Orchard autonomous robot collaborative operation method and system based on dynamic role allocation

Abstract

The invention relates to the field of robots, in particular to an orchard autonomous robot collaborative operation method and system based on dynamic role allocation. The method comprises the following steps: acquiring an orchard plane operation panoramic data set, performing two-dimensional plane operation feature extraction and modeling according to the orchard plane operation panoramic data set, further performing robot capability matching and initial role allocation decision under a two-dimensional plane, and generating a basic role scheduling scheme set; obtaining an orchard operation disturbance factor set, and based on the basic role scheduling scheme set and the orchard operation disturbance factor set, performing three-dimensional space conflict resolution and global collaborative optimization to generate an optimized role scheduling collaborative instruction set; and based on the optimized role scheduling collaborative instruction set, executing a role switching instruction and collaborative operation control facing heterogeneous robot individuals, and generating a multi-robot collaborative operation scheduling report. In the collaborative operation process of the robot, the orchard operation efficiency and stability are greatly improved.

Description

Technical Field

[0001] This application relates to the field of robotics, and in particular to a collaborative operation method and system for autonomous robots in orchards based on dynamic role assignment. Background Technology

[0002] In the field of smart agriculture and precision orchard operations, autonomous robot systems have gradually become key equipment for completing complex tasks such as fruit picking, plant protection spraying, and material transportation. Multi-robot collaborative operations have core value in improving the management efficiency of large-scale orchards, reducing labor costs, and coping with agricultural time constraints, and are an important direction for the intelligent transformation of modern agriculture.

[0003] Existing orchard robot operation scheduling methods are mostly based on two-dimensional planar planning, lacking collaborative modeling and real-time response capabilities for multiple factors of fruit trees. This results in low robot operation efficiency, collision risks, and an inability to dynamically adjust roles and perform global collaborative optimization based on real-time environmental and task changes, which seriously restricts the system's reliability, adaptability, and overall operational efficiency. Summary of the Invention

[0004] This application provides a method and system for collaborative operation of autonomous robots in orchards based on dynamic role allocation, in order to solve the above-mentioned technical problems.

[0005] In a first aspect, this application provides a method for collaborative operation of autonomous robots in orchards based on dynamic role allocation. The method includes: acquiring a panoramic dataset of orchard operations on a planar surface; extracting and modeling two-dimensional planar operation features based on the panoramic dataset, and then performing robot capability matching and initial role allocation decisions in the two-dimensional plane to generate a basic role scheduling scheme set; acquiring a set of orchard operation disturbance factors; performing three-dimensional spatial conflict resolution and global collaborative optimization based on the basic role scheduling scheme set and the orchard operation disturbance factors set to generate an optimized role scheduling collaborative instruction set; and executing role switching instructions and collaborative operation control for heterogeneous robot individuals based on the optimized role scheduling collaborative instruction set to generate a multi-robot collaborative operation scheduling report.

[0006] By precisely matching orchard operations with robot capabilities, the above technical solutions can flexibly address issues such as sudden environmental changes and equipment malfunctions, avoid operational conflicts and resource waste, reduce fruit damage and cost losses, significantly improve orchard operation efficiency and stability, provide data support for managers, and facilitate the implementation of intelligent and refined collaborative operations in orchards.

[0007] Optionally, the generation of the basic role scheduling scheme set includes: the orchard planar operation panoramic dataset includes orchard two-dimensional digital map data, robot body state data, and operation task feature data; based on the operation task feature data, the urgency and priority of tasks are quantitatively calculated to generate an orchard task sequence with time constraints; combined with the orchard two-dimensional digital map data, the orchard task sequence with time constraints is preprocessed by geographic location-based clustering and path planning to generate optimized task groups with expected execution paths and energy consumption estimates; based on the robot body state data, the tasks in the optimized task groups are matched to robots with matching states and assigned initial operation roles through a dynamic allocation method based on rules and filtering mechanisms to generate the basic role scheduling scheme set.

[0008] Optionally, the dynamic allocation method includes: establishing a dynamic capability profile for each robot based on the remaining battery power, current position, and types of loaded materials and equipment in the robot's body status data; matching the dynamic capability profile for each optimized task group according to its operation type, expected path complexity, and total energy consumption estimate, wherein the dynamic capability profile includes at least: operation capability specialization matching and battery life safety redundancy; initiating internal matching from qualified robot candidates, and allocating tasks to the best matched robot based on the estimated time and energy consumption cost of completing the task; assigning an initial operation role to the winning robot according to the operation type corresponding to the assigned task, including but not limited to fixed-point harvester, row sprayer, cross-regional transporter, or composite inspector.

[0009] Optionally, the step of generating the optimized role scheduling collaborative instruction set includes: constructing an orchard operation environment voxel map reflecting the three-dimensional canopy structure, terrain undulation, and obstacle distribution of fruit trees based on the orchard planar operation panoramic dataset and the orchard operation disturbance factor set; predicting the spatiotemporal region required for each robot assigned to a task group to perform its task based on the basic role scheduling scheme set and the orchard operation environment voxel map; performing conflict detection and pre-detection based on three-dimensional space and time dimensions in the set of spatiotemporal regions of all robots to identify conflict region sets with spatial overlap or time window competition; and initiating dynamic avoidance negotiation between robots with the goal of minimizing global operation delay for each conflict region set, and generating the optimized role scheduling collaborative instruction set.

[0010] Optionally, the process of constructing the orchard operation environment voxel map includes: the orchard operation disturbance factor set includes real-time weather data, crop 3D planting data, and obstacle 3D spatial data; discretizing the orchard operation space into a uniform-sized 3D cubic grid based on the horizontal range of the orchard 2D digital map and the vertical planting structure defined by the crop 3D planting data, with each grid cell defined as an orchard basic voxel; assigning a comprehensive attribute label to each orchard basic voxel, the comprehensive attribute label including: a vertical planting layer attribute defined based on the crop 3D planting data, and a spatial occupancy status attribute determined based on the obstacle 3D spatial data; and performing a dynamic impact assessment on the orchard basic voxels based on the real-time weather data to generate the orchard operation environment voxel map for robot 3D spatial operation planning.

[0011] Optionally, the prediction process for the spatiotemporal region includes: before matching robots to the optimized task group, upgrading its task requirements to three-dimensional task requirements that include a clearly defined required vertical working space; upgrading the robot's dynamic capability profile to three-dimensional physical capabilities including its maximum working height, robotic arm working envelope, and body foldability; based on the orchard working environment voxel map, performing spatiotemporal conflict prediction, and searching for other qualified robots with matched capabilities in the same spatiotemporal region within the same time period; if a spatial conflict is found where regions intersect, the robot's application is not immediately rejected, but a dynamic avoidance negotiation strategy is executed as a prerequisite for the final task allocation and instruction generation; the dynamic avoidance negotiation strategy includes a conflict resolution sub-strategy based on spatiotemporal resource scheduling and a collaborative optimization sub-strategy based on dynamic reorganization of tasks and roles.

[0012] Optionally, the conflict resolution sub-strategy includes: evaluating and implementing a time-domain delay scheme, that is, under the premise of satisfying the urgency constraints of both parties in the conflicting task, guiding one robot role to temporarily switch to a waiter, and adjusting its task time window to avoid space occupation; evaluating and implementing a spatial domain reconstruction scheme, that is, based on the voxel map of the orchard operation environment, guiding one robot role to temporarily switch to a detour explorer, and planning and allocating a three-dimensional alternative path that can avoid the current spatial conflict.

[0013] Optionally, the collaborative optimization sub-strategy includes: performing task and role decoupling and reorganization, that is, analyzing the operational compatibility between conflicting tasks; if they are determined to be mutually exclusive, dynamically inserting region closure instructions or adjusting operational parameters to isolate the impact; if they are determined to be complementary, dynamically adjusting the operational role or parameters of one robot to collaboratively utilize the spatiotemporal resources; and performing cross-robot role fusion, that is, in extremely narrow or high-collaboration-requirement scenarios, merging two conflicting robots and their tasks, forming a temporary collaborative unit with composite functions through operational logic combination, and allocating a unified spatiotemporal pipeline and collaborative instructions to this unit.

[0014] Optionally, generating the multi-robot collaborative operation scheduling report includes: summarizing the final execution sequence of all robots, actual spatiotemporal trajectories, dynamic negotiation events, and task completion status based on the execution feedback of the optimized role scheduling collaborative instruction set; evaluating the overall timeliness, energy consumption, and conflict resolution effectiveness of the collaborative operation based on the orchard operation disturbance factor set; and generating the multi-robot collaborative operation scheduling report containing optimization suggestions for this round of operations and the initial role scheduling pre-configuration for the next cycle based on the comprehensive feedback data and evaluation results.

