Method and system for controlling pick-and-place time sequence of sky rail robot and multiple machining devices

Abstract

The invention relates to the technical field of robot control, and discloses a pick-and-place time sequence control method and system for a sky rail robot and multiple machining equipments.The pick-and-place time sequence control method for the sky rail robot and the multiple machining equipments.The pick-and-place time sequence control method for the sky rail robot and the multiple machining equipmentsincludes the steps that a machining equipment set and cache capacity parameters are read, and a service urgency index is calculated; operating state information is collected, the abnormal probability is calculated, and the processing equipment is divided into a fault isolation set and a serviceable set; global optimization is carried out on the service path of the sky rail robot, and an action sequence is generated and executed; performing isolation operation on the processing equipment in the fault isolation set, and smoothly incorporating the recovered processing equipment; the whole process of taking and placing operation is recorded, and the estimated machining completion moment is updated; according to the method, the service path of the sky rail robot is globally optimized through the ant colony algorithm, a high-quality action sequence is found in a complex combinatorial optimization problem, downstream accumulation is prevented through automatic adjustment when a fault occurs, and stable operation of a production system is guaranteed.

Description

Method and system for controlling the loading and unloading sequence of a ceiling-rail robot and multiple processing devices. Technical Field

[0001] This invention relates to the field of robot control technology, and more specifically, to a method and system for controlling the timing of picking up and placing multiple processing devices using a ceiling-mounted robot. Background Technology

[0002] With the rapid development of industrial automation technology, ceiling-mounted robots are widely used in material handling scenarios in multi-station processing lines due to their advantages such as high space utilization, large range of motion, and no need to occupy ground space. In fields such as crystal processing and precision parts manufacturing, ceiling-mounted robots typically need to serve multiple processing devices arranged in series or parallel, completing the tasks of picking up, placing, and transferring workpieces between devices according to a preset time sequence plan.

[0003] Chinese Patent CN111823238B discloses a robot timing control method, a power timing management unit, and a control system. When the main circuit is powered on again after a power outage, the system determines the output status of the power module. If the power module is in power-off protection mode, it determines whether the robot controller is powered off. If so, it outputs a power-on trigger signal to the robot controller to restart its operating system. If not, it monitors the status of the teach pendant program within the robot controller or the power-off signal from the robot controller's mainboard in real time. If the teach pendant program is detected to be closed, it delays for a preset time to allow the robot controller to power off. After the delay ends or a power-off signal from the robot controller's mainboard is received, it outputs a power-on trigger signal to the robot controller to restart its operating system.

[0004] However, when faced with sudden failures, existing technologies lack an effective time-series reconstruction mechanism after a failure occurs. Robots may continue to try to access the faulty equipment according to the original time sequence, causing non-faulty equipment to be forced to stop and wait due to lack of timely service, triggering a chain of production stoppages. When the faulty equipment is repaired and put back into production, existing technologies cannot smoothly integrate it into the current operating time sequence system, which may cause material backlog, buffer overflow or cycle disorder due to improper time sequence switching, resulting in material loss imbalance and affecting the overall efficiency of the production line. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for controlling the loading and unloading sequence of a ceiling-mounted robot and multiple processing devices in order to solve the above-mentioned problems.

[0006] This invention provides a method for controlling the loading and unloading sequence of a ceiling-mounted robot and multiple processing devices, comprising the following steps:

[0007] Read the set of processing equipment and the cache capacity parameters of each processing equipment, establish hard constraints based on the cache capacity parameters, and calculate the service urgency index to quantify the service priority of the processing equipment.

[0008] The operating status information of each processing equipment is collected at a preset sampling period. The abnormal probability of each processing equipment is calculated based on multi-source signals. The processing equipment is divided into a fault isolation set and a serviceable set according to the abnormal probability and operating status information. When any processing equipment is transferred from the serviceable set to the fault isolation set, a dynamic switching operation is performed.

[0009] Based on the service urgency index, the ant colony algorithm is used to globally optimize the service path of the overhead rail robot. Each processing device in the service set is mapped to a node in the ant colony algorithm, and the robot's movement path between processing devices is mapped to an edge between nodes. An action sequence is generated and executed through iterative search and pheromone update.

