth and wheel slippage under complex working conditions, leading to operational misalignment and process contamination risks. Furthermore, they lack the ability to analyze the dynamic motion envelope and process interaction effects in real time, making it difficult to achieve efficient and accurate multi-module collaborative operations.
A hierarchical timing-based scheduling method is adopted. The target operation trajectory is planned by using environmental perception data and operation sensor data. Combined with timing logic and mutual exclusion lockout constraints, the operation timing is dynamically adjusted, the risks of physical interference and process crossover are solved in real time, and the collaborative or lockout instructions are generated to control the dynamic switching of heterogeneous operation modules.
It enables precise collaborative operation of multiple modules under complex working conditions, avoiding misalignment of operations and process pollution, improving operational efficiency and agronomic safety, and reducing waste of pesticides and fertilizers.
Smart Images

Figure CN123067546A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of robotics, and in particular to a multi-task collaborative scheduling method and system for rice robots based on hierarchical temporal sequence. Background Technology
[0002] Rice field management encompasses various agronomic processes such as weeding, fertilization, and pesticide spraying. To improve operational efficiency, current technologies tend to integrate multiple operational modules onto a single mobile platform, enabling multi-functional composite operations. In such multi-module integrated systems, coordinating the execution sequence of heterogeneous operational modules in time and space to avoid mutual interference is a key technical challenge for achieving efficient and precise field management.
[0003] Currently, the common approach to multi-task collaborative scheduling of rice-growing robots is to uniformly schedule each task module based on a preset global operation path and a fixed task sequence table. Some solutions calculate the start-stop delay between modules using simple physical distance and travel speed to maintain approximate alignment between preceding and following processes; others set fixed no-entry zones or switching thresholds to shut down specific modules under certain conditions to avoid conflicts. These scheduling strategies typically rely on offline pre-set rules, with each module executing sequentially or synchronously according to a predetermined plan during operation.
[0004] However, the aforementioned existing technologies have the following shortcomings: First, fixed-time scheduling is difficult to adapt to the real-time speed fluctuations caused by complex working conditions such as changes in mud depth in paddy fields and wheel slippage. The inter-module coordination delay calculated based on static parameters is prone to deviating from the actual spatial alignment requirements, resulting in operational misalignment, such as the fertilization position deviating from the weeding trajectory. Second, the preset conflict avoidance rules lack the ability to analyze the dynamic motion envelope and process interaction effects of each module in actual operation in real time, making it difficult to balance the efficiency and safety of cross-module parallel operation. For example, it is impossible to accurately determine the locking timing of the spraying module when turning, which can easily lead to pesticide damage or missed spraying. It is also impossible to effectively identify the process pollution risk caused by the mud and water stirred up by high-speed weeding to foliar spraying. Summary of the Invention
[0005] The purpose of this application is to provide a multi-task collaborative scheduling method and system for rice robots based on hierarchical timing, which solves the problem of inaccurate and conflict-prone multi-module collaboration under complex working conditions.
[0006] To address the aforementioned technical problems, the embodiments of this application provide the following technical solutions: The first aspect of this application provides a multi-task cooperative scheduling method for rice robots based on hierarchical temporal sequence, the method comprising: Acquire environmental perception data, operational sensing data, and operational status data of each heterogeneous operation module of the rice robot, and plan the target operation trajectory with task attributes accordingly. Based on the environmental perception data and the task attributes of the target operation trajectory, the running state machine model corresponding to each heterogeneous operation module is called and initialized. The running state machine models are pre-set with constraint rules based on time-series logic, including time-series coordination constraints and mutual exclusion locking constraints. The actual travel speed is obtained from the operation sensor data, and coordinated scheduling is determined by combining the physical interval distance of each heterogeneous operation module on the vehicle body, the operation status data, and the constraint rules. When it is determined that the timing coordination constraint is satisfied, a timing coordination instruction is generated to control the timing coupling of the operation actions of each heterogeneous operation module at the target position; when it is determined that there is physical interference or process cross-interference between the heterogeneous operation modules and the mutual exclusion lockout constraint is satisfied, a module lockout instruction is generated to forcibly lock the target operation module. The timing coordination instruction or the module locking instruction is used as a trigger event and input into the corresponding running state machine model to drive the corresponding running state machine model to dynamically switch between standby, running or locked states, and output the underlying control instruction according to the state after the switch to control each heterogeneous operation module to perform operations.
[0007] A second aspect of this application provides a hierarchical temporal-based multi-task collaborative scheduling system for rice robots, the system comprising: The acquisition module is used to acquire environmental perception data, operational sensing data, and operational status data of each heterogeneous operation module of the rice robot, and to plan the target operation trajectory with task attributes accordingly. An initialization module is used to call and initialize the running state machine model corresponding to each heterogeneous operation module based on the environmental perception data and the task attributes of the target operation trajectory. The running state machine models are pre-set with constraint rules based on time-series logic, including time-series coordination constraints and mutual exclusion locking constraints. The collaborative scheduling determination module is used to obtain the actual travel speed from the operation sensor data, and combine it with the physical interval distance of each heterogeneous operation module on the vehicle body, the operation status data, and the constraint rules to make collaborative scheduling determination. The instruction generation module is used to generate a timing coordination instruction when it is determined that the timing coordination constraint is satisfied, in order to control the timing coupling of the operation actions of each heterogeneous operation module at the target position; when it is determined that there is physical interference or process cross-interference between the heterogeneous operation modules and the mutual exclusion lockout constraint is satisfied, the module lockout instruction is generated to forcibly lock the target operation module. The control operation module is used to input the timing coordination instruction or the module lock instruction as a trigger event into the corresponding running state machine model, drive the corresponding running state machine model to dynamically jump between standby, running or locked states, and output the underlying control instruction according to the state after the jump to control each heterogeneous operation module to perform operations.