[0015] Secondly, this application provides a collaborative operation system for autonomous robots in orchards based on dynamic role allocation, the system comprising:

[0016] The basic role analysis module is used to acquire a panoramic dataset of orchard operations on a flat plane. Based on this dataset, it extracts and models two-dimensional planar operation features, then performs robot capability matching and initial role allocation decisions in the two-dimensional plane, generating a basic role scheduling scheme set. The optimized role analysis module is used to acquire a set of orchard operation disturbance factors. Based on the basic role scheduling scheme set and the orchard operation disturbance factor set, it performs three-dimensional spatial conflict resolution and global collaborative optimization, generating an optimized role scheduling collaborative instruction set. The scheduling report generation module is used to execute role switching instructions and collaborative operation control for heterogeneous robot individuals based on the optimized role scheduling collaborative instruction set, generating a multi-robot collaborative operation scheduling report. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of an application scenario provided in an embodiment of this application;

[0019] Figure 2 A flowchart illustrating a collaborative operation method for autonomous robots in an orchard based on dynamic role allocation, provided in an embodiment of this application;

[0020] Figure 3 This is a schematic diagram of the structure of an orchard autonomous robot collaborative operation system based on dynamic role allocation, provided as an embodiment of this application. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0022] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article, unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship.

[0023] The embodiments of this application will now be described in further detail with reference to the accompanying drawings.

[0024] Existing orchard robot operation scheduling methods are mostly based on two-dimensional planar planning, lacking collaborative modeling and real-time response capabilities for multiple factors of fruit trees. This results in low robot operation efficiency, collision risks, and an inability to dynamically adjust roles and perform global collaborative optimization based on real-time environmental and task changes, which seriously restricts the system's reliability, adaptability, and overall operational efficiency.

[0025] Based on this, this application provides a method and system for collaborative operation of autonomous robots in orchards based on dynamic role allocation. First, a panoramic dataset of orchard operations is acquired. Through 2D planar operation feature extraction and modeling, and combined with the capabilities of heterogeneous robots, matching is performed to formulate initial role allocation decisions and generate a basic role scheduling scheme set. Next, a set of orchard operation disturbance factors is collected. Based on the basic scheme, 3D spatial conflict resolution and global collaborative optimization are carried out to form an optimized role scheduling collaborative instruction set. Finally, according to this instruction set, role switching and collaborative operation control for each heterogeneous robot are executed, operation data is recorded in real time, a multi-robot collaborative operation scheduling report is generated and output to orchard staff, adapting to the needs of orchard operation scenarios throughout the process. By accurately matching orchard operations with robot capabilities, it flexibly responds to problems such as sudden environmental changes and equipment failures, avoids operational conflicts and resource waste, reduces fruit damage and cost losses, significantly improves orchard operation efficiency and stability, provides data support for managers, and facilitates the implementation of intelligent and refined collaborative operations in orchards.

[0026] Figure 1 This is a schematic diagram illustrating an application scenario provided by this application. In the process of robot collaborative operation, the method provided in this application significantly improves the efficiency and stability of orchard operations, facilitating the implementation of intelligent and refined collaborative operations in orchards.

[0027] Specifically, the method of this application is applied to any server that communicates with multiple types of acquisition devices and a multi-dimensional sensor network. The server acquires a panoramic dataset of orchard operations from the multiple types of acquisition devices and a set of orchard operation disturbance factors from the multi-dimensional sensor network. First, the panoramic dataset of orchard operations is acquired. Then, through two-dimensional planar operation feature extraction and modeling, and combined with the capabilities of heterogeneous robots, matching is completed, initial role allocation decisions are made, and a basic role scheduling scheme set is generated. Next, the orchard operation disturbance factor set is acquired. Based on the basic scheme, three-dimensional spatial conflict resolution and global collaborative optimization are carried out to form an optimized role scheduling collaborative instruction set. Finally, based on this instruction set, role switching and collaborative operation control for each heterogeneous robot are executed, operation data is recorded in real time, a multi-robot collaborative operation scheduling report is generated and output to orchard staff, fully adapting to the needs of orchard operation scenarios.

[0028] For specific implementation details, please refer to the following examples.

[0029] Figure 2 This is a flowchart illustrating a collaborative operation method for autonomous robots in an orchard based on dynamic role allocation, as provided in one embodiment of this application. The method of this embodiment can be applied to the server in the above scenario. Figure 2 As shown, the method includes:

[0030] S201. Obtain the orchard planar operation panoramic dataset. Based on the orchard planar operation panoramic dataset, perform two-dimensional planar operation feature extraction and modeling, and then perform robot capability matching and initial role allocation decision in the two-dimensional plane to generate a basic role scheduling scheme set.

[0031] The orchard planar operation panoramic dataset is a comprehensive data set that fully reflects the two-dimensional planar operation environment and operational requirements of an orchard. It includes orchard two-dimensional digital map data, robot body status data, and operational task feature data, sourced from various types of acquisition devices deployed within the orchard. Two-dimensional planar operation feature extraction and modeling involves extracting key feature information influencing robot task allocation from the orchard planar operation panoramic dataset. Based on these extracted key features, a visualized and structured orchard two-dimensional operation environment model is constructed, clearly presenting the environmental constraints and operational requirements of each work area. Robot capability matching is the process of accurately adapting the functions and performance parameters of each robot participating in the orchard operation to the operational task requirements in the two-dimensional planar operation environment model. Initial role allocation decision-making is the decision-making process of formulating role allocation rules and determining the initial work role for each robot based on the two-dimensional planar operation feature model and the robot capability matching results. The basic role scheduling scheme set is a standardized set of schemes containing information such as the initial work role of each robot, the corresponding work area, the specific work task, the work sequence arrangement, and the basic movement path.

[0032] Specifically, traditional orchard robot operations often adopt a "single-role fixed execution" model, lacking precise adaptation to the orchard environment and robot capabilities. The drawbacks are particularly prominent in real-world scenarios. For example, during the apple orchard ripening season, having irrigation robots, which are only adapted to low-precision operations, perform picking tasks will lead to a significant increase in fruit damage from bumps and knocks. In the citrus orchard seedling area, having a 1-ton transport robot perform irrigation work will not only waste energy but may also crush the seedlings due to its excessive weight. At the same time, without a panoramic dataset, it is impossible to know the distribution of obstacles such as irrigation pipes and tree trunks in the peach orchard, and the robot may be delayed due to path obstruction. Without feature modeling, it is difficult to distinguish the planting density of different areas in a strawberry orchard, resulting in the fertilization robot applying fertilizer evenly, causing fertilizer burn in the seedling area and insufficient fertility in the fruiting area. In addition, without an initial role allocation plan, multiple robots may simultaneously flood into a certain picking area in a pear orchard, causing congestion, while the spraying task in another area is left unattended. This step integrates aerial photographs, GIS maps, and other data to construct a semantically labeled 2D digital map of the orchard. The algorithm automatically identifies key features such as fruit tree locations and road networks. Subsequently, based on the robot's capability list (such as size and dexterity), the matching degree for each work area is calculated. Through multi-objective optimization, one or more basic role scheduling schemes are generated, clarifying the initial tasks, positions, and sequences of each robot. This transforms chaotic static information into a structured operational blueprint, achieving precise pre-matching of robot expertise with the heterogeneous needs of the orchard. This effectively avoids planning blind spots and resource mismatches caused by complex spatial structures, ensuring that operations are established on an efficient and professional basis from the outset. It provides a clear and reliable optimization starting point for subsequent responses to dynamic changes, significantly improving the rationality and initial efficiency of the overall system planning.

[0033] S202. Obtain the orchard operation disturbance factor set. Based on the basic role scheduling scheme set and the orchard operation disturbance factor set, perform three-dimensional spatial conflict resolution and global collaborative optimization to generate an optimized role scheduling collaborative instruction set.

[0034] The orchard operation disturbance factor set can be a collection of various factors that may affect robot operation and coordination during orchard operations, including real-time weather data, 3D crop planting data, and 3D obstacle spatial data. This data originates from a multi-dimensional sensor network deployed in the orchard. 3D spatial conflict resolution addresses potential conflicts such as path intersections, overlapping work areas, and motion interference that may occur when multiple robots operate in 3D space. Global collaborative optimization aims to maximize overall orchard operation efficiency by combining a basic role scheduling scheme set with the orchard operation disturbance factor set to comprehensively adjust and optimize each robot, ensuring efficient collaboration in a dynamic environment. The optimized role scheduling and collaborative instruction set is a set of instructions generated after 3D spatial conflict resolution and global collaborative optimization, containing information such as the adjusted role, task, execution time, movement route, and collaborative requirements for each robot.

