Real-time task allocation system and real-time task allocation method for autonomous transport robots in orchards

Abstract

The system of the present invention is a real-time work assignment system for performing a harvest transport operation using a plurality of autonomous transport robots in an orchard environment, comprising: a robot state management unit that manages in real-time state information including the position, load amount, and operation mode of each of the plurality of autonomous transport robots; and a central work assignment control unit that receives a harvest collection request from a worker and selects an optimal robot to perform the collection request based on the state information and assigns the work, wherein the central work assignment control unit calculates a cost function by summing the total estimated travel distance (d_total) of the candidate robot and the value obtained by multiplying the distance traveled with the harvest loaded (d_loaded) among the total estimated travel distance by a pre-set weight (W), and selects the robot that minimizes the cost function value as the optimal robot.

Description

Technology Field

[0001] The present invention relates to autonomous robot operation technology and agricultural automation technology. More specifically, it relates to a system for dynamically assigning an optimal robot in response to real-time work requests that occur during the process of transporting harvested produce using a plurality of autonomous transport robots in an irregular agricultural environment such as an orchard.

[0002] In particular, the present invention relates to a real-time task assignment system and method that prevents degradation of harvest quality during the transport process and maximizes the operational efficiency of the entire system by reflecting the distance traveled with harvest loaded, beyond the simple distance traveled, in the cost calculation when assigning tasks to a robot. Background Technology

[0003] To address the issues of declining and aging agricultural populations, the introduction of autonomous transport robots to automate the transport of harvested produce in orchard environments is being actively researched. For the efficient operation of robots in such environments, technology that assigns tasks to appropriate robots based on worker collection requests is essential.

[0004] Conventional autonomous robot task assignment technologies have primarily been developed for structured environments such as logistics warehouses or manufacturing plants, and have mainly utilized pre-planning-based assignment methods or simple shortest-distance-based assignment methods.

[0005] However, orchard environments possess unique characteristics, such as irregular worker locations, harvest times, and yields, as well as uneven and often unpaved terrain. When conventional simple distance-based assignment methods are applied to orchards, a problem may arise where a specific robot is assigned solely because it is close to the requested work location, resulting in it traveling long distances while loaded with harvest. Moving for extended periods on uneven surfaces while carrying harvest is a direct cause of fruit drop or reduced marketability due to vibrations and shocks. Nevertheless, conventional technology has failed to account for this "distance traveled while loaded" in the calculation of work assignment costs.

[0006] Furthermore, most conventional job assignment systems were designed with a blocking structure that assigns new tasks only when a robot is in an idle state. This had limitations, as it caused resources to be inefficiently wasted on robots returning to the collection center or waiting at the work site, even when there was available loading space, thereby slowing down the overall system response speed and reducing the job throughput.

[0007] (Patent Document 1) KR 10-2626502 B The problem to be solved

[0008] The present invention was devised to solve the problems of the prior art as described above, and its main purpose is to solve the problems of inefficiency and quality degradation that occur during the transport of harvested produce using an autonomous transport robot in an orchard environment.

[0009] The present invention aims to preserve the quality of harvested produce by actively suppressing long-distance movement while loaded, through the introduction of a new cost function that considers not only the total movement distance of the robot but also the distance traveled while loaded as independent cost variables and assigns weights to them.

[0010] Another technical objective of the present invention is to maximize the utilization of robot resources by implementing a non-blocking task assignment structure that manages the robot's state in real-time by subdividing it into Idle, Waiting, To_home, etc., and allows for the dynamic reallocation of new tasks even when the robot is not in an idle state, provided there is spare load capacity.

[0011] Furthermore, the present invention has another technical objective of performing optimal matching that minimizes the overall system cost by comprehensively considering multiple work requests and multiple robot status information occurring in real time, thereby distributing the workload in a balanced manner and improving the system's responsiveness to dynamic orchard environments. means of solving the problem

[0012] The system of the present invention is a real-time work assignment system for performing a harvest transport operation using a plurality of autonomous transport robots in an orchard environment, comprising: a robot state management unit that manages in real-time state information including the position, load amount, and operation mode of each of the plurality of autonomous transport robots; and a central work assignment control unit that receives a harvest collection request from a worker and selects an optimal robot to perform the collection request based on the state information and assigns the work, wherein the central work assignment control unit calculates a cost function by summing the total estimated travel distance (d_total) of the candidate robot and the value obtained by multiplying the distance traveled with the harvest loaded (d_loaded) among the total estimated travel distance by a pre-set weight (W), and selects the robot that minimizes the cost function value as the optimal robot. Effects of the invention

[0013] The real-time work assignment system for an autonomous transport robot for an orchard according to the present invention has the following effects.