[0010] Isolate the processing equipment within the fault isolation set, monitor the downstream accumulation risk caused by the fault, activate the virtual back pressure control mechanism, and smoothly integrate the recovered processing equipment.

[0011] The entire process of each pick-up and place operation performed by the overhead rail robot is recorded, and the estimated processing completion time in the operation status information is updated.

[0012] Furthermore, the cache capacity parameter includes inbound cache capacity and outbound cache capacity;

[0013] The hard constraints include buffer capacity constraints and interlock safety constraints. The buffer capacity constraint requires that for any time and any processing equipment, the inbound buffer occupancy of the processing equipment must be greater than or equal to zero and not exceed the inbound buffer capacity of the processing equipment, and the outbound buffer occupancy of the processing equipment must be greater than or equal to zero and not exceed the outbound buffer capacity of the processing equipment. The interlock safety constraint indicates whether the processing equipment allows the robot to perform pick-and-place operations through an interlock permission signal.

[0014] The service urgency index is calculated by multiplying the preset first weighting coefficient by the outbound cache urgency, adding the preset second weighting coefficient by the inbound cache urgency, and adding the preset third weighting coefficient by the processing completion time urgency. The outbound cache urgency indicates the degree to which the outbound cache exceeds the warning threshold, the inbound cache urgency indicates the degree to which the inbound cache is below the warning threshold, and the processing completion time urgency is the preset processing completion urgency baseline duration divided by the remainder, where the remainder is the remaining processing time plus the preset processing completion urgency baseline duration.

[0015] Furthermore, performing dynamic switching operations includes:

[0016] Operational status information includes processing equipment status, inbound buffer occupancy, outbound buffer occupancy, and estimated processing completion time;

[0017] The method for calculating the anomaly probability is to normalize each multi-source signal, multiply each normalized multi-source signal by its corresponding preset weight coefficient, and then sum them up.

[0018] If the abnormal probability of a certain processing equipment is greater than or equal to a preset abnormal probability threshold, or if the processing equipment is in a fault state, then the processing equipment is added to the fault isolation set; otherwise, the processing equipment is added to the serviceable set.

[0019] Maintain the execution sequence and shadow sequence;

[0020] Determine the number of actions that have been completed in the execution sequence and obtain the current state information;

[0021] Find the longest consecutive executable prefix subsequence in the shadow sequence that meets the following conditions: starting from the first action in the shadow sequence, perform action feasibility verification and state prediction sequentially and remove actions involving new faulty processing equipment; if the length of the prefix subsequence is greater than zero, replace the actions that have not yet been executed in the execution sequence with the prefix subsequence; if the length of the prefix subsequence is zero, immediately re-plan and generate a new execution sequence.

[0022] Furthermore, the multi-source signals include processing equipment fault codes, abnormal temperature signals, abnormal vibration signals, sudden changes in processing cycle time signals, and the number of times interlock rejections occurred;

[0023] The execution sequence is the sequence of actions that will be issued to the overhead rail robot for execution. Its length is the number of steps specified by the preset rolling planning window length. It includes material picking actions, material placing actions, moving actions or waiting actions and the corresponding target processing equipment. The shadow sequence is the next round of optimal action sequence pre-calculated under the premise that the current fault isolation set remains unchanged. The shadow sequence is generated when the background calculation starts immediately after the execution sequence begins to execute.

[0024] Status information includes the actual status of the overhead rail robot, robot vehicle occupancy, and inbound and outbound buffer occupancy of each processing equipment;

[0025] Feasibility verification of an action should include at least hard constraints.

[0026] Furthermore, the generation and execution of action sequences through iterative search and pheromone updates include:

[0027] Perform the initial configuration of the ant colony algorithm, setting the preset number of ants, preset maximum number of iterations, preset pheromone importance coefficient, preset heuristic pheromone importance coefficient, preset pheromone volatility coefficient, and preset pheromone increment constant;

[0028] A problem model for the ant colony algorithm is constructed, which maps each processing device in the service set to a node in the ant colony algorithm, maps the robot's movement path between processing devices to an edge between nodes, initializes the pheromone concentration on each edge to a preset initial pheromone concentration value, and calculates the heuristic pheromone value of each edge.

[0029] Introduce action feasibility constraints and state prediction mechanisms into the problem model, set robot carrier capacity and robot carrier occupancy, and set material picking feasibility conditions, material placement feasibility conditions and composite picking and placement feasibility conditions.