[0035] Specifically, the orchard operating environment is dynamic and ever-changing. The initial role scheduling plan cannot cope with various emergencies. The lack of a dynamic optimization mechanism will lead to operation interruption or a sharp drop in efficiency. For example, if a sudden rainstorm hits a cherry orchard, and the picking robot continues to operate according to the initial plan, not only will the fruit get wet and spoil, but the robot's electronic components may also short-circuit and be damaged. If a large-scale aphid infestation occurs in a citrus orchard, and the plan for scheduling the spraying robot is not adjusted in time, the pests and diseases will spread rapidly throughout the orchard. Robot malfunctions are also frequent. For example, if the fertilization robot in an apple orchard suddenly runs out of battery power and cannot complete the fertilization task, the fruit trees will suffer from nutrient deficiency and be affected in their growth if the role is not reassigned. At the same time, multiple robots operating can easily cause spatial conflicts: the movement path of the transport robot in the peach orchard overlaps with the working range of the spraying robot, and the mechanical arm of the picking robot in the strawberry orchard interferes with the height of the spray nozzle of the irrigation robot. If these conflicts are not resolved, they may cause equipment collisions or failure of operation. This step continuously monitors dynamic information flows from environmental sensors and robot status feedback. Once a critical disturbance is detected, optimization is triggered. The algorithm uses the current execution base plan and real-time status as input to predict conflicts in a simulated three-dimensional spacetime. It then uses a collaborative decision-making algorithm to search for a new scheduling scheme that can eliminate conflicts and optimize the global objective, ultimately compiling it into precise collaborative instructions. This endows the system with strong environmental adaptability and anti-interference resilience. Through real-time three-dimensional conflict resolution and global collaborative optimization, the system can safely and smoothly cope with various sudden disturbances (such as machine failures and weather changes), minimizing the impact. This ensures that the robot team can not only "work according to plan" but also "intelligently respond to changes," achieving a leap from static scheduling to dynamic intelligent scheduling and greatly improving the robustness and overall execution efficiency of the operating system.

[0036] S203. Based on the optimized role scheduling and collaborative instruction set, execute role switching instructions and collaborative operation control for heterogeneous robot individuals, and generate a multi-robot collaborative operation scheduling report.

[0037] Role switching commands can be specific instructions generated by the central control system for each heterogeneous robot, based on an optimized role scheduling and coordination command set, to adjust its operational role and working status. Collaborative operation control involves the central control system acquiring the operational status of each heterogeneous robot in real time via a communication network, and dynamically coordinating and controlling the robots' actions based on the optimized role scheduling and coordination command set, ensuring efficient cooperation and conflict-free operation among multiple robots. A multi-robot collaborative operation scheduling report can be a comprehensive report recording key data and results throughout the entire multi-robot collaborative operation process.

[0038] Specifically, the successful implementation of optimized scheduling schemes relies on precise execution and collaborative control, which is particularly important in orchards with intensive heterogeneous robot operations. Orchards employ a variety of robot types; for example, apple orchards use Modbus control robots, while pear orchards use CANopen. If a uniform set of role-switching commands is directly issued, the robots cannot recognize them, leading to situations where, in cherry orchards, spraying robots cannot promptly switch from their transport roles when pests or diseases occur. Furthermore, a lack of collaborative control can result in operational disjointedness: for instance, in peach orchards, after a picking robot completes harvesting in a certain area, it may fail to promptly notify the transport robot to transfer the fruit, causing the fruit to accumulate and rot; in vineyards, pruning robots and irrigation robots may have conflicting operating sequences, leading to irrigation before the pruned branches are cleared, potentially causing mold growth. Additionally, without scheduling reports, managers cannot access data on the progress, pesticide consumption, and frequency of malfunctions of each robot in strawberry orchards, making it difficult to assess operational effectiveness and equipment status, which hinders subsequent scheme optimization. This step compiles and distributes the optimized global instructions to each heterogeneous robot. Upon receiving the instructions, the robots load the corresponding parameters and execute them. Simultaneously, the background automatically records all logs, and after the operation, analyzes and generates a scheduling report covering indicators such as task completion, role switching records, and disturbance handling. The generated digital scheduling report not only achieves transparent management of the operation process but also provides valuable data insights for analyzing efficiency bottlenecks, optimizing robot configurations, and developing long-term operational strategies, completing a closed-loop decision-making process from automating single operations to continuously intelligentizing farm management.

[0039] The method provided in this embodiment first acquires a panoramic dataset of orchard operations on a planar scale. Through 2D planar operation feature extraction and modeling, and combined with the capabilities of heterogeneous robots, matching is achieved, initial role allocation decisions are made, and a basic role scheduling scheme set is generated. Next, a set of orchard operation disturbance factors is collected. Based on the basic scheme, 3D spatial conflict resolution and global collaborative optimization are carried out to form an optimized role scheduling collaborative instruction set. Finally, according to this instruction set, role switching and collaborative operation control for each heterogeneous robot are executed, operation data is recorded in real time, a multi-robot collaborative operation scheduling report is generated and output to orchard staff, fully adapting to the needs of orchard operation scenarios. By accurately matching orchard operations with robot capabilities, it flexibly responds to problems such as sudden environmental changes and equipment failures, avoiding operational conflicts and resource waste, reducing fruit damage and cost losses, significantly improving orchard operation efficiency and stability, providing data support for managers, and facilitating the implementation of intelligent and refined collaborative operations in orchards.

[0040] In some embodiments, the orchard planar operation panoramic dataset includes orchard two-dimensional digital map data, robot body state data, and operation task feature data. Based on the operation task feature data, the urgency and priority of tasks are quantitatively calculated to generate an orchard task sequence with time constraints. Combining the orchard two-dimensional digital map data, the orchard task sequence with time constraints is preprocessed by geographic location-based clustering and path planning to generate optimized task groups with expected execution paths and energy consumption estimates. Based on the robot body state data, the tasks in the optimized task groups are matched to robots with matching states and assigned initial operation roles through a rule-based and filtering mechanism-based dynamic allocation method, generating a basic role scheduling scheme set.

[0041] Two-dimensional digital map data for orchards can be structured data presenting the orchard's geographical environment in a two-dimensional plane, covering information such as orchard boundaries, crop planting area divisions (e.g., apple or citrus planting areas), work path widths, coordinates of fixed obstacles (tree trunks, irrigation pipes, electrical boxes), and key work points (harvesting assembly points, material resupply points). Robot status data can be a set of real-time status parameters fed back by each heterogeneous robot participating in the operation, including remaining battery power, current geographical location, quantity of loaded materials (e.g., remaining pesticide solution for a spraying robot), type of mounted equipment (e.g., harvesting robotic arm), and equipment operating status (e.g., whether spray pressure is stable). Task characteristic data can be a set of information characterizing the core attributes of various orchard tasks, including task type (harvesting, spraying, fertilizing, transportation, inspection, etc.), work area range, task requirements (e.g., fruit maturity standards for harvesting), task initiation time, and associated crop growth status (e.g., whether it is in a high-incidence period of pests and diseases). The orchard task sequence can be a list of tasks with a defined execution order and latest completion deadline, after quantification of urgency and priority. Each task is marked with corresponding time constraints (e.g., "Complete the picking of Fuji apples in area A before 18:00 tomorrow"). Optimized task groups can be sets of tasks formed after geographic clustering and path planning preprocessing. Each group contains multiple spatially adjacent tasks, with defined expected execution paths and energy consumption estimates, facilitating the matching of suitable robots.

[0042] Specifically, traditional orchard robot scheduling often uses static or polling allocation, ignoring the spatiotemporal characteristics of tasks and the dynamic changes in the robot's real-time status, resulting in low operational efficiency. For example, a fixed-allocation picking robot may stop midway due to insufficient power (e.g., only 25% remaining), or delay the optimal operation window (e.g., treatment within 2 hours before the spread of pests and diseases) because the task location is not considered (e.g., emergency prevention areas are too far apart). To address these issues, this step first uses a drone to conduct low-altitude mapping of the entire orchard, and combines this with GPS positioning to collect the coordinates of planting blocks such as Fuji apple and mandarin orange areas, tree trunks, and irrigation pipes, generating a two-dimensional digital map of the orchard. Then, using the power, location, and material sensors on the robots, it collects the robot's status data, such as 85% remaining power, current location at the north supply station, and high-precision robotic arm (operational accuracy ±1mm). Finally, it integrates the "mandarin orange fertilization" plan entered by the management personnel with the "apple orchard powdery mildew spraying" task triggered by the sensors to construct a panoramic dataset of orchard operations. Based on task characteristics, cherry harvesting (to be completed within 48 hours, highest urgency) and powdery mildew spraying (to be completed within 72 hours, high urgency) are prioritized and time-constrained task sequences are generated. Using a 2D map, three adjacent cherry harvesting subtasks are clustered, and a path planning algorithm is used to plan the path "Replenishment Station → Sub-area 1 → Sub-area 2 → Sub-area 3 → Replenishment Station," estimating energy consumption at approximately 30% of battery power, thus forming optimized task groupings. Finally, based on the robot's dynamic capability profile, the harvesting task is matched to the aforementioned high-precision harvesting robot (assigned the role of "fixed-point harvester"), and the spraying task is matched to a spraying robot loaded with fungicide and with 70% remaining battery power (assigned the role of "inter-row sprayer"). All matching results are integrated to generate a basic role scheduling scheme set.