[0014] This invention significantly reduces the distance a robot travels while carrying harvested produce by assigning tasks based on a cost function that assigns weights to the distance traveled while loaded. This minimizes the risk of damage, quality degradation, and fruit drop caused by vibrations or shocks during transport, thereby greatly contributing to maintaining the marketability of agricultural products.

[0015] The present invention provides a non-blocking structure that allows a robot to be assigned new tasks not only when idle but also while returning or waiting. This reduces unnecessary idle time for the robot and maximizes utilization, thereby improving the overall task processing efficiency and productivity of the system.

[0016] The present invention can respond quickly and flexibly to work requests occurring in real time, even in irregular and dynamic work environments such as orchards. By immediately reflecting changes in the robot's dynamic state and reassigning tasks, it minimizes worker waiting time and facilitates the smooth flow of the overall harvesting operation. Brief explanation of the drawing

[0017] FIG. 1 is a block diagram illustrating the configuration of the system of the present invention. FIGS. 2 and FIGS. 3 are a graph and a result table of simulation results according to an embodiment of the present invention. Specific details for implementing the invention

[0018] Hereinafter, a preferred embodiment of a real-time work assignment system for an autonomous transport robot for an orchard according to the present invention will be described in detail with reference to the attached drawings. The system of the present invention aims to simultaneously achieve transport efficiency and quality preservation of harvested produce in an irregular and dynamic environment such as an orchard.

[0019] FIG. 1 is a block diagram illustrating the configuration of the system of the present invention.

[0020] The system of the present invention is largely composed of a user request receiving unit (100), a robot state management unit (200), a central work assignment control unit (300), and a robot motion control unit (400). These components are organically connected to each other to perform real-time data exchange, and establish and execute an optimal robot operation strategy for irregular and dynamic work requests occurring in a special work environment such as an orchard.

[0021] First, the user request receiving unit (100) is an interface module that receives and processes harvest collection requests in real time from a worker performing harvest work in an orchard. The worker can generate collection requests through various input means, such as a portable terminal, a smartphone, a wearable device, or a voice recognition device.

[0022] The work request data received by the user request receiving unit (100) may include the following information. First, a unique user ID is included to identify the worker, enabling analysis of the specific worker's work patterns and history management. Second, the worker's current location information, expressed in the form of GPS coordinates or an orchard zone code, is included to enable the robot to identify the exact destination. Third, information on the requested quantity, such as the estimated weight of the harvest to be collected or the number of boxes, is included and utilized for planning the robot's loading capacity. Fourth, information on the time of request occurrence in the form of a timestamp is recorded and used for determining work priority and analyzing waiting time.

[0023] The user request receiving unit (100) has the flexibility to sequentially register received requests in a first-in, first-out (FIFO) work queue, but to reorder them according to urgency or priority. It may also include an intelligent queue management function that integrates and processes multiple requests that occur consecutively from the same worker, or groups and manages requests from multiple workers in adjacent locations.

[0024] Next, the robot state management unit (200) is a core module that collects, updates, stores, and manages dynamic state information of each of the multiple autonomous transport robots operated in the orchard in real time. This module continuously receives state information by linking with the sensor system and communication module mounted on each robot and systematically stores it in a central database.

[0025] The status information managed by the robot status management unit (200) is broadly classified into location information, loading status information, operation mode information, and performance indicator information. The location information includes the robot's real-time coordinates and movement path history obtained through GPS, IMU, vision sensors, etc. The loading status information tracks the current load quantity (carry_quantity), maximum load capacity (MAX_CARRY), and remaining load capacity in real time, and ensures accuracy through a weight sensor or a vision-based volume measurement system.