[0008] Compared to existing technologies, the multi-task collaborative scheduling method and system for rice robots based on hierarchical temporal sequence provided in this application effectively solves the problem of inaccurate and conflict-prone multi-module collaboration under complex paddy field conditions by constructing a multi-task collaborative scheduling method based on hierarchical temporal sequence, which deeply integrates environmental perception, trajectory planning, and state machine constraints based on temporal logic. By introducing temporal collaborative constraints and mutual exclusion locking constraints, and combining real-time travel speed, attitude changes, and physical distance between modules for collaborative scheduling judgment, heterogeneous modules such as weeding, fertilizing, and spraying can adaptively adjust their operation sequence according to the actual degree of slippage and vehicle dynamics, realizing precise relay operation of multiple modules at the same target location. At the same time, by real-time calculation of spatial motion envelope and medium drift boundary, the risk of physical interference and process cross-interference can be detected in advance and forced locking can be executed, fundamentally avoiding module collisions and cross-contamination, significantly improving operation efficiency and agronomic safety, and reducing waste of pesticides and fertilizers. Attached Figure Description
[0009] The above and other objects, features, and advantages of exemplary embodiments of this application will become readily understood by reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of this application are illustrated by way of example and not limitation, with the same or corresponding reference numerals denoteing the same or corresponding parts, wherein: Figure 1 A flowchart illustrating a multi-task cooperative scheduling method for rice robots based on hierarchical timing is shown schematically. Figure 2 A schematic diagram of a multi-task collaborative scheduling system for rice robots based on hierarchical timing is shown. Detailed Implementation
[0010] Exemplary embodiments of this application will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of this application are shown in the drawings, it should be understood that this application may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of this application and to fully convey the scope of this application to those skilled in the art.
[0011] It should be noted that, unless otherwise stated, the technical or scientific terms used in this application shall have the ordinary meaning as understood by one of ordinary skill in the art to which this application pertains.
[0012] The methods described in the embodiments of this application will be explained in detail below.
[0013] This embodiment provides a hierarchical temporal scheduling method for multi-task collaborative scheduling of rice robots. This method can be applied to multi-task collaborative operation scenarios such as weeding, fertilizing, and spraying in rice paddy management. In this embodiment, the algorithm can flexibly deploy tasks on a cloud server or an edge computing terminal mounted on the robot body, depending on the actual needs of the scenario. The tasks are completed collaboratively by the cloud server and the robot terminal, or by the robot terminals themselves, through a wireless communication network.
[0014] The proposed solution adopts a three-layer control architecture consisting of a decision-making layer, a scheduling layer, and an execution layer. Each layer has a clearly defined function and progresses step by step, enabling precise collaboration among multiple modules under complex working conditions.
[0015] The decision-making layer is responsible for global task planning and environmental adaptation modeling. This layer receives environmental perception data from multiple sensors, constructs a gridded environmental adaptation model, and plans global operation paths and multi-task execution sequence plans based on agronomic requirements and task priorities. It then divides tasks into meta-action sequences and distributes them to the scheduling layer.
[0016] The scheduling layer introduces a temporal logic parsing algorithm to establish separate state machine models for heterogeneous tasks such as weeding, fertilizing, and spraying. This layer dynamically parses the temporal coordination constraints and mutual exclusion constraints between the state machines based on real-time operating conditions (such as travel speed, plant density, and vehicle posture). When conditions are met, cross-module parallel operation instructions are triggered; when a critical state occurs, module locking instructions are triggered, achieving millisecond-level scheduling response.
[0017] The execution layer consists of controllers for various heterogeneous operation modules distributed on the robot, connected to the scheduling layer via a fieldbus. This layer receives timing coordination commands or module locking commands from the scheduling layer, drives the actions of each actuator, and simultaneously feeds back the actual status words of each module to the scheduling layer in real time, forming a closed-loop verification.
[0018] In this application, the heterogeneous operation modules mounted on the rice-growing robot refer to independent operating implements with different functions, structures, and working principles. These implements are connected to the main controller via a CAN bus or other fieldbuses and are subject to unified scheduling by the scheduling layer. Common heterogeneous operation modules include, but are not limited to, the following categories: Mechanical weeding module Composed of actuators such as weeding wheels, weeding claws, or weeding rollers, this module removes weeds between rice rows through high-speed rotation or reciprocating motion. It is typically equipped with a lifting mechanism to adjust the working depth and height as needed.
[0019] Fertilization module This includes devices such as side-deep fertilization guns, surface fertilization discs, or variable displacement fertilization screws. Side-deep fertilization guns precisely apply fertilizer deep into the soil, while surface fertilization discs evenly distribute fertilizer on the surface. This module is equipped with a fertilizer discharge shaft speed sensor and a fertilizer blockage detector, providing real-time feedback on fertilizer application rate and operating status.
[0020] Spraying module It consists of a medicine tank, a liquid pump, a solenoid valve, a spray boom, and nozzles. The spray boom can be folded and unfolded, and the nozzles can be independently controlled to open and close. This module can perform variable spraying based on prescription maps or real-time sensing results, and can also perform targeted point spraying on specific locations.
[0021] Foliar fertilizer spraying module This module functions similarly to the pesticide spraying module, but uses liquid foliar fertilizer as the medium. Because foliar fertilizer and certain pesticides may react chemically during spraying, potentially causing phytotoxicity, there is a chemical process mutual exclusion relationship between this module and the pesticide spraying module, requiring mutual exclusion and interlocking control at the scheduling layer.
[0022] Surface leveling module (scraper) Used in the land preparation stage before rice transplanting, this module levels the muddy surface using a leveling shovel. Equipped with a resistance sensor, it detects soil hardness and depth, providing a basis for adjusting operating parameters for the subsequent fertilization module.
[0023] Crop growth monitoring module Composed of a multispectral camera or an RGB-D camera, it is installed at the front of the vehicle to collect real-time spectral information of the rice canopy, calculate growth indicators such as the Normalized Difference Vegetation Index (NDVI), as well as the degree of pests and diseases and the density of weed distribution.
[0024] Soil / mud depth detection module Using lidar or ultrasonic sensors, high-frequency distance measurement is performed on the mud surface of paddy fields. Combined with the wheel axle subsidence displacement, data on mud depth and subsidence degree are extracted in real time.
[0025] Walking chassis module This module includes drive wheels, tracks or wheeled locomotives, and steering servos. It is responsible for the robot's movement and steering, and the data output from its wheel speed encoders and inertial measurement units forms the basis for calculating slippage rate and attitude changes.
[0026] Step 110: Obtain environmental perception data, operational sensor data, and operational status data of each heterogeneous operation module of the rice robot, and plan the target operation trajectory with task attributes accordingly.
[0027] In this step, the system collects three types of heterogeneous data in real time through multi-source sensors mounted on the rice robot, which serves as the data basis for subsequent state machine construction and collaborative scheduling determination.
[0028] Environmental perception data is used for macroscopic agronomic feature identification and microscopic paddy field geophysical characteristic perception. This may include, but is not limited to: two-dimensional / three-dimensional visual image data collected by front-view / rear-view cameras and multispectral cameras to identify the physical edges of rice rows and field ridge boundaries, and to quantitatively calculate crop growth and weed distribution density; and high-frequency point ranging of paddy field mud surface by lidar or depth camera, combined with wheel axle sinking displacement, to extract mud foot depth sequence or subsidence degree sequence, in order to characterize the changes in mud resistance caused by fluctuations in paddy field soil moisture content and viscosity.