[0043] By integrating orchard 2D digital map, robot status and task characteristics data, the urgency and priority of tasks are quantified, neighboring tasks are clustered and paths are planned, and then the robot is dynamically matched and adapted. This ensures that tasks are executed in an orderly manner according to their urgency, reduces ineffective robot movement and energy consumption, and achieves precise "task-robot" adaptation, laying a scientific foundation for subsequent collaborative operations.

[0044] In some embodiments, a dynamic capability profile is established for each robot based on its remaining battery power, current location, and types of loaded materials and equipment in the robot's body status data. For each optimized task group, the dynamic capability profile is matched according to its operation type, expected path complexity, and total energy consumption estimate. The dynamic capability profile includes at least: operation capability specialization matching and battery life safety redundancy. An internal matching is initiated from the qualified robot candidates. Each robot applies for a task based on the estimated time and energy consumption cost to complete the task. The task is assigned to the robot with the best overall match. According to the operation type corresponding to the assigned task, an initial operation role is assigned to the winning robot, including but not limited to a fixed-point harvester, a row sprayer, a cross-regional transporter, or a composite inspector.

[0045] A dynamic capability profile is a dynamically updated, personalized capability file built based on real-time robot status data, comprehensively reflecting the robot's adaptability to various tasks. Specialized task matching is a core indicator in the dynamic capability profile representing the robot's professional adaptability to specific task types, such as the adaptability of a harvesting robot to high-precision harvesting tasks or a spraying robot to quantitative spraying tasks. Battery life safety redundancy is an indicator of the robot's remaining battery power, ensuring it can complete the target task while reserving a certain safety margin (to prevent power depletion in case of emergencies). This needs to be dynamically calculated based on the total energy consumption estimate and the robot's remaining battery power. The optimal match is the robot candidate with the highest score in the internal matching process, considering multiple dimensions such as estimated time, energy cost, and task quality assurance capabilities; this robot will ultimately be assigned to the target task group. The initial task role is the core task identity assigned to the winning robot based on the assigned task type, including but not limited to fixed-point harvester, row sprayer, cross-regional transporter, and multi-functional inspector, clearly defining the robot's core operational functions and responsibilities.

[0046] Specifically, traditional orchard robot scheduling often employs static role allocation based on preset types or rotating assignments. Its fundamental flaw lies in treating robots as tools with a constant state, rather than as intelligent agents that change in real time. For example, the actual availability of a device labeled "transport robot" depends on its current battery level, location, and load. If it has just completed a long-distance transport, has only 20% battery remaining, and is located in a remote corner of the orchard, issuing a new cross-regional transport instruction could easily cause it to malfunction en route, blocking critical operational channels. To address these issues, this step uses the robot's battery level, location, material sensors, and equipment detection modules to collect data such as the apple orchard picking robot's remaining battery (85%), current location near the northern picking area, and equipped with a high-precision flexible robotic arm (operating accuracy ±1mm); and the citrus orchard spraying robot's remaining battery (70%), 50L of fungicide loaded, and spraying accuracy ±0.1L / ㎡. This data is used to construct a dynamic capability profile that includes specialized operational capabilities and a safety redundancy in battery life (10% safety reserve). For tasks such as "Picking Fuji apples in Area A of the apple orchard" (estimated energy consumption 30%) and "Spraying pesticides to kill aphids in Area B of the citrus orchard" (estimated energy consumption 25%), a pool of candidate robots with specialized matching capabilities and sufficient battery life was selected. Task details were sent to the candidate robots to initiate internal matching. One candidate robot for apple orchard picking had an estimated completion time of 1.5 hours and energy cost of 25%, while another candidate had an estimated completion time of 2 hours and energy cost of 28%. The robot was scored based on a combination of "estimated time (weight 60%) + energy cost (weight 40%)" to select the best overall performer. Finally, based on the task type, the winning picking robot was assigned the role of "fixed-point picker," and the spraying robot was assigned the role of "row sprayer." The integrated results were then updated into the basic role scheduling scheme set.

[0047] The method provided in this embodiment, based on the construction of a dynamic capability profile that includes job specialization matching and battery life redundancy, combined with an internal competition mechanism to select the best comprehensive robot, is to solve the blindness of the traditional "one-size-fits-all" allocation, ensure that the task is matched with a "specialized, efficient, and long-lasting" robot, reduce resource waste and operational errors, and improve the quality and efficiency of orchard operations.

[0048] In some embodiments, based on the orchard planar operation panoramic dataset and the orchard operation disturbance factor set, an orchard operation environment voxel map reflecting the three-dimensional canopy structure of fruit trees, terrain undulations, and obstacle distribution is constructed; based on the basic role scheduling scheme set and the orchard operation environment voxel map, the spatiotemporal region to be occupied by each robot assigned to a task group is predicted; in the set of spatiotemporal regions of all robots, conflict detection and pre-detection based on three-dimensional space and time dimensions are performed to identify conflict region sets with spatial overlap or time window competition; for each conflict region set, dynamic avoidance negotiation is initiated between robots with the goal of minimizing global operation delay, and an optimized role scheduling cooperative instruction set is generated.

[0049] The three-dimensional canopy structure of fruit trees can be the three-dimensional spatial morphological data of the tree canopy, including information such as canopy height, canopy width, branch and leaf distribution density, and fruit layer height, which directly determines the vertical space requirements and avoidance range of robot operations. Topographic relief can be the elevation changes of the orchard ground, including terrain elements such as slope, aspect, depressions, and protrusions, affecting robot path planning and operational stability. The spatiotemporal region can be the combination of "spatial range + time window" required by the robot to perform a specific task. The spatial range encompasses three-dimensional spatial coordinates, and the time window refers to the start and end time period of the operation; the combination constitutes the robot's dedicated resources. The orchard operational environment voxel map can be a three-dimensional environmental model that integrates orchard terrain, crop structure, obstacles, and disturbance factors, using three-dimensional voxels as basic units. It can accurately reflect the three-dimensional canopy structure of fruit trees, topographic relief, and obstacle distribution, providing three-dimensional environmental support for robot spatiotemporal region prediction and conflict detection. A conflict region set can be a collection of multiple spatiotemporal regions that overlap spatially or compete for time windows, selected after conflict detection. Each set contains spatiotemporal conflict information for at least two robots.

[0050] Specifically, traditional orchard robot scheduling suffers from drawbacks such as two-dimensional environmental modeling, fragmented spatiotemporal planning, and passive conflict handling. It relies solely on planar maps for path planning, neglecting three-dimensional factors like the tree canopy and terrain undulations, often leading to equipment collisions. For example, an apple orchard picking robot (with a robotic arm working height of 1.5-2.5 meters) collides with a spraying robot (with a spraying range of 1.2-2.0 meters) due to vertical spatial overlap. Furthermore, adjustments are only made passively after a conflict occurs, delaying operations. To address these issues, this step integrates a basic role scheduling scheme set, along with orchard two-dimensional digital maps, 1.5-2.5 meter fruit layer data from UAV three-dimensional mapping, real-time wind speed data (level 3) collected by meteorological sensors, and three-dimensional coordinates of irrigation pipes obtained through GPS and visual recognition. This process discretizes the orchard operating space into three-dimensional voxels with sides of 0.5 meters, labeling each voxel with vertical planting layers (such as fruit layer and leaf layer) and spatial occupancy status (such as pipe occupancy), generating an orchard operating environment voxel map. Based on this map and basic scheme, the spatiotemporal areas of the apple orchard picking robot (initial role: fixed-point picker, working area X100-150 meters, Y50-100 meters, time 10:00-12:00, Z-axis 1.5-2.5 meters) and the pesticide spraying robot (initial role: row sprayer, same plane area, time 10:30-12:30, Z-axis 1.5-2.0 meters) are predicted. By comparing the three-dimensional spatial overlap and time window, a 1.5-hour "spatial + temporal" conflict is detected between the two. Dynamic avoidance negotiation is initiated, and it is finally determined that the pesticide spraying robot is adjusted to a three-dimensional detour path (along the Z-axis at the edge of the area, 2.0-2.5 meters). An optimized role scheduling and coordination instruction set including path correction and height adjustment is generated to ensure collision-free collaborative operation.

[0051] The method provided in this embodiment, based on the construction of a three-dimensional voxel map, prediction of robot spatiotemporal regions, early detection of conflicts and negotiation of avoidance, is the core solution to the three-dimensional conflict blind spots and passive handling problems of traditional scheduling. It ensures that multiple robots can cooperate efficiently in three-dimensional space and time dimensions, reduce equipment damage and operation delays, and improve the stability of orchard collaborative operations.

[0052] In some embodiments, the orchard operation disturbance factor set includes real-time weather data, crop 3D planting data, and obstacle 3D spatial data; the orchard operation space is discretized into a uniform-sized 3D cubic grid based on the horizontal range of the orchard 2D digital map and the vertical planting structure defined by the crop 3D planting data, with each grid cell defined as an orchard basic voxel; each orchard basic voxel is assigned a comprehensive attribute label, which includes: vertical planting layer attributes defined based on crop 3D planting data, and spatial occupancy status attributes determined based on obstacle 3D spatial data; based on real-time weather data, a dynamic impact assessment of the orchard basic voxels is performed to generate an orchard operation environment voxel map for robot 3D spatial operation planning.