[0026] In particular, the robot's operating modes can be managed by subdividing them. While conventional technology simply classified the robot's state as either Idle or Busy, the present invention defines and manages subdivided modes as follows. Idle mode refers to a state where the robot waits at the collection center and is ready for immediate work. Move mode (To_user) refers to a state where the robot is moving to an assigned worker's location, moving while empty. Waiting mode indicates a state where the robot waits for a certain period of time after arriving at the worker's location and completing loading, considering the possibility of additional work. Return mode (To_home) refers to a state where the robot is returning to the collection center with the harvested produce loaded.

[0027] This granular mode management serves as the basis for enabling non-blocking task assignment, which is the core of the present invention. The robot state management unit (200) clearly defines the switching conditions and switching times between each mode, and supports dynamic reassignment by immediately notifying the central task assignment control unit (300) whenever a mode switching event occurs.

[0028] Next, the central work assignment control unit (300) is a module that implements the core intelligence of the present invention and performs optimal work-robot matching by comprehensively analyzing the work request received from the user request receiving unit (100) and the robot status information received from the robot status management unit (200).

[0029] The job assignment process of the central job assignment control unit (300) above is largely composed of three stages.

[0030] The first step is the allocability determination stage. In this stage, a pool of robot candidates capable of receiving a task at the current time is selected. One of the innovative features of the present invention is the implementation of a non-blocking allocation structure. In conventional technology, only robots in an idle state could be assigned new tasks; however, in the present invention, robots in standby mode or return mode are also included as allocable candidates if they satisfy specific conditions. Specifically, a robot in standby mode is determined to be eligible for allocation if the difference between its current load and maximum load is greater than the expected quantity of new task requests. A robot in return mode is determined to be able to perform additional tasks by changing its path if the distance from its current location to the new task location is within a threshold and there is sufficient load capacity. This flexible allocability determination maximizes the utilization of robot resources and minimizes the waiting time of workers.

[0031] The second step is the cost calculation step. This involves the introduction of the weighted cost function used in this step. The cost function is defined as follows:

[0032] Cost = d_total + W × d_loaded

[0033] Here, d_total represents the total estimated distance the robot travels from its current position to the worker position and from the worker position to the collection center. This can be calculated using an appropriate method among Manhattan distance, Euclidean distance, or a path distance reflecting the actual road network of the orchard. d_loaded represents the distance the robot travels with harvested produce loaded out of the total travel distance, i.e., the distance from the worker position to the collection center. W is a weight parameter that can be set by the system administrator to adjust the strength of the penalty for moving while loaded.

[0034] Specifically, the larger the value of the weight W, the more actively the system avoids moving while loaded with harvest. For example, when W=0, it operates the same as the conventional simple distance-based allocation, but when W=5, the distance traveled while loaded is reflected with a weight of 6 times in the cost calculation, significantly suppressing long-distance movement while loaded with harvest. This brings about a practical effect of protecting the harvest from vibrations and shocks occurring on unpaved roads in the orchard, and preventing fruit drop and quality degradation.

[0035] The third step is the optimal matching step. The central task assignment control unit (300) constructs a cost matrix based on the costs calculated for all assignable robot-task pairs. This cost matrix contains the costs for all possible combinations between m task requests and n assignable robots. The system uses this cost matrix as input to find the optimal matching that minimizes the total system cost.

[0036] To this end, optimization techniques such as the Hungarian Algorithm, the Auction Algorithm, or the Kuhn-Munkres Algorithm can be utilized. Since these algorithms guarantee a global optimal solution with polynomial time complexity, they are suitable for real-time task assignment. Additionally, if the number of task requests exceeds the number of robots, virtual robots can be added to convert the matrix into a square matrix, or a priority-based sequential assignment method can be applied.

[0037] Next, the robot motion control unit (400) is an execution module that converts and controls work instructions received from the central work assignment control unit (300) into actual physical movements of the robot. This module can be implemented by being distributed to each robot to operate autonomously, or by transmitting remote control commands from a central server.

[0038] The robot motion control unit (400) plans an optimal path to perform assigned tasks. When planning the path, it considers terrain information of the orchard, obstacle locations, and the movement paths of other robots to avoid collisions and ensure safe movement. In particular, when carrying harvested produce, it limits sudden changes in direction or acceleration to ensure the stability of the harvested produce.