[0030] The ant colony algorithm is executed in an iterative search process. In each iteration, each ant starts from the robot's current position and selects the next target processing equipment node in turn until a complete service path containing the number of steps specified by the preset rolling planning window length is constructed. When selecting the next target processing equipment node, the ant adopts a probabilistic selection mechanism. After each ant completes a complete path, the overall fitness value of the path is calculated and the path with the highest fitness value in the current iteration is recorded as the optimal path in this round.

[0031] After completing one iteration, a pheromone update operation is performed, including pheromone evaporation and pheromone release;

[0032] Repeat the iterative search and pheromone update process until the preset maximum number of iterations is reached or the convergence condition is met. After the algorithm terminates, it outputs the globally optimal path as the optimal service sequence for the sky-rail robot.

[0033] Based on the optimal service sequence, the target processing equipment for each step is determined sequentially, and the feasibility of the action and the action type are verified and determined at each step based on the expected arrival time. The action type includes material picking action, material unloading action or combined picking and unloading action, and the generated action is added to the action sequence.

[0034] Furthermore, the heuristic information value is the service urgency item divided by the time cost item. The time cost item is the sum of the movement cost and the waiting cost plus a preset time cost benchmark value. The service urgency item is the service urgency index plus a preset urgency benchmark value. Here, the movement cost represents the time required for the robot to move from its current position to the processing equipment, and the waiting cost represents the time the robot needs to wait for the processing equipment to complete after arriving at the processing equipment. The waiting cost is calculated by subtracting the sum of the current time and the movement cost from the estimated processing completion time of the processing equipment. If the difference is positive, the difference is taken as the waiting cost; if the difference is negative or zero, the waiting cost is zero.

[0035] The feasible conditions for picking up materials are that the outbound buffer occupancy of the target processing equipment is greater than zero, the interlock permission signal is in the permission state, and the robot carrier occupancy is less than the robot carrier capacity. The feasible conditions for unloading materials are that the inbound buffer occupancy of the target processing equipment is less than the inbound buffer capacity, the interlock permission signal is in the permission state, and the robot carrier occupancy is greater than zero. The feasible conditions for combined picking up and unloading are that the same target processing equipment simultaneously meets the feasible conditions for unloading and picking up materials at the same arrival time, and the execution order of the combined picking up and unloading actions is unloading first and then picking up materials.

[0036] State prediction is used to advance the state at the arrival time when constructing candidate solutions. The calculation method of state prediction is to accumulate the movement cost and waiting cost of each step based on the current time of the overhead rail robot and the robot vehicle occupancy according to the candidate action sequence to obtain the expected arrival time of each step. At each step arrival time, the feasibility of the action is verified by combining the interlock permission signal, the occupancy of the inbound buffer, the occupancy of the outbound buffer and the estimated processing completion time.

[0037] Furthermore, the probabilistic selection mechanism includes, for each candidate processing device of the current ant, first calculating, based on state prediction, whether the candidate processing device has at least one executable action at the expected arrival time and eliminating candidate processing devices that do not meet the action feasibility constraints. The executable actions include material picking, material releasing, or a combination of picking and releasing. Then, the probability of the remaining candidate processing devices being selected is calculated. Specifically, this includes taking the pheromone concentration on the edge from the current node to the candidate processing device, increasing it to a power of a preset pheromone importance coefficient to obtain an enhanced pheromone concentration, increasing the heuristic information value of the edge to a power of a preset heuristic information importance coefficient to obtain an enhanced heuristic information value, multiplying the enhanced pheromone concentration by the enhanced heuristic information value as the selection tendency value of the candidate processing device, and dividing the selection tendency value of the candidate processing device by the sum of the selection tendency values ​​of all candidate processing devices to obtain the probability of the candidate processing device being selected. The ant selects the next target processing device node using a roulette wheel method based on the above probability distribution.

[0038] The fitness value of the path is calculated by summing the service urgency index of the target processing equipment corresponding to each step in the path to obtain the total service urgency, and summing the movement cost and waiting cost of each step in the path to obtain the total time cost. The fitness value of the path is the total service urgency divided by the total time cost item, and the total time cost item is the sum of the total time cost and the preset fitness time benchmark value.