[0053] Real-time weather data can be meteorological parameters that change in real time during orchard operations, including wind speed, wind direction, rainfall, temperature, and humidity, which directly affect the adjustment of robot operation parameters and path planning. Crop 3D planting data can be structured data characterizing the 3D spatial growth status of fruit trees, including the 3D canopy structure (canopy height, canopy width, branch and leaf distribution density), vertical planting layer division (such as fruiting layer, branch and leaf layer, trunk layer), and spatial growth characteristics corresponding to crop varieties. Obstacle 3D spatial data can be the 3D coordinates and shape data of all objects in the orchard that obstruct robot operations, including fixed obstacles (tree trunks, irrigation pipes, electrical boxes, fences) and temporary obstacles (temporarily stacked tools, maintenance equipment, temporary irrigation hoses). Orchard basic voxels can be individual cubic units in a 3D cubic grid, the smallest spatial unit constituting the voxel map, with each voxel corresponding to a fixed spatial range in the actual orchard. Vertical planting layer attributes can be crop growth layer information (such as fruiting layer, branch and leaf layer, no-crop layer) labeled on voxels based on crop 3D planting data, clearly defining the crop growth status corresponding to each voxel. The space occupancy status attribute can be based on the three-dimensional spatial data of obstacles, and the space occupancy status marked on voxels (such as free, occupied by fixed obstacles, occupied by temporary obstacles), which clarifies whether the voxel can be used by the robot.

[0054] Specifically, traditional orchard environment modeling suffers from two-dimensional, static, and coarse-grained defects. A flat map alone cannot reflect the vertical planting structure of fruit trees and dynamic environmental changes, often leading to robot conflicts: for example, apple orchard picking robots (operating height 1.5-2.5 meters) and spraying robots (1.2-2.0 meters) collide due to vertical overlap; temporarily laid irrigation hoses are not marked, causing transport robots to run over them and cause malfunctions. To address these issues, this step first collects basic and dynamic data: It uses a two-dimensional digital map of the orchard (600 meters east-west and 400 meters north-south) generated by UAV mapping; it obtains three-dimensional planting data of apple orchard crops (fruit layer 1.5-2.5 meters, branch and leaf layer 2.5-3.5 meters) using UAV 3D LiDAR and ground-based supplementary surveys; it collects real-time weather data (wind speed level 3, no rainfall) using meteorological sensors; and it obtains temporary obstacle data (irrigation hoses in area B, X300-320 meters, Y150-160 meters, Z0.1-0.2 meters) through robot vision recognition and GPS positioning. Using the horizontal range of the 2D map and the vertical structure of the crops (Z0-4.0 meters) as boundaries, the space is discretized into a 3D cubic grid with sides of 0.5 meters. Each voxel corresponds to a 0.5m × 0.5m × 0.5m real space. Each voxel is assigned a comprehensive attribute label; for example, voxels in the apple orchard X100-100.5m, Y50-50.5m, and Z1.5-2.0m are labeled "Result Layer + Idle," while the voxel corresponding to the temporary hose is labeled "Temporary Obstacle Occupied." Combined with real-time wind speed, voxels above Z2.5m are labeled "Medium Risk of High-Altitude Operation." After dynamic impact assessment, a voxel map of the orchard's operating environment is generated to support the robot's 3D operation planning.

[0055] The method provided in this embodiment integrates real-time weather, 3D crop planting, and 3D obstacle data to discretize the orchard space into voxels and label their attributes. The core is to solve the pain point that traditional modeling cannot support 3D operation planning, so that the environmental model can accurately adapt to the dynamic scene and provide a reliable spatial basis for robot conflict avoidance and collaborative operation.

[0056] In some embodiments, before matching robots to optimized task groups, the task requirements are upgraded to three-dimensional task requirements that include a clearly defined required vertical working space; the robot's dynamic capability profile is upgraded to three-dimensional physical capabilities that include its maximum working height, robotic arm working envelope, and body foldability; based on a voxel map of the orchard working environment, spatiotemporal conflict prediction is performed, and it is searched whether there are other robots with matching capabilities that have won bids for tasks in the same spatiotemporal region within the same time period; if a spatial conflict is found where the regions intersect, the robot application is not immediately rejected, but a dynamic avoidance negotiation strategy is executed as a prerequisite for the final task allocation and instruction generation; the dynamic avoidance negotiation strategy includes a conflict resolution sub-strategy based on spatiotemporal resource scheduling and a collaborative optimization sub-strategy based on dynamic reorganization of tasks and roles.

[0057] Three-dimensional task requirements can be based on existing optimized task grouping requirements, supplementing the task description with clearer vertical working space requirements. This includes the three-dimensional coordinate range of the operation, vertical height constraints, and three-dimensional spatial operation accuracy, upgrading the task requirements from two-dimensional to three-dimensional. Maximum working height refers to the highest vertical height at which the robot can stably perform the task, such as the maximum extension height of a harvesting robot's arm or the maximum coverage height of a spraying robot's spray. The robotic arm's working envelope is the closed area formed by all positions that the robot arm can reach in three-dimensional space, reflecting the robotic arm's three-dimensional working range and activity limits. Body foldability refers to whether the robot body has folding and retraction capabilities, and the changes in size after folding, affecting the robot's ability to move and operate in narrow three-dimensional spaces (such as densely planted orchards or under low canopies). Spatiotemporal conflict prediction can be based on a voxel map of the orchard's working environment, searching for whether there is three-dimensional spatial overlap or time window competition between different robots' predetermined spatiotemporal areas within the same time period. The core is "advance prediction and proactive avoidance." Dynamic avoidance negotiation strategies can be systematic approaches to resolve predicted spatial conflicts. Instead of directly rejecting robot task requests, these strategies utilize temporal adjustments, spatial reconstruction, and task reorganization to mitigate the conflict. This includes conflict resolution sub-strategies and collaborative optimization sub-strategies. Conflict resolution sub-strategies, centered on "avoiding conflict and ensuring task progress," directly resolve conflicts by adjusting time windows or reconstructing work paths. Collaborative optimization sub-strategies, centered on "integrating resources and improving overall efficiency," transform conflicts into collaborative opportunities through task and role reorganization or robot fusion, achieving high resource utilization.

[0058] Specifically, traditional orchard robot task matching suffers from drawbacks such as two-dimensional task requirements, incomplete capability profiles, and simplistic conflict handling. It focuses only on planar area allocation, ignoring vertical working space and the robot's three-dimensional physical capabilities. Furthermore, it directly rejects robot requests when conflicts arise, leading to resource waste and frequent operational conflicts. For example, in apple orchards, picking robots (operating height 1.5-2.5 meters) collide with pruning robots (2.5-4.0 meters) due to vertical overlap; in densely planted citrus orchards, assigning robots without folding capabilities (1.2 meters in height) to narrow passages results in them getting entangled in vines; and highly adaptable picking robots are left idle due to minor time conflicts with inspection robots, leading to the selection of suboptimal robots and resulting in fruit damage. To address the above issues, this step first upgrades the optimized task groups to three-dimensional task requirements. For example, "Picking in Area A of the Fuji Apple Orchard" specifies the three-dimensional area as X120-180 meters, Y60-100 meters, and Z1.5-2.5 meters (result layer), with a three-dimensional operation accuracy of ±1mm; "Fertilizing in Area B of the Tangerine Orchard" specifies a vertical operating space of 0.8-1.2 meters, avoiding the tree trunk's three-dimensional coordinates of X220 meters, Y100 meters, and Z0-0.8 meters. Next, the robot's dynamic capability profile is upgraded to three-dimensional physical capabilities. For example, the picking robot is supplemented with a maximum operating height of 3.0 meters and a robotic arm working envelope radius of 1.5 meters, while the fertilizing robot is supplemented with a foldable body (0.8 meters high after folding) and a vertical coverage range of 0.5-1.2 meters. Based on a voxel map of the orchard's operating environment with a side length of 0.5 meters, the spatiotemporal regions of each robot are predicted. For example, the picking robot operates in the "10:00-12:00 + the above three-dimensional region", and the pruning robot operates in the "10:30-12:30 + X120-180 meters, Y60-100 meters, Z2.5-3.5 meters". After detecting a critical conflict in the vertical space, the robot does not reject the application but initiates a dynamic avoidance negotiation strategy. It adopts a spatial reconstruction scheme to adjust the operating height of the pruning robot to 3.0-4.0 meters to avoid the conflict, which is a prerequisite for the final task allocation.

[0059] The method provided in this embodiment utilizes the upgraded 3D task requirements and the robot's 3D physical capabilities to predict spatiotemporal conflicts in advance and perform dynamic avoidance negotiation. The core is to solve the pain points of inaccurate matching and rigid conflict handling in traditional methods, and achieve "precise matching, advance avoidance of conflicts, and efficient collaboration", thereby avoiding equipment damage and waste of resources and ensuring the smooth progress of orchard operations.