[0029] The operational sensor data is used to characterize the kinematic and dynamic state of the rice robot body at high frequency. This may include, but is not limited to: the drive wheel speed sequence directly measured by the wheel encoder, and the actual travel speed sequence representing the actual displacement of the vehicle body relative to the ground calculated by the integrated navigation system; and the multi-degree-of-freedom attitude change characteristics of the vehicle body output by the inertial measurement unit, including heading angle, roll angle, pitch angle and corresponding angular velocity and acceleration time sequence characteristics, which are used to characterize the dynamic attitude fluctuations of the robot caused by uneven ground or turning.
[0030] The operational status data of each heterogeneous operation module is fed back in real time by the controllers of heterogeneous implements such as weeding, fertilizing, and spraying. This data may include, but is not limited to: dynamic mechanical structural parameters of the actuators of each module, such as the extension length of the weeding arm, the unfolding and folding angle of the spray boom, and the lifting height and position of the cutter. These parameters are combined with the physical spacing between each module on the vehicle body to calculate the physical occupancy of each implement in three-dimensional space in real time; and the jet status of the process medium of each module, such as spraying pressure, nozzle flow rate opening, fan speed, and fertilizer discharge shaft speed, to characterize whether each module is currently releasing the working medium and the dynamic diffusion characteristics of the medium.
[0031] After acquiring the above three types of data, the system further combines the pre-stored field geographic information and the spatial distribution map of operation parameters, performs spatiotemporal alignment and fusion of real-time environmental perception data and the spatial distribution map of operation parameters, generates the target operation trajectory with the physical boundary of the field as a constraint, and generates operation behavior labels based on crop growth and weed distribution density, and maps the operation behavior labels to the corresponding spatial location of the target operation trajectory as task attributes.
[0032] In a specific implementation, planning a target operation trajectory with task attributes can be achieved through the following steps: Obtain the pre-stored field geographic information and spatial distribution map of operation parameters; Real-time environmental perception data is spatiotemporally aligned and fused with the spatial distribution map of operational parameters to identify the current crop growth and weed distribution density in the field. Using the physical boundaries defined by the geographic information of the field as constraints, a target operation trajectory is generated, and operation behavior labels are generated based on the distribution density of weeds and the growth of crops. The operation behavior labels are then mapped to the corresponding spatial locations of the target operation trajectory as task attributes.
[0033] Specifically, the system pre-stores the geographical information of the fields, including basic data such as field boundaries, shape, and area, as well as a spatial distribution map of operational parameters. This map indicates the types and amounts of operations to be performed at different locations within the field, such as which areas need fertilization and how much, and which areas need spraying and what pesticides to use. During real-time operations, the system spatiotemporally aligns real-time environmental data collected by sensors such as LiDAR and multispectral cameras with the pre-stored spatial distribution map of operational parameters. Spatiotemporal alignment means unifying data collected from different times and different sensors to the same coordinate system and the same point in time, ensuring that subsequent processing will not encounter errors due to data misalignment. After alignment, the system identifies the crop growth and weed density at different locations within the field based on the analysis results of the multispectral images.
[0034] With this information, the system uses the geographical boundaries of the field as constraints to plan a complete operational trajectory within those boundaries. During the planning process, the system assigns corresponding operational behavior labels to each spatial location on the trajectory based on the previously identified weed density and crop growth. For example, locations with dense weeds are labeled "key weeding," locations with poor crop growth are labeled "additional fertilization," and locations without abnormalities are labeled "routine passage." These labels attached to the trajectory form a target operational trajectory with task attributes. During subsequent scheduling, the system directly reads these labels to determine what operation should be performed at each location.
[0035] Step 120: Based on the task attributes of environmental perception data and target operation trajectory, call and initialize the running state machine model corresponding to each heterogeneous operation module. There are pre-set constraint rules based on time logic between the running state machine models. The constraint rules include time coordination constraints and mutual exclusion locking constraints.
[0036] To ensure the coordinated and safe operation of diverse functional modules such as weeding, spraying, and fertilizing in the complex and ever-changing paddy field environment, this system employs a finite state machine-based temporal state control approach, rather than relying on numerous preset fixed conditions. After generating the target operation trajectory with task attributes in step 110, based on the acquired environmental perception data and the task attributes bound to the current trajectory segment (such as heavy weeding or variable spraying), the corresponding running state machine model is invoked for each heterogeneous module that needs to participate in the operation, and initialization is completed.
[0037] Each running state machine model has pre-embedded timing logic-based constraint rules. These rules uniformly define the sequence of actions and safety interlocking relationships between modules from the bottom layer. Specifically, they include two types: one is timing coordination constraints to guide cross-module relay operations, and the other is mutual exclusion interlocking constraints to prevent collisions or cross-contamination when multiple machines are running at the same time.
[0038] In a specific embodiment, step 120 can be implemented by the following steps: Based on environmental perception data and the task attributes of the target operation trajectory, a dynamic knowledge graph of equipment, environment and process under the current task is constructed. The graph reasoning algorithm is used to extract the process dependencies, spatial physical interference and chemical process mutual exclusion relationships between nodes in the graph. The process dependencies are transformed into temporal collaborative constraints, and the spatial physical interference and chemical process mutual exclusion relationships are transformed into mutual exclusion locking constraints. Using temporal coordination constraints and mutual exclusion locking constraints as boundary conditions, the running state machine model corresponding to each heterogeneous operation module is generated.
[0039] Specifically, based on the environmental perception data and the task attributes of the target operation trajectory input in step 110, the scheduling system constructs a dynamic knowledge graph of "machinery-environment-process" that reflects the current operation scenario.
[0040] The nodes in this graph consist of three types of entities: first, equipment nodes, such as weeding wheels, spray valves, and chassis; second, environmental nodes, such as rice plants, sudden mudslide areas, and crosswind direction; and third, agronomic process nodes, such as physical weed damage red lines and pesticide dilution thresholds. Nodes are connected by directed edges, which describe physical connections, relative spatial relationships, and chemical process influences. Through this structure, the graph semantically depicts the complete system boundary of the rice-growing robot operating on its current trajectory segment in real time.
[0041] After the graph is constructed, the system runs a preset graph reasoning algorithm to mine and extract three types of hidden underlying relationships between nodes in the graph, and transforms them into mathematical hard constraints that can be used by the state machine: Process dependency transformation: The system extracts the sequence of each module in agronomic processes. For example, weeding must be completed before spraying, otherwise the weeding wheel will shake off the pesticide. Graph reasoning algorithms automatically transform such process dependencies into temporal logic expressions with temporal coordination constraints.