[0039] When the robot arrives at the worker's location, the robot motion control unit (400) manages the loading process. During this process, it receives a loading completion signal through an interface with the worker and measures the actual loading amount through a weight sensor. After loading is complete, it determines the next action by comparing the current loading amount with the maximum loading amount. If the maximum loading amount is reached, it immediately begins returning to the collection center. On the other hand, if additional loading is possible, it waits at the current location for a pre-set waiting time (WAIT_TIME) and prepares to accept additional work requests that may occur nearby during this time.

[0040] One of the features of the present invention is the dynamic path replanning capability of the robot motion control unit (400). When the robot is moving toward a collection center in return mode or waiting at a worker location in standby mode, it may receive a new work assignment instruction from the central work assignment control unit (300). In this case, the robot motion control unit (400) immediately stops the current operation and replans the path toward the new destination. The robot that was returning creates a path from its current location to the new worker location, and the robot that was waiting ends its standby and immediately begins moving. This dynamic replanning capability significantly improves the responsiveness and efficiency of the entire system.

[0041] Additionally, the robot motion control unit (400) handles exceptional situations that may occur during movement. When situations such as the discovery of unexpected obstacles on the path, low battery, or mechanical failure occur, it immediately notifies the robot state management unit (200) of the state change, and allows the central task assignment control unit (300) to exclude the robot from the list of assignable candidates or reassign the task to another robot.

[0042] An embodiment of the present invention will be described with reference to FIGS. 2 and FIGS. 3.

[0043] A simulation example performed to verify the effectiveness of the real-time work assignment system for an autonomous transport robot for an orchard according to the present invention will be described in detail. In particular, the impact of the cost function, which assigns weights to the loading state travel distance—the core of the present invention—on the preservation of harvest quality and work efficiency will be described in detail.

[0044] FIG. 2 is a graph of the simulation results according to an embodiment of the present invention, and FIG. 3 is a table of the simulation results.

[0045] [Simulation Environment and Conditions]

[0046] To verify the validity of the present invention in a virtual orchard environment, the following simulation environment was constructed.

[0047] Working environment: As shown in Fig. 2, a two-dimensional space with a unit area of ​​100 x 100 was set as an orchard work area.

[0048] Robots and Users: Three autonomous transport robots (R1, R2, R3) were deployed, and requests for harvest collection were set to occur randomly in real time at approximately 15 user locations (U01~U15).

[0049] Execution time: The simulation was performed for a total of 6,000 time steps.

[0050] Task assignment criteria: Task assignment was performed based on the following cost function proposed in the present invention.

[0051] Cost = d_total + W × d_loaded

[0052] [Experimental Design (Comparative Examples and Examples)]

[0053] The purpose of this experiment is to analyze the effect of changes in the weight (W) for the loading state movement distance (d_loaded) on system performance. To this end, comparative examples and exemplary cases were distinguished and performed as follows according to the setting of the weight (W).

[0054] Comparison Example (W=0): Set weight (W) to 0.

[0055] This corresponds to conventional technology that assigns work based only on the total distance traveled without considering the loading status, as Cost = d_{total}.

[0056] Example 1 (W=1): When a weight of 1 is assigned to the loading state movement distance according to the present invention.

[0057] Example 2 (W=2): When a weight of 2 is assigned to the loading state movement distance according to the present invention.

[0058] Example 3 (W=5): A case in which the minimization of the loading state movement is prioritized by assigning a weight of 5 to the loading state movement distance according to the present invention.

[0059] [Simulation Results and Analysis]

[0060] Figure 2 is a snapshot at a specific point in time (Time Step 6000) of the simulation, showing multiple robots dynamically assigned tasks and moving in response to multiple user requests. Additionally, through the status window on the left, it can be seen that the movement distance (Dist), idle time (Idle), and load state movement time (LoadMove) of each robot are being managed in real time.

[0061] Figure 3 shows the quantitative results of the simulation for each comparative example and embodiment. The performance indicators of the three robots are summarized in Table 1 below.

[0062] [Table 1] Summary of Simulation Results According to Changes in Weight (W)

[0063]

[0065] According to the present embodiment, there is an effect of preserving the quality of the harvested product (significant reduction in the time of transport in the loaded state).

[0066] A very significant change was observed in the 'time to move while loaded with harvest (Load_move),' which is one of the objectives of the present invention. In the comparative example (W=0) corresponding to the prior art, a total of 10,450 unit times were required, but in Example 3 (W=5) with increased weighting applied to the present invention, only 6,848 unit times were required.