[0060] In some embodiments, a time-domain delay scheme is evaluated and implemented, that is, under the premise of satisfying the urgency constraints of both parties in the conflicting tasks, one robot role is guided to temporarily switch to a waiter and its task time window is adjusted to avoid space occupation; a spatial domain reconstruction scheme is evaluated and implemented, that is, based on the voxel map of the orchard operation environment, one robot role is guided to temporarily switch to a detour explorer and a three-dimensional alternative path that can avoid the current spatial conflict is planned and assigned to it.

[0061] The temporal delay scheme can be an important branch of the conflict resolution sub-strategy. Without changing the robot's workspace and core task, it resolves conflict by adjusting the task time windows of one or both conflicting parties, staggering their time periods of space occupation. The waiting robot, in the temporal delay scheme, temporarily switches roles. Its core responsibility is to pause work or remain on standby during a designated time period, resuming its original role and continuing its task after the conflict is resolved, without adding any additional work. The spatial reconstruction scheme can be another important branch of the conflict resolution sub-strategy. Without changing the task time window, it plans a completely new 3D work path for the conflicting robot based on a voxel map of the orchard's work environment, avoiding the conflict area, thus resolving the conflict. The detour explorer, in the spatial reconstruction scheme, temporarily switches roles. Its core responsibility is to perform work according to the newly planned 3D alternative path, avoiding conflict by detouring. After completing the task, it can resume its original role or maintain that role to adapt to subsequent paths. A three-dimensional alternative path can be a redesigned operational path for the conflict robot, which includes three-dimensional spatial coordinates (XY plane position + Z-axis height). It must avoid the original conflict area and meet the robot's three-dimensional physical capabilities and orchard environmental constraints (such as crop height and obstacle distribution).

[0062] Specifically, traditional orchard robot conflict handling suffers from drawbacks such as a single approach, neglect of task urgency, and limitation to two-dimensional adjustments. When conflicts arise, random stoppages and waiting are often employed without considering task priority and the three-dimensional environment, leading to delays in critical tasks and wasted resources. For example, in apple orchards, the conflict between ripe fruit harvesting (within a 3-day window) and routine inspections results in overripe fruit due to the traditional method of waiting for harvesting; in hilly citrus orchards (12° slope), the overlap between transport and spraying robots in three-dimensional space due to the terrain's inclination cannot be resolved by two-dimensional path adjustments; and in densely planted vineyards, robots face no alternative paths when conflicts occur, forcing them to stop abruptly and disrupting operations. To address the above issues, this step addresses the conflict in Apple Orchard Area A: the harvesting robot (a fixed-point harvester, operating from 10:00 to 12:00, with a 3-day fruit ripening window, extremely high urgency) and the inspection robot (10:30 to 11:30, moderate urgency) overlap in three-dimensional space. A time-domain delay scheme is implemented. Under the constraint of urgency, the inspection robot is temporarily switched to a waiting position, with its time window adjusted to 12:00 to 13:00. X150 meters, Y100 meters, and Z0-1.2 meters are designated as standby areas. Operations resume after harvesting is completed. Regarding the conflict in Area B of the hilly citrus orchard (slope 12°): the transport robot (transferring citrus from 14:00 to 15:00, high urgency) and the spraying robot (spraying pesticides at the same time, 1.2-1.8 meters high, high urgency) overlapped in three-dimensional space. Based on a voxel map of the orchard's operating environment with a side length of 0.5 meters, a spatial reconstruction scheme was implemented, allowing the transport robot to temporarily switch to a detour explorer and plan a three-dimensional alternative path (X200-260 meters, Y80-120 meters, Z0.5-1.2 meters) to avoid the spraying height and the original conflict area. It arrived at the picking point on time at 14:25, and both completed their tasks as planned, successfully resolving the conflict.

[0063] The method provided in this embodiment, which adjusts time and plans three-dimensional alternative paths according to urgency and combines them with temporary role switching, solves the pain points of rigid and poor adaptability in traditional conflict handling, ensures that key tasks are prioritized and conflicts are resolved flexibly, and guarantees continuous and efficient orchard operations.

[0064] In some embodiments, task and role decoupling and reorganization are performed, that is, the operational compatibility between conflicting tasks is analyzed. If they are determined to be mutually exclusive, a region closure instruction is dynamically inserted or the operational parameters are adjusted to isolate the impact. If they are determined to be complementary, the operational role or parameters of one robot are dynamically adjusted to coordinate the use of the time and space resources. Cross-robot role fusion is performed, that is, in extremely narrow or high-coordination scenarios, two conflicting robots and their tasks are merged and combined into a temporary collaborative unit with composite functions through operational logic. A unified time and space pipeline and collaborative instructions are assigned to this unit.

[0065] Task and role decoupling and recombination can break the traditional model of "fixed tasks bound to fixed roles." By analyzing the operational compatibility of conflicting tasks, the task execution logic and robot roles can be recombined, either isolating mutually exclusive influences or strengthening complementary and collaborative processing methods. Operational compatibility refers to the degree of compatibility between two or more conflicting tasks in terms of work processes, work parameters, and environmental requirements, and is divided into mutually exclusive (conflicting work logic, cannot be performed simultaneously) and complementary (related work objectives, can be promoted collaboratively). Area closure instructions can be instructions to temporarily delineate a dedicated work area for mutually exclusive tasks and prohibit other robots from entering, avoiding conflicts through spatial isolation. Adjusting work parameters can be specific adjustable indicators for different types of orchard operations, mainly used to isolate the influence of mutually exclusive tasks or strengthen the collaboration of complementary tasks. Extremely narrow scenarios can be areas in orchards with limited work space, such as row widths of less than 1.2 meters in densely planted orchards or canopy heights of less than 1.0 meter in dwarf fruit trees, where a single robot has no redundant space and multiple robots are prone to conflict. Temporary collaborative units can be temporary work combinations formed by the fusion of cross-robot roles. The two robots each perform their own duties and work together. After the task is completed, they automatically disband and resume their independent roles.

[0066] Specifically, traditional orchard robot conflict handling only reaches the level of "passive avoidance," lacking collaborative awareness. It fails to effectively isolate mutually exclusive tasks, easily leading to secondary conflicts or operational contamination. It also wastes the collaborative value of complementary tasks and cannot adapt to extremely confined environments. For example, in densely planted cherry orchards (row width 0.8 meters), a conflict between picking and pruning robots would traditionally necessitate a work stoppage; in apple orchards, picking and high-concentration spraying are mutually exclusive, and the lack of isolation leads to fruit contamination; in vineyards, pruning and spraying are complementary, but the avoidance mechanism reduces the effectiveness of pest control. To address the above issues, this step addresses the conflict between the spraying robot (0.3% fungicide, 9:00-10:30) and the harvesting robot in the apple orchard. Based on a voxel map of the orchard's operating environment, a closed area (X300-350 meters, Y100-150 meters, Z0.5-2.5 meters) is defined, and a closed area command is issued. The harvesting robot's time window is adjusted to 10:30-12:00, requiring pre-harvest testing of fruit for pesticide residues (<0.01 mg / kg) to isolate the mutually exclusive impact. Regarding the conflict between the pruning robot (14:00-15:30) and the spraying robot in the vineyard, the spraying robot's role is adjusted to collaborative spraying, with parameters set to an atomized particle size of 8 micrometers and a pressure of 0.2 MPa. A following path is planned (lagging the pruning robot by 5 meters), sharing spatiotemporal resources to achieve "pruning as spraying." For extreme scenarios in densely planted cherry orchards (0.8 meters between rows): where there is a conflict between the picking robot (0.5 meters wide) and the pruning robot (0.4 meters wide), the pruning robot can be folded down to 0.35 meters to form a temporary collaborative unit. Dedicated time and space channels are allocated (X200-250 meters, Y80-130 meters, Z0.7-1.0 meters, 10:00-12:00). The pruning robot clears branches in front, and the picking robot harvests behind, completing the task simultaneously.

[0067] The method provided in this embodiment enables a rapid response to unexpected events within critical agricultural time windows. By reorganizing resources and reallocating tasks, it ensures that core production activities are not interrupted or that interruptions are minimized, effectively guaranteeing the stability of orchard production and the reliability of output, and reducing direct economic losses and risks caused by unexpected shutdowns.

[0068] In some embodiments, based on the execution feedback of the optimized role scheduling collaborative instruction set, the final role sequence, actual spatiotemporal trajectory, dynamic negotiation events and task completion status of all robots are summarized; based on the orchard operation disturbance factor set, the overall timeliness, energy consumption and conflict resolution effectiveness of collaborative operations are evaluated; based on the comprehensive feedback data and evaluation results, a multi-robot collaborative operation scheduling report containing optimization suggestions for this round of operations and the initial role scheduling pre-configuration for the next cycle is generated.