[0042] Spatial physical interference transformation: Based on the vehicle attitude nodes and machine space nodes, the system infers whether there is a risk of 3D spatial envelope collision between two or more machines in their current position and action state. The graph inference algorithm transforms this risk into mutually exclusive locking constraints that prohibit the simultaneous occurrence of specific states.
[0043] Chemical process mutual exclusion transformation: The system infers the impact of the equipment medium on the external environment based on process nodes. For example, the turbid mud splashed up by the weeding module can instantly cover the leaf surface, causing the pesticide sprayed by the spraying module to fail because it cannot reach the leaf surface. After capturing such process mutual exclusions, the graph reasoning algorithm also transforms them into mutual exclusion locking constraints that require related modules to not operate simultaneously.
[0044] After completing the above reasoning, the system uses the obtained temporal coordination constraints and mutual exclusion locking constraints as boundary conditions for state transitions, i.e., the trigger thresholds or monitoring conditions for state jumps, and dynamically writes them into the underlying finite state machine templates of each heterogeneous operation module. In this way, each module obtains an instantiated running state machine model tailored to the current field operation environment.
[0045] Step 130: Obtain the actual travel speed from the operating sensor data, and combine it with the physical interval distance of each heterogeneous operation module on the vehicle body, operation status data, and constraint rules to make a collaborative scheduling judgment.
[0046] In this step, the system obtains the robot's actual travel speed from operational sensor data, combines this with the physical distances between the various heterogeneous operation modules on the vehicle body, the real-time operational status data fed back by each module, and the established temporal coordination constraints and mutual exclusion constraints, to make a unified judgment on the multi-module collaborative scheduling under the current working condition. Depending on the judgment objective, this step is divided into two parallel judgment branches: one is the temporal coordination branch, used to determine when to trigger sequential relay operations across modules; the other is the mutual exclusion branch, used to determine when it is necessary to forcibly shut down a module to avoid conflicts.
[0047] Branch 1: Timing Coordination Determination The purpose of this branch is to ensure that upstream and downstream modules with sequential dependencies achieve temporal coupling of their operations at the same ground coordinates. Regarding how to obtain accurate collaborative delay times, this solution provides two parallel technical approaches.
[0048] Option A: Physical correction method based on wheel slip compensation The soil conditions in paddy fields differ significantly from those in dry land. Due to prolonged soaking, the mud depth typically ranges from 10 to 30 centimeters, with high surface mud viscosity and low load-bearing capacity. When robots navigate in this environment, significant relative slippage occurs between the drive wheels or tracks and the mud surface, known as wheel-ground slippage. Specifically, the theoretical linear velocity corresponding to the actual rotational speed of the drive wheels is significantly higher than the actual travel speed of the vehicle relative to the ground. If the actual travel speed is used directly to calculate the coordination delay time between upstream and downstream work modules, the additional time lag caused by slippage will be ignored, resulting in the downstream module starting prematurely and the work position lagging behind the expected target point.
[0049] Therefore, before formally determining the cooperative mechanism, this scheme first extracts the drive wheel speed from the operational sensor data and compares the actual travel speed with the drive wheel speed. Because the mud in paddy fields is soft, the actual rotational speed of the drive wheels is often higher than the vehicle's actual travel speed relative to the ground; the difference reflects the current degree of slippage. Based on this, the system calculates a wheel-ground slippage correction coefficient, which is used to compensate for the initial spatial delay calculated from the actual travel speed and physical distance, resulting in a cooperative delay parameter that more closely reflects real-world conditions.
[0050] Therefore, before making a collaborative scheduling decision, the following process must be followed: The driving wheel speed of the rice robot is extracted from the operational sensor data. Based on the difference between the actual travel speed and the driving wheel speed, the wheel-ground slip correction coefficient, which characterizes the degree of wheel-ground slippage, is calculated. By using the wheel-ground slip correction coefficient, dynamic time compensation is performed on the initial spatial delay calculated from the actual travel speed and physical distance to obtain the cooperative delay parameter.
[0051] Specifically, the system synchronously extracts two sets of speed data from the operational sensor data: one is the theoretical linear velocity (i.e., drive wheel speed) obtained by converting the drive wheel rotation speed measured by the wheel speed encoder; the other is the actual travel speed calculated by RTK-GNSS Doppler velocimetry or visual odometry. By comparing the two frame by frame, the wheel-to-ground slip correction coefficient at the current moment can be expressed as:
[0052] in, The theoretical linear velocity converted from the driving wheel speed. This represents the actual speed of the vehicle relative to the ground. When When it equals 1, it indicates no slippage; when A value greater than 1 indicates slippage, and the larger the ratio, the more severe the slippage. In actual paddy field operations, this coefficient usually fluctuates between 1.05 and 1.30, and can reach over 1.50 in cases of severe subsidence.
[0053] Next, the system calculates the initial spatial delay based on the physical installation spacing of each module on the vehicle body. Taking the weeding module and the fertilizing module installed one after the other as an example, if the installation spacing between them along the direction of travel is D, and the current actual travel speed is... The initial spatial delay is:
[0054] This initial delay is the time elapsed from when the weeding module reaches a certain point to when the fertilizing module reaches the same point, assuming no slippage. However, due to slippage, the robot's actual movement efficiency is lower than the theoretical value; the time required for the fertilizing module to actually reach the target point is longer than the theoretical value. Longer. Therefore, the system uses a wheel-ground slip correction coefficient to dynamically compensate for the initial spatial delay in time:
[0055] Compensated Cooperative Delay Parameters By fully considering the slippage caused by the current mud depth and soil viscosity, this parameter more accurately reflects the actual time interval required for the downstream module to reach the target point previously worked by the upstream module from its current position. This parameter will serve as the basis for controlling the downstream module's startup delay in subsequent collaborative scheduling decisions, ensuring precise timing coupling between upstream and downstream modules at the same ground coordinates.
[0056] Determining the cooperative delay parameters The subsequent collaborative scheduling determination can be implemented through the following steps: Real-time monitoring of the task attributes of the current trajectory segment to determine whether cross-module collaborative operations are triggered; If triggered, the action status of the upstream operation module in the operation status data will be used as the starting reference. The starting timing of the downstream operation module will be delayed according to the coordination delay parameter so that the upstream and downstream operation modules with timing coordination constraints can achieve operation action timing coupling at the same ground coordinate. When the actual execution timing of the downstream job module is detected to be consistent with the expected timing after delay control, it is determined that the timing coordination constraint is satisfied.