[0067] This is a significant result, showing a decrease of approximately 34.5% compared to the comparative example. In addition, the ratio of the time spent carrying items to the total time also decreased significantly from 17.42% in the comparative example to 11.42% in Example 3.

[0068] This result clearly demonstrates that the weighted cost function of the present invention induces the system to actively avoid long-distance movement while loaded. Through this, the effect of preventing quality degradation and fruit drop was achieved by minimizing the time harvested produce is exposed to vibration or shock during transport.

[0069] As a result of the analysis of work efficiency maintenance and trade-offs, there was no significant difference in the 'total number of processed orders,' which represents the overall productivity of the system, between the comparative example (701 cases) and examples 1 to 3 (707 cases). This demonstrates that the present invention can maximize the effect of preserving the quality of harvested products without compromising existing work efficiency.

[0070] Meanwhile, as the weight (W) increased, the robot's 'Total Distance' in Example 3 (22346.5) increased slightly compared to Comparative Example (19732.4). This is because the system selected an allocation method that intentionally minimizes movement in a loaded state (e.g., a path that moves more in an empty state) even if the total distance increases slightly, in order to preserve the quality of the harvest.

[0071] Through the simulation results above, it was verified that applying the loading state-based weighted cost function according to the present invention can significantly reduce the time and rate at which the robot moves while loaded with harvested produce, while maintaining the overall work processing efficiency of the system.

[0072] The system of the present invention configured as described above provides an optimized robot operation solution that maximizes work efficiency while preserving the quality of harvests in a special agricultural environment such as an orchard. In particular, the organic combination of a weighted cost function considering the loading state travel distance and a non-blocking dynamic reallocation structure realizes a level of performance improvement that conventional technology could not achieve.

Claims

Claim 1 A real-time work assignment system for performing harvest transport operations using multiple autonomous transport robots in an orchard environment, comprising: a robot state management unit that manages in real-time state information including the position, load capacity, and operating mode of each of the multiple autonomous transport robots; and a central work assignment control unit that receives a harvest collection request from a worker, selects the optimal robot to perform the collection request based on the state information, and assigns the work. The system includes a robot motion control unit that controls the movement of the plurality of autonomous transport robots according to work instructions received from the central work assignment control unit, wherein the operation modes include an idle mode, a user mode, a waiting mode, and a home mode; the central work assignment control unit, in selecting candidate robots to perform the collection request, includes not only robots in idle mode but also robots in waiting mode or home mode whose current load is less than a preset maximum load; in calculating the matching cost between the candidate robot and the collection request, it calculates a cost function by summing the total estimated travel distance (d_total) of the candidate robot and the value obtained by multiplying the distance traveled while loaded with harvested material (d_loaded) out of the total estimated travel distance by a preset weight (W), and selects the robot with the minimum cost function value as the optimal robot; and when a new task is assigned to a robot in waiting mode or home mode, the robot motion control unit stops the currently performing operation and immediately resets the movement path to the new work location. A real-time task assignment system characterized by controlling movement. Claim 2 A real-time work assignment system according to claim 1, wherein the central work assignment control unit generates a cost matrix using the cost function for all combinations between a plurality of collection requests and a plurality of candidate robots, and determines a work-robot matching combination that minimizes the total cost of the entire system by applying an optimization algorithm. Claim 3 A real-time task assignment method for a central control server to perform a harvest transport operation using multiple autonomous transport robots in an orchard environment, comprising: (a) receiving a harvest collection request from a worker; (b) verifying status information including the position, load, and operation mode of each of the multiple autonomous transport robots, wherein the operation mode includes an idle mode, a user mode, a waiting mode, and a home mode; (c) selecting, based on the status information, as candidate robots, robots in the idle mode as well as robots in the waiting or home mode whose current load is less than a preset maximum load; (d) for each of the candidate robots, calculating a cost function by summing the total estimated travel distance (d_total) and the value obtained by multiplying the distance traveled while loaded with harvest (d_loaded) within the total estimated travel distance by a preset weight (W); and (e) assigning the task for the collection request to the candidate robot that has the minimum cost function value. and (f) a step of stopping the currently performing operation and immediately resetting the movement path to the work position when a task is assigned to a candidate robot in the standby mode or the return mode; a real-time task assignment method comprising. Claim 4 delete Claim 5 delete

Images