[0069] Role sequence can be a complete record of a single robot's role changes during this task, from its initial role to temporary role switching (such as waiter, detour explorer), and then to restoration or update of its role, reflecting the entire process of dynamic adjustment of robot roles. Actual spatiotemporal trajectory can be a continuous record of the actual three-dimensional spatial coordinates and time nodes occupied by the robot when performing the task. Comparison with the predicted spatiotemporal region can assess the accuracy of path planning and conflict resolution. Dynamic negotiation events can be records of inter-robot negotiation events triggered by conflicts and disturbance factors during the task, including the reason for negotiation, participating robots, negotiation strategy (temporal delay / spatial reconstruction / cross-robot fusion), negotiation results and impacts, etc. Task completion status can be the final execution result of each optimized task group, including whether it is completed, completion time, task quality (such as fruit damage rate, pesticide coverage rate), resource consumption (electricity, materials) and other core indicators.

[0070] Specifically, traditional orchard robot collaborative operations lack closed-loop management, resulting in fragmented operational data, one-sided performance evaluation, and a lack of continuous optimization mechanisms. This leads to difficulties in tracing the root causes of problems and hinders the improvement of operational efficiency. For example, in apple orchards, the high rate of fruit breakage during harvesting makes it impossible to determine whether it's due to improper role matching or changes in conflict resolution strategies; in citrus orchards, poor pesticide application results are unclear whether it's due to deviations in spraying trajectories or delays in negotiation regarding optimal timing; and in densely planted areas, frequent conflicts in the previous round of operations fail to lead to targeted optimizations in the next round. To address the above issues, this step first summarizes the execution feedback data: The robot's role sequence of "fixed-point harvester → waiter (conflict with spraying robot) → harvester" is collected, and its actual spatiotemporal trajectory is recorded (X120-180 meters, Y60-100 meters, Z1.5-2.5 meters, 10:00-12:30). Three dynamic negotiation events are analyzed (time-domain delay resolves harvesting and spraying conflicts, spatial domain reconstruction resolves transportation and fertilization conflicts, cross-machine integration adapts to harvesting and pruning in densely planted areas). The completion status of four task groups is statistically analyzed (Apple A area harvesting 2.5 hours, damage rate 2%, power consumption 40%; Citrus B area spraying coverage 98%, pesticide consumption 30L). Combining real-time weather (wind speed level 2), crop 3D planting data, and other disturbance factors, the overall efficiency is evaluated: total duration 6 hours, delay 0.5 hours (due to transport robot detours), conflict resolution success rate 100%, total power consumption 185%. Finally, optimization suggestions (enhancing narrow passage path rehearsal and pre-setting temporary role equipment parameter templates) and pre-configuration for the next cycle (setting "cherry picking + transportation" as a collaborative group and adding folding robots to densely planted areas) are generated and integrated to form a multi-robot collaborative operation scheduling report.

[0071] The scheduling report, provided in this embodiment, comprehensively records dynamic data such as robot role switching, spatiotemporal trajectory, and negotiation events, providing a complete basis for tracing operational problems. For example, when the fruit breakage rate is abnormal, the role sequence can be checked to see if there are any improper role switching, and the actual spatiotemporal trajectory can be compared to check for path collisions, quickly locating the root cause of the problem, avoiding repeated trial and error, and improving the stability of operational quality.

[0072] Figure 3 This is a schematic diagram of the structure of an orchard autonomous robot collaborative operation system based on dynamic role allocation, provided in one embodiment of this application. Figure 3 As shown, the orchard autonomous robot collaborative operation system 300 based on dynamic role allocation in this embodiment includes: a basic role analysis module 301, an optimized role analysis module 302, and a scheduling report generation module 303.

[0073] The basic role analysis module 301 is used to acquire a panoramic dataset of orchard operations, extract and model two-dimensional planar operation features based on the dataset, and then perform robot capability matching and initial role allocation decisions in the two-dimensional plane to generate a basic role scheduling scheme set. The optimized role analysis module 302 is used to acquire a set of orchard operation disturbance factors, and perform three-dimensional spatial conflict resolution and global collaborative optimization based on the basic role scheduling scheme set and the orchard operation disturbance factor set to generate an optimized role scheduling collaborative instruction set. The scheduling report generation module 303 is used to execute role switching instructions and collaborative operation control for heterogeneous robot individuals based on the optimized role scheduling collaborative instruction set to generate a multi-robot collaborative operation scheduling report.

[0074] Optionally, the basic role analysis module 301, when generating the basic role scheduling scheme set, specifically performs the following: the orchard planar operation panoramic dataset includes orchard two-dimensional digital map data, robot body state data, and operation task feature data; based on the operation task feature data, it performs quantitative calculation of task urgency and priority to generate an orchard task sequence with time constraint attributes; combined with the orchard two-dimensional digital map data, it performs geographical location-based clustering and path planning preprocessing on the orchard task sequence with time constraint attributes to generate optimized task groups with expected execution paths and energy consumption estimates; based on the robot body state data, it matches the tasks in the optimized task groups to robots with matching states and assigns them initial operation roles through a dynamic allocation method based on rules and filtering mechanisms, thereby generating the basic role scheduling scheme set.

[0075] Optionally, the basic role analysis module 301, when based on the dynamic allocation method, is specifically used for: establishing a dynamic capability profile for each robot based on the remaining battery power, current position, and types of loaded materials and equipment in the robot body status data; matching the dynamic capability profile for each optimized task group according to its operation type, expected path complexity, and total energy consumption estimate, wherein the dynamic capability profile includes at least: operation capability specialization matching and battery life safety redundancy; initiating internal matching from qualified robot candidates, and allocating tasks to the best-matched robot based on the estimated time and energy consumption cost of completing the task; assigning an initial operation role to the winning robot according to the operation type corresponding to the assigned task, including but not limited to fixed-point harvester, row sprayer, cross-regional transporter, or composite inspector.

[0076] Optionally, the optimized role analysis module 302, when generating the optimized role scheduling and coordination instruction set, is specifically used for: constructing an orchard operation environment voxel map reflecting the three-dimensional canopy structure, terrain undulation, and obstacle distribution of fruit trees based on the orchard planar operation panoramic dataset and the orchard operation disturbance factor set; predicting the spatiotemporal area required for each robot assigned to a task group to perform its task based on the basic role scheduling scheme set and the orchard operation environment voxel map; performing conflict detection and pre-detection based on three-dimensional space and time dimensions in the set of spatiotemporal areas of all robots, identifying conflict area sets with spatial overlap or time window competition; and initiating dynamic avoidance negotiation between robots with the goal of minimizing global operation delay for each conflict area set, and generating the optimized role scheduling and coordination instruction set.

[0077] Optionally, the optimized role analysis module 302, during the construction process of the orchard operation environment voxel map, specifically performs the following: the orchard operation disturbance factor set includes real-time weather data, crop 3D planting data, and obstacle 3D spatial data; using the horizontal range of the orchard 2D digital map and the vertical planting structure defined by the crop 3D planting data, the orchard operation space is discretized into a uniform-sized 3D cubic grid, with each grid cell defined as an orchard basic voxel; a comprehensive attribute label is assigned to each orchard basic voxel, the comprehensive attribute label including: a vertical planting layer attribute defined based on the crop 3D planting data, and a spatial occupancy status attribute determined based on the obstacle 3D spatial data; based on the real-time weather data, a dynamic impact assessment is performed on the orchard basic voxels to generate the orchard operation environment voxel map for robot 3D spatial operation planning.

[0078] Optionally, the optimized role analysis module 302, during the prediction process based on the spatiotemporal region, specifically performs the following: before matching robots to the optimized task groups, it upgrades the task requirements to three-dimensional task requirements that include a clearly defined required vertical working space; it upgrades the robot's dynamic capability profile to three-dimensional physical capabilities that include its maximum working height, robotic arm working envelope, and body foldability; based on the orchard working environment voxel map, it performs spatiotemporal conflict prediction and searches whether there are other qualified robots with matched capabilities in the predetermined spatiotemporal regions of the same time period; if a spatial conflict is found between the regions, it does not immediately reject the robot's application, but instead executes a dynamic avoidance negotiation strategy as a prerequisite for the final task allocation and instruction generation; the dynamic avoidance negotiation strategy includes a conflict resolution sub-strategy based on spatiotemporal resource scheduling and a collaborative optimization sub-strategy based on dynamic reorganization of tasks and roles.

[0079] Optionally, the optimized role analysis module 302, when based on the conflict resolution sub-strategy, is specifically used to: evaluate and execute a time-domain delay scheme, that is, under the premise of satisfying the urgency constraints of both parties in the conflicting task, guide one robot role to temporarily switch to a waiter, and adjust its task time window to avoid space occupation; evaluate and execute a spatial domain reconstruction scheme, that is, according to the orchard operation environment voxel map, guide one robot role to temporarily switch to a detour explorer, and plan and allocate a three-dimensional alternative path that can avoid the current spatial conflict.