[0057] Specifically, after obtaining the collaborative delay parameters, the system enters the actual collaborative determination phase. First, the system reads the task attribute labels on the current trajectory segment in real time to determine whether the current location requires joint operation of multiple modules. For example, if the label indicates that the area needs "weeding before fertilizing," the system uses the actual action state of the upstream module (such as the weeding module) as the start reference time and performs a countdown delay according to the previously calculated collaborative delay parameters. After the delay ends, the system sends a start command to the downstream module (such as the fertilizing module). The purpose of this is to use the time difference to compensate for the installation distance between the upstream and downstream modules on the vehicle body, so that the operation of the downstream module falls exactly at the same coordinate position on the ground where the upstream module has just performed its operation, achieving precise spatial docking between the upstream and downstream processes. When the system detects that the actual execution sequence of the downstream module matches the expected sequence preset by the delay control, it determines that the current working condition meets the timing collaboration constraints.
[0058] Option B: Prediction method based on time series forecasting models Considering that the depth and slippage of mud in paddy fields are constantly changing during operation, simple physical corrections may not be able to fully capture these nonlinear fluctuations. This solution adopts a data-driven approach, inputting the robot's historical operational data sequence accumulated during the current operation period, including the actual travel speed sequence, the drive wheel speed sequence, and the mud depth sequence extracted from environmental perception data, into a pre-trained time-series prediction model. This model models the nonlinear coupling relationship between speed, wheel speed, and mud depth, directly predicting the nonlinear time delay required for the robot to reach the target cooperative position from its current position, and outputting a cooperative delay prediction value.
[0059] In a specific implementation, scheme B can be achieved by the following steps: The historical operation data sequence of the rice robot during the current operation period is obtained. The historical operation data sequence includes the actual travel speed sequence, the drive wheel speed sequence, and the mud depth or subsidence degree sequence extracted from the environmental perception data. The historical operation data sequence is input into the preset time series prediction model. The time series prediction model is used to perform time series modeling on the nonlinear coupling relationship between the actual travel speed, drive wheel speed and mud foot depth, predict the nonlinear time delay experienced by the rice robot from the current position to the target cooperative position, and output the cooperative delay prediction value. Based on the predicted collaborative delay value, the startup timing of downstream operation modules with timing collaborative constraints is delayed and controlled. When it is determined that the actual execution timing of the downstream operation module is consistent with the expected timing determined based on the predicted collaborative delay value, the timing collaborative constraint is satisfied.
[0060] Specifically, this solution provides a time-series prediction-based path. This path no longer relies on the slippage coefficient at a specific moment, but instead treats historical operational data over a period of time as a whole sequence, allowing the model to learn the hidden patterns of change within it.
[0061] First, the system maintains a sliding time window, continuously caching historical operational data for the most recent few seconds within the current work period. This includes three sets of synchronously acquired time series: the first set is the actual travel speed sequence output by RTK-GNSS or visual odometry, reflecting the varying speed of the vehicle's movement relative to the ground; the second set is the drive wheel speed sequence output by the wheel speed encoder, reflecting the theoretical linear velocity of the drive wheels; and the third set is the mud depth or subsidence degree sequence extracted from environmental perception data, obtained by high-frequency ranging of the mud surface using lidar or depth cameras and then filtering. These three sets of sequences are aligned with the same timestamp to form the historical operational data sequence.
[0062] Next, the system inputs this historical operational data sequence into a pre-defined time-series prediction model. The model's input is a multi-dimensional time-series matrix, with each column corresponding to a variable (actual travel speed, drive wheel speed, and mud depth), and each row corresponding to a sampling time. The model performs time-series modeling of the nonlinear coupling relationship between these three variables. Specifically, the model learns how the drive wheel speed increases with a lag relative to the actual travel speed as the mud depth increases, and how this increase evolves over time. Based on the learned coupling pattern, the model predicts the nonlinear time delay required for the robot to reach the target cooperative position ahead, starting from its current position and in its current state, and outputs a predicted cooperative delay value.
[0063] Finally, the system uses this prediction value to control the startup timing of downstream operation modules. When the actual execution timing of the downstream module is detected to be consistent with the expected timing determined based on the prediction value, it is determined that the timing coordination constraint is satisfied.
[0064] In a preferred embodiment of this solution, the time-series prediction model employed is a Long Short-Term Memory (LSTM) network, a recurrent neural network structure specifically designed for processing time-series data. LSTM controls the retention and discarding of information through internal gating mechanisms—including a forget gate, an input gate, and an output gate. The forget gate determines which historical information needs to be forgotten, the input gate determines which new information needs to be written into memory, and the output gate determines which information is output at the current moment. This structure allows LSTM to remember key information over a relatively long time span, such as the continued impact of a deep mud area experienced tens of seconds ago on the current slippage state, while avoiding the gradient vanishing problem seen in ordinary recurrent neural networks. After inputting the three interdependent variables—actual travel speed, driving wheel speed, and mud depth—into the LSTM, the network can automatically extract the temporal dependencies between them and output predictions of future time delays.
[0065] Model training was completed offline. First, the robot was manipulated to navigate in paddy fields with varying mud depths and soil conditions, recording the actual travel speed, drive wheel speed, and mud depth sequences. The actual arrival times of preset ground markers were used as training labels. After collecting data covering various working conditions from slight to severe slippage, long sequences were divided into fixed-duration training samples using a sliding window, and the input sequences were normalized. During training, samples were input into the time-series prediction model in batches, using mean squared error as the loss function. Parameters were iteratively updated through backpropagation and gradient descent until convergence. Finally, the trained model parameters were solidified and deployed to the robot's edge computing terminal for online inference.
[0066] Branch 2: Mutual Exclusion Locking Determination The purpose of this branch is to promptly detect physical collision risks or cross-contamination risks when multiple modules are operating simultaneously, and to trigger interlocks accordingly. This solution also provides two parallel technical paths.
[0067] Option C: Physical determination method based on motion envelope and drift boundary calculation In a specific implementation, scheme C can be achieved by the following steps: Extract vehicle attitude change features from operational sensor data; Cross-calibrate the operational status data, actual travel speed, and vehicle attitude change characteristics, and combine them with physical interval distances to dynamically calculate the spatial motion envelope of each heterogeneous operational module and the drift boundary of the operational medium under the current driving state. The spatial motion envelope is used to determine whether there is spatial physical interference between modules, and the drift boundary, actual travel speed and physical interval distance are used to determine whether there is process cross-interference between modules. When it is determined that there is spatial physical interference or process cross-interference between heterogeneous operation modules, the mutual exclusion lockout constraint is satisfied.
[0068] Specifically, this solution is used to determine whether there is a risk of physical collision or process cross-interference when multiple modules are operating simultaneously. Its core idea is to use real-time vehicle attitude data and the operating status of each module to dynamically calculate the actual occupancy range and medium diffusion range of each module in three-dimensional space, and then determine whether to trigger interlocking through geometric comparison.
[0069] Specifically, the system uses the fixed installation coordinates of each module on the vehicle body as a reference, combined with the vehicle body attitude angles (roll angle, pitch angle, yaw angle) output in real time by the inertial measurement unit, and the current joint variables of each module's actuators (such as the lifting height of the weeding arm and the extension angle of the spray boom), to transform the mechanical end-effector positions of each module from the module's own coordinate system to the global coordinate system through coordinate transformation. As the vehicle body attitude fluctuates and the joints move in real time, the reachable range of the module's end-effector in space constantly changes, and the system uses the envelope boundary of the set of all possible end-effector position points as the spatial motion envelope of the module.
[0070] The dispersion boundary of the working medium refers to the farthest diffusion range that pesticide droplets or fertilizer particles can reach after being sprayed from the nozzle / fertilizer outlet, under the combined effects of vehicle speed, natural wind, their own spray velocity, and gravitational settling. Taking pesticide droplets as an example, the initial velocity of the droplets when they are sprayed from the nozzle is determined by the spraying pressure, and the nozzle installation height and spray angle are known modular structural parameters. The movement of the droplets in the air is affected by the combined influence of initial velocity, gravitational settling, air resistance, and the relative airflow (equivalent to lateral wind speed) brought about by the vehicle's movement. The farthest horizontal distance of droplet dispersion can be approximated as:
[0071] in, This represents the actual speed of travel. Natural wind speed, This is the time required for droplets to settle from the nozzle height to the crop canopy height due to gravity. Settling time can be estimated from the nozzle height and the terminal velocity of the droplets. When the drift boundary of the upstream module overlaps with the operating area of the downstream module, a risk of process cross-interference is identified.
[0072] After the calculation is completed, the system performs two judgments: First, based on the spatial motion envelopes of each module, it checks whether there are overlapping envelope regions; if so, it determines that there is a risk of spatial physical interference. Second, based on the drift boundary, actual travel speed, and physical interval distance, it checks whether the medium released by one module will drift into the working area of another module and cause pollution or interference; if so, it determines that there is a risk of process cross-interference. When either judgment is valid, it is determined that the current operating condition satisfies the mutual exclusion lockout constraint.
[0073] Option D: Prediction method based on conflict prediction model This solution uses actual travel speed, vehicle attitude change characteristics, physical interval distance, and operational status data of each module as input features, which are then fed into a pre-trained conflict prediction model. This model predicts the spatial trajectory and media dispersion range of each module within a future time window, outputting the probability of physical interference and process cross-interference between modules. When any risk probability exceeds a preset threshold, it is determined that physical interference or process cross-interference exists between the corresponding modules, and that mutual exclusion constraints are satisfied.
[0074] In a specific implementation, scheme D can be achieved by the following steps: Extract vehicle attitude change features from operational sensor data, and use actual travel speed, vehicle attitude change features, physical interval distance, and operational status data as input features; Input the input features into the preset conflict prediction model, and use the conflict prediction model to predict the spatial motion trajectory and media drift range of each heterogeneous operation module in the future time window, and output the probability of physical interference risk and the probability of process cross-interference risk between each heterogeneous operation module. When any risk probability exceeds a preset threshold, it is determined that there is physical interference or process cross-interference between the corresponding heterogeneous operation modules, and the mutual exclusion lockout constraint is satisfied.
[0075] Specifically, a conflict prediction model is used instead of geometric calculations to predict whether physical collisions or process interference will occur between modules in the future. In practice, the system extracts vehicle attitude change features from operational sensor data, and uses the actual travel speed, attitude change features, physical distances between modules on the vehicle, and the current operating status of each module as input features, feeding them into a pre-trained conflict prediction model. The model predicts the spatial trajectory and media dispersion range of each module within a future time window, outputting the probability of physical interference and process interference between modules. When any risk probability exceeds a preset safety threshold, the system determines that there is a conflict risk between the corresponding two modules, and that the current operating condition satisfies mutual exclusion constraints.
[0076] The conflict prediction model is used to predict in advance whether multiple heterogeneous operation modules will collide or cross-contaminate in subsequent actions. In a preferred embodiment of this solution, the model can adopt a spatiotemporal graph neural network structure. The basic idea is to abstract the spatial layout of each module on the vehicle body as nodes of a graph, and the physical distance and process coupling relationship between modules as edges of the graph. Through node update and edge message passing mechanisms, the model simultaneously considers the current motion trend of each module, changes in vehicle body attitude, and the release state of the operating medium to extrapolate the spatial occupancy of each node and the medium diffusion area at future moments, ultimately outputting the probability of conflict risk between each pair of modules.
[0077] Model training is completed offline. Training data is collected by manipulating the robot under different mud depths, turning radii, and wind directions. Each sample includes the robot's travel speed, attitude angle, module spacing, and operational status as input features, and manually labeled "collision occurred" and "cross-interference occurred" as supervision labels. During training, samples are input into the model in batches. The model parameters are iteratively updated using backpropagation and gradient descent by calculating the cross-entropy loss between the predicted risk probability and the true label until convergence. After training, the model parameters are fixed and deployed to the robot's edge computing terminal. During online inference, only one forward computation is needed to output the conflict risk probability of all module pairs.
[0078] Step 140: When it is determined that the timing coordination constraint is met, a timing coordination instruction is generated to control the timing coupling of the operation actions of each heterogeneous operation module at the target position; when it is determined that there is physical interference or process cross-interference between each heterogeneous operation module and the mutual exclusion lockout constraint is met, a module lockout instruction is generated to forcibly lock the target operation module.
[0079] Specifically, this step generates two types of scheduling instructions: When step 130 determines that the timing coordination constraint is met, a timing coordination instruction is generated. This instruction is used to control the timing coupling of the operation actions of upstream and downstream modules with process dependencies at the target ground location. For example, it ensures that the fertilization module accurately follows the same coordinate point as the weeding module.
[0080] When step 130 determines that there is physical interference or process cross-interference between the heterogeneous operation modules and the mutual exclusion lockout constraint is satisfied, a module lockout command is generated. This command is used to forcibly close the specified target operation module, such as closing the spraying module when turning to prevent pesticide drift, or locking the foliar fertilizer spraying module to prevent cross-contamination when the weeding wheel is rotating at high speed and stirring up mud and water.
[0081] Step 150: Input the timing coordination instruction or module locking instruction as a trigger event into the corresponding running state machine model, drive the corresponding running state machine model to dynamically jump between standby, running or locked states, and output the underlying control instruction according to the state after the jump to control each heterogeneous operation module to perform operations.