[0080] Optionally, the optimized role analysis module 302, when based on the collaborative optimization sub-strategy, is specifically used for: performing task and role decoupling and reorganization, that is, analyzing the operational compatibility between conflicting tasks; if they are determined to be mutually exclusive, dynamically inserting region closure instructions or adjusting operational parameters to isolate the influence; if they are determined to be complementary, dynamically adjusting the operational role or parameters of one robot to collaboratively utilize the spatiotemporal resources; and performing cross-robot role fusion, that is, in extremely narrow or high-collaboration scenarios, merging two conflicting robots and their tasks, forming a temporary collaborative unit with composite functions through operational logic combination, and allocating a unified spatiotemporal pipeline and collaborative instructions to this unit.

[0081] Optionally, the scheduling report generation module 303 is specifically used for: summarizing the final executed role sequence, actual spatiotemporal trajectory, dynamic negotiation events, and task completion status of all robots based on the execution feedback of the optimized role scheduling and coordination instruction set; evaluating the overall timeliness, energy consumption, and conflict resolution effectiveness of the collaborative operation based on the orchard operation disturbance factor set; and generating the multi-robot collaborative operation scheduling report containing optimization suggestions for this round of operation and the initial role scheduling pre-configuration for the next cycle based on the comprehensive feedback data and evaluation results. The system in this embodiment can be used to execute the method of any of the above embodiments, and its implementation principle and technical effects are similar, so they will not be described again here.

Claims

1. A collaborative operation method for autonomous robots in orchards based on dynamic role allocation, characterized in that, include: Obtain a panoramic dataset of orchard operations, extract and model two-dimensional operation features based on the orchard panoramic dataset, and then perform robot capability matching and initial role allocation decisions in the two-dimensional plane to generate a basic role scheduling scheme set. Obtain the set of orchard operation disturbance factors, and based on the set of basic role scheduling schemes and the set of orchard operation disturbance factors, perform three-dimensional spatial conflict resolution and global collaborative optimization to generate an optimized role scheduling collaborative instruction set; Based on the optimized role scheduling and collaborative instruction set, role switching instructions and collaborative operation control for heterogeneous robot individuals are executed to generate a multi-robot collaborative operation scheduling report.

2. The method according to claim 1, characterized in that, The generated basic role scheduling scheme set includes: The orchard planar operation panoramic dataset includes orchard two-dimensional digital map data, robot body status data, and operation task feature data. Based on the task feature data, the urgency and priority of the tasks are quantitatively calculated to generate an orchard task sequence with time constraints. Combining the orchard two-dimensional digital map data, the orchard task sequence with time constraints is preprocessed by geographic location-based clustering and path planning to generate optimized task groups with expected execution paths and energy consumption estimates. Based on the robot's physical state data, tasks in the optimized task group are matched to robots with matching states and assigned initial job roles through a dynamic allocation method based on rules and filtering mechanisms, thereby generating the basic role scheduling scheme set.

3. The method according to claim 2, characterized in that, The dynamic allocation method includes: Based on the remaining battery power, current location, and types of loaded materials and equipment in the robot's status data, a dynamic capability profile is created for each robot. For each optimized task group, the dynamic capability profile is matched based on its job type, expected path complexity and total energy consumption estimate. The dynamic capability profile includes at least: job capability specialization matching and power consumption safety redundancy. An internal matching process is initiated from qualified robot candidates. Each robot, based on its estimated time to complete the task and energy consumption cost, applies for the task and assigns the task to the robot with the best overall match. Based on the job type corresponding to the assigned task, assign an initial job role to the winning robot, including but not limited to fixed-point picker, row sprayer, cross-regional transporter, or composite inspector.

4. The method according to claim 3, characterized in that, The generated optimized role scheduling coordination instruction set includes: Based on the orchard planar operation panoramic dataset and the orchard operation disturbance factor set, an orchard operation environment voxel map reflecting the three-dimensional canopy structure of fruit trees, terrain undulation and obstacle distribution is constructed. Based on the basic role scheduling scheme set and the orchard operation environment voxel map, predict the spatiotemporal area that each robot in the assigned task group needs to occupy to perform its task. Within the set of spatiotemporal regions of all robots, conflict detection and pre-detection based on three-dimensional space and time dimensions are performed to identify sets of conflict regions with spatial overlap or time window competition. For each set of conflict zones, dynamic avoidance negotiation is initiated between robots with the goal of minimizing global operation delay, and the optimized role scheduling coordination instruction set is generated.

5. The method according to claim 4, characterized in that, The process of constructing the voxel map of the orchard operating environment includes: The orchard operation disturbance factor set includes real-time weather data, three-dimensional crop planting data, and three-dimensional spatial data of obstacles; Using the horizontal range of the orchard's two-dimensional digital map and the vertical planting structure defined by the crop's three-dimensional planting data, the orchard's operational space is discretized into a three-dimensional cubic grid of uniform size, with each grid cell defined as an orchard basic voxel. Each of the orchard basic voxels is assigned a comprehensive attribute label, which includes: a vertical planting layer attribute defined based on the crop three-dimensional planting data, and a spatial occupancy status attribute determined based on the obstacle three-dimensional spatial data; Based on the real-time weather data, a dynamic impact assessment is performed on the basic voxels of the orchard to generate a voxel map of the orchard's operational environment for robot 3D spatial operation planning.

6. The method according to claim 5, characterized in that, The prediction process for the spatiotemporal region includes: Before performing robot matching on the optimized task group, its task requirements are upgraded to three-dimensional task requirements that include a clearly defined required vertical working space. The robot's dynamic capability profile is upgraded to include its maximum working height, robotic arm working envelope, and body foldability in three-dimensional physical capabilities. Based on the voxel map of the orchard operation environment, spatiotemporal conflict prediction is performed, and the system searches whether there are other qualified robots that have won the bid for the task in the same spatiotemporal region within the same time period. If a spatial conflict is found where regions intersect, the robot's application will not be rejected immediately. Instead, a dynamic avoidance negotiation strategy will be implemented as a prerequisite for the final task allocation and instruction generation. The dynamic avoidance negotiation strategy includes a conflict resolution sub-strategy based on spatiotemporal resource scheduling and a collaborative optimization sub-strategy based on dynamic reorganization of tasks and roles.

7. The method according to claim 6, characterized in that, The conflict resolution sub-strategy includes: Evaluate and implement a time-domain delay scheme, which means, under the premise of satisfying the urgency constraints of both parties in the conflicting tasks, guide one of the robots to temporarily switch to the waiting role and adjust its task time window to stagger space occupation. Evaluate and implement the airspace reconstruction plan, which involves guiding one robot to temporarily switch to a detour explorer based on the voxel map of the orchard's operating environment, and planning and assigning it a three-dimensional alternative path that can avoid the current spatial conflict.

8. The method according to claim 6, characterized in that, The collaborative optimization sub-strategy includes: The task and role decoupling and reorganization involves analyzing the operational compatibility between conflicting tasks. If they are determined to be mutually exclusive, a region closure command is dynamically inserted or the operational parameters are adjusted to isolate the impact. If they are determined to be complementary, the operational role or parameters of one robot are dynamically adjusted to make collaborative use of the space-time resources. Perform cross-robot role fusion, that is, in extremely narrow or highly collaborative scenarios, merge two conflicting robots and their tasks, combine them through operational logic to form a temporary collaborative unit with composite functions, and assign a unified spatiotemporal pipeline and collaborative instructions to this unit.

9. The method according to claim 8, characterized in that, The generation of the multi-robot collaborative operation scheduling report includes: Based on the execution feedback of the optimized role scheduling and coordination instruction set, the final role sequence, actual spatiotemporal trajectory, dynamic negotiation events, and task completion status of all robots are summarized. Based on the set of orchard operation disturbance factors, the overall timeliness, energy consumption and conflict resolution effectiveness of collaborative operations are evaluated. Based on the comprehensive feedback data and evaluation results, a multi-robot collaborative operation scheduling report is generated, which includes optimization suggestions for this round of operations and the initial role scheduling pre-configuration for the next cycle.

10. An orchard autonomous robot collaborative operation system based on dynamic role allocation, characterized in that, The method applied to any one of claims 1-9 includes: The basic role analysis module is used to acquire a panoramic dataset of orchard operations, extract and model two-dimensional planar operation features based on the panoramic dataset, and then perform robot capability matching and initial role allocation decisions in the two-dimensional plane to generate a basic role scheduling scheme set. The optimized role analysis module is used to obtain the orchard operation disturbance factor set. Based on the basic role scheduling scheme set and the orchard operation disturbance factor set, three-dimensional spatial conflict resolution and global collaborative optimization are performed to generate an optimized role scheduling collaborative instruction set. The scheduling report generation module is used to execute role switching instructions and collaborative operation control for heterogeneous robot individuals based on the optimized role scheduling and collaboration instruction set, and generate a multi-robot collaborative operation scheduling report.

Images