[0082] Specifically, the system uses the timing coordination command or module locking command generated in step 140 as a trigger event and inputs it into the corresponding running state machine model initialized in step 120. After receiving the trigger event, the state machine automatically completes dynamic transitions between standby, running, or locked states according to internally preset state transition rules. After the state transition is completed, the state machine outputs the corresponding underlying control command based on the current state, such as PWM speed control signals, solenoid valve switching quantities, or joint angle control quantities, to directly drive the actuators of various heterogeneous operation modules such as weeding, fertilizing, and spraying, thereby achieving precise timing coordination or emergency avoidance interlocking of multiple modules.
[0083] This application constructs a multi-task collaborative scheduling method based on hierarchical temporal sequence, deeply integrating environmental perception, trajectory planning, and state machine constraints based on temporal logic. This effectively solves the problem of inaccurate and conflict-prone multi-module collaboration under complex paddy field conditions. By introducing temporal collaborative constraints and mutual exclusion locking constraints, and combining real-time travel speed, attitude changes, and physical distances between modules for collaborative scheduling determination, heterogeneous modules such as weeding, fertilizing, and spraying can adaptively adjust their operation sequence according to the actual degree of slippage and vehicle dynamics, achieving precise relay operation of multiple modules at the same target location. Simultaneously, by real-time calculation of the spatial motion envelope and medium dispersion boundary, it can detect physical interference and process cross-interference risks in advance and execute forced locking, fundamentally avoiding module collisions and cross-contamination, significantly improving operational efficiency and agronomic safety, and reducing pesticide and fertilizer waste.
[0084] Based on the same inventive concept, as an implementation of the above-mentioned hierarchical timing-based multi-task collaborative scheduling method for rice robots, this application also provides a hierarchical timing-based multi-task collaborative scheduling system for rice robots. Figure 2 This is a structural diagram of the system in an embodiment of this application. See also... Figure 2 As shown, the system includes: The acquisition module is used to acquire environmental perception data, operational sensing data, and operational status data of each heterogeneous operation module of the rice robot, and to plan the target operation trajectory with task attributes accordingly. The initialization module is used to call and initialize the running state machine model corresponding to each heterogeneous operation module based on the task attributes of environmental perception data and target operation trajectory. The running state machine models are pre-set with constraint rules based on time logic, including time coordination constraints and mutual exclusion locking constraints. The collaborative scheduling determination module is used to obtain the actual travel speed from the operation sensor data, and combine it with the physical interval distance of each heterogeneous operation module on the vehicle body, operation status data, and constraint rules to make collaborative scheduling determination. The instruction generation module is used to generate timing coordination instructions when the timing coordination constraints are met, which are used to control the timing coupling of the operation actions of each heterogeneous operation module at the target position; when it is determined that there is physical interference or process cross-interference between the heterogeneous operation modules and the mutual exclusion lockout constraints are met, the module locking instruction is generated to forcibly lock the target operation module. The control module is used to input timing coordination instructions or module locking instructions as trigger events into the corresponding running state machine model, drive the corresponding running state machine model to dynamically jump between standby, running or locked states, and output low-level control instructions according to the state after the jump to control each heterogeneous operation module to perform operations.
[0085] This application constructs a multi-task collaborative scheduling method based on hierarchical temporal sequence, deeply integrating environmental perception, trajectory planning, and state machine constraints based on temporal logic. This effectively solves the problem of inaccurate and conflict-prone multi-module collaboration under complex paddy field conditions. By introducing temporal collaborative constraints and mutual exclusion locking constraints, and combining real-time travel speed, attitude changes, and physical distances between modules for collaborative scheduling determination, heterogeneous modules such as weeding, fertilizing, and spraying can adaptively adjust their operation sequence according to the actual degree of slippage and vehicle dynamics, achieving precise relay operation of multiple modules at the same target location. Simultaneously, by real-time calculation of the spatial motion envelope and medium dispersion boundary, it can detect physical interference and process cross-interference risks in advance and execute forced locking, fundamentally avoiding module collisions and cross-contamination, significantly improving operational efficiency and agronomic safety, and reducing pesticide and fertilizer waste.
[0086] This application also provides an electronic device, comprising: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a program for a hierarchical timing-based multi-task cooperative scheduling method for rice robots. When the processor executes the computer program, it implements all the steps in the aforementioned hierarchical timing-based multi-task cooperative scheduling method for rice robots, such as S110 to S150 and their sub-steps. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the aforementioned system embodiments.
[0087] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete this application. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.
[0088] The electronic device may be a computing device such as a server or cloud server deployed in a hierarchical timing-based rice robot multi-task collaborative scheduling system. The electronic device may include, but is not limited to, processors and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include data acquisition cards and output cards for connecting sensors and actuators, network communication devices, buses, etc.
[0089] The processor can be a central processing unit, or other general-purpose processors, digital signal processors, application-specific integrated circuits (ASICs), off-the-shelf programmable gate arrays (OPGs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting various parts of the entire electronic device via various interfaces and lines.
[0090] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function, etc.; the data storage area may store data created according to system usage, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart memory card, secure digital card, flash memory card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.
[0091] Wherein, if the modules / units integrated in the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory, a random access memory, an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.
[0092] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.
Claims
1. A multi-task cooperative scheduling method for rice robots based on hierarchical temporal sequence, characterized in that, The method includes: Acquire environmental perception data, operational sensing data, and operational status data of each heterogeneous operation module of the rice robot, and plan the target operation trajectory with task attributes accordingly. Based on the environmental perception data and the task attributes of the target operation trajectory, the running state machine model corresponding to each heterogeneous operation module is called and initialized. The running state machine models are pre-set with constraint rules based on time-series logic, including time-series coordination constraints and mutual exclusion locking constraints. The actual travel speed is obtained from the operation sensor data, and coordinated scheduling is determined by combining the physical interval distance of each heterogeneous operation module on the vehicle body, the operation status data, and the constraint rules. When it is determined that the timing coordination constraint is satisfied, a timing coordination instruction is generated to control the timing coupling of the operation actions of each heterogeneous operation module at the target position; when it is determined that there is physical interference or process cross-interference between the heterogeneous operation modules and the mutual exclusion lockout constraint is satisfied, a module lockout instruction is generated to forcibly lock the target operation module. The timing coordination instruction or the module locking instruction is used as a trigger event and input into the corresponding running state machine model to drive the corresponding running state machine model to dynamically switch between standby, running or locked states, and output the underlying control instruction according to the state after the switch to control each heterogeneous operation module to perform operations.
2. The method according to claim 1, characterized in that, The planned target operation trajectory with task attributes includes: Obtain the pre-stored field geographic information and spatial distribution map of operation parameters; The real-time acquired environmental perception data is spatiotemporally aligned and fused with the spatial distribution map of the operation parameters to identify the current crop growth and weed distribution density in the field. Using the physical boundaries defined by the geographic information of the field as constraints, a target operation trajectory is generated, and operation behavior tags are generated based on the weed distribution density and the crop growth. The operation behavior tags are then mapped to the corresponding spatial locations of the target operation trajectory as task attributes.
3. The method according to claim 1, characterized in that, The step of calling and initializing the running state machine model corresponding to each heterogeneous task module based on the environmental perception data and the target operation trajectory includes: Based on the environmental perception data and the task attributes of the target operation trajectory, a dynamic knowledge graph of equipment, environment and process under the current task is constructed. The process dependencies, spatial physical interferences, and chemical process mutual exclusions between nodes in the graph are extracted using a graph reasoning algorithm. The process dependencies are transformed into the temporal coordination constraints, and the spatial physical interferences and chemical process mutual exclusions are transformed into the mutual exclusion locking constraints. Using the temporal coordination constraints and the mutual exclusion locking constraints as boundary conditions, a running state machine model corresponding to each heterogeneous operation module is generated.
4. The method according to claim 1, characterized in that, Before performing the collaborative scheduling determination, the method further includes: The driving wheel speed of the rice robot is extracted from the operational sensor data, and a wheel-ground slip correction coefficient, which characterizes the degree of wheel-ground slippage, is calculated based on the difference between the actual travel speed and the driving wheel speed. Using the wheel-ground slip correction coefficient, dynamic time compensation is performed on the initial spatial delay calculated from the actual travel speed and the physical interval distance to obtain the cooperative delay parameter.
5. The method according to claim 4, characterized in that, The step of obtaining the actual travel speed from the operational sensor data, and combining it with the physical intervals between the heterogeneous operation modules on the vehicle body, the operation status data, and the constraint rules to perform collaborative scheduling determination includes: Real-time monitoring of the task attributes of the current trajectory segment to determine whether cross-module collaborative operations are triggered; If triggered, the action status of the upstream operation module in the operation status data will be used as the starting reference, and the starting sequence of the downstream operation module will be delayed according to the coordination delay parameter so that the upstream and downstream operation modules with timing coordination constraints can achieve operation action timing coupling at the same ground coordinate. When the actual execution timing of the downstream operation module is detected to be consistent with the expected timing after the delay control, it is determined that the timing coordination constraint is satisfied.
6. The method according to claim 1, characterized in that, The step of obtaining the actual travel speed from the operational sensor data, and combining it with the physical intervals between the heterogeneous operation modules on the vehicle body, the operation status data, and the constraint rules to perform collaborative scheduling determination includes: The historical operation data sequence of the rice robot during the current working period is obtained. The historical operation data sequence includes the actual travel speed sequence, the drive wheel speed sequence, and the mud depth or subsidence degree sequence extracted from the environmental perception data. The historical operation data sequence is input into a preset time-series prediction model. The time-series prediction model is used to perform time-series modeling on the nonlinear coupling relationship between the actual travel speed, drive wheel speed and mud depth, predict the nonlinear time delay experienced by the rice robot from the current position to the target cooperative position, and output the cooperative delay prediction value. Based on the predicted collaborative delay value, the startup timing of downstream operation modules with timing collaborative constraints is delayed and controlled. When it is determined that the actual execution timing of the downstream operation module is consistent with the expected timing determined based on the predicted collaborative delay value, it is determined that the timing collaborative constraint is satisfied.
7. The method according to claim 1, characterized in that, The step of obtaining the actual travel speed from the operational sensor data, and combining it with the physical intervals between the heterogeneous operation modules on the vehicle body, the operation status data, and the constraint rules to perform collaborative scheduling determination includes: Extract vehicle attitude change features from the operational sensor data; The operation status data, the actual travel speed, and the vehicle attitude change characteristics are cross-calibrated and combined with the physical interval distance to dynamically calculate the spatial motion envelope of each heterogeneous operation module and the drift boundary of the operation medium under the current driving state. Based on the spatial motion envelope, it is determined whether there is spatial physical interference between each module, and based on the drift boundary, the actual travel speed and the physical interval distance, it is determined whether there is process cross-interference between each module; When it is determined that there is spatial physical interference or process cross-interference between the heterogeneous operation modules, the mutual exclusion lockout constraint is satisfied.
8. The method according to claim 1, characterized in that, The step of obtaining the actual travel speed from the operational sensor data, and combining it with the physical intervals between the heterogeneous operation modules on the vehicle body, the operation status data, and the constraint rules to perform collaborative scheduling determination includes: The vehicle attitude change features are extracted from the operational sensor data, and the actual travel speed, the vehicle attitude change features, the physical interval distance, and the operation status data are used as input features. The input features are input into a preset conflict prediction model. The conflict prediction model is used to predict the spatial motion trajectory and media drift range of each heterogeneous operation module in the future time window, and output the probability of physical interference risk and the probability of process cross-interference risk between each heterogeneous operation module. When any risk probability exceeds a preset threshold, it is determined that there is physical interference or process cross-interference between the corresponding heterogeneous operation modules, and the mutual exclusion lockout constraint is satisfied.
9. A multi-task collaborative scheduling system for rice robots based on hierarchical temporal sequence, characterized in that, The system includes: The acquisition module is used to acquire environmental perception data, operational sensing data, and operational status data of each heterogeneous operation module of the rice robot, and to plan the target operation trajectory with task attributes accordingly. An initialization module is used to call and initialize the running state machine model corresponding to each heterogeneous operation module based on the environmental perception data and the task attributes of the target operation trajectory. The running state machine models are pre-set with constraint rules based on time-series logic, including time-series coordination constraints and mutual exclusion locking constraints. The collaborative scheduling determination module is used to obtain the actual travel speed from the operation sensor data, and combine it with the physical interval distance of each heterogeneous operation module on the vehicle body, the operation status data, and the constraint rules to make collaborative scheduling determination. The instruction generation module is used to generate a timing coordination instruction when it is determined that the timing coordination constraint is satisfied, in order to control the timing coupling of the operation actions of each heterogeneous operation module at the target position; when it is determined that there is physical interference or process cross-interference between the heterogeneous operation modules and the mutual exclusion lockout constraint is satisfied, the module lockout instruction is generated to forcibly lock the target operation module. The control operation module is used to input the timing coordination instruction or the module lock instruction as a trigger event into the corresponding running state machine model, drive the corresponding running state machine model to dynamically jump between standby, running or locked states, and output the underlying control instruction according to the state after the jump to control each heterogeneous operation module to perform operations.