[0039] The convergence condition is that the globally optimal path remains unchanged in multiple consecutive iterations, and the number of consecutive iterations is a preset convergence judgment round.

[0040] Furthermore, activating the virtual backpressure control mechanism and smoothly integrating the recovered processing equipment includes:

[0041] Perform isolation operations on the processing equipment within the fault isolation set, and prohibit the overhead rail robot from entering the pick-up and put-out interlock area of ​​the faulty processing equipment;

[0042] The system monitors the downstream back pressure risk caused by the fault. When the ratio of the outbound buffer occupancy of any processing equipment to the outbound buffer capacity of the processing equipment is detected to be greater than or equal to a preset back pressure trigger threshold, a virtual back pressure control mechanism is activated. The virtual back pressure control mechanism includes calculating the back pressure coefficient.

[0043] The back pressure coefficient is used to control the flow rate by adjusting the upstream feeding cycle. When the back pressure coefficient is greater than or equal to the preset back pressure stop threshold, the standard feeding interval time of the upstream feeding processing equipment is divided by the back pressure coefficient to obtain the adjusted feeding interval time. When the back pressure coefficient is less than the preset back pressure stop threshold, the upstream feeding is suspended until the back pressure coefficient recovers.

[0044] When a processing device in the fault isolation set reports a recovery signal or an anomaly probability less than a preset anomaly probability threshold, the processing device is transferred from the fault isolation set to the observation recovery set. A stability window determination is performed on the processing devices in the observation recovery set. If a processing device simultaneously meets the conditions of not being in a fault state and having an interlock allowance rate greater than or equal to a preset allowance rate threshold within a continuous preset stability observation window duration, then the processing device is allowed to be included in the serviceable set.

[0045] For recovery processing equipment that passes the stability window assessment, a gradual quota mechanism is used for smooth inclusion.

[0046] Furthermore, the back pressure coefficient is calculated as follows: for each processing device in the serviceable set, 1 is subtracted from the cache occupancy item, where the cache occupancy item is the ratio of the outbound cache occupancy to the outbound cache capacity of the processing device. If the result is negative, it is taken as zero. Then, the minimum value is taken among the above calculation results of all processing devices in the serviceable set, and the minimum value is the back pressure coefficient.

[0047] The interlock permission rate is the ratio of the number of successful pick-up and put-down requests to the total number of requests within a preset stable observation window.

[0048] The progressive quota mechanism allocates service weights to the recovery processing equipment. The service weights are calculated using a ramp function. The calculation method is to take the current time and subtract the time when the processing equipment passes the stability window determination, and then divide the difference by a preset ramp duration parameter. If the ratio is greater than one, it is taken as one; otherwise, the ratio is taken as the service weight. By gradually increasing the service weight from zero to one, the recovery processing equipment gradually transitions from participating in a small number of scheduling tasks to normal competitive scheduling. When the service weight equals one, the recovery processing equipment enters a fully normal scheduling state.

[0049] This invention provides a timing control system for the pick-and-place of a ceiling-mounted robot and multiple processing devices, which stores computer-readable instructions. When the computer-readable instructions are read, the aforementioned timing control method for the pick-and-place of the ceiling-mounted robot and multiple processing devices can be executed. The system includes:

[0050] The priority determination module reads the set of processing equipment and the cache capacity parameters of each processing equipment, establishes hard constraints based on the cache capacity parameters, and calculates the service urgency index to quantify the service priority of the processing equipment.

[0051] The status awareness module collects the operating status information of each processing device at a preset sampling period, calculates the abnormal probability of each processing device based on multi-source signals, and divides the processing devices into a fault isolation set and a serviceable set according to the abnormal probability and operating status information. When any processing device is transferred from the serviceable set to the fault isolation set, a dynamic switching operation is performed.

[0052] The path optimization module, based on the service urgency index, uses the ant colony algorithm to globally optimize the service path of the overhead rail robot. It maps each processing device in the service set to a node in the ant colony algorithm, maps the robot's movement path between processing devices to an edge between nodes, and generates and executes an action sequence through iterative search and pheromone update.

[0053] The fault control module performs isolation operations on the processing equipment within the fault isolation set, monitors the downstream accumulation risk caused by the fault, activates the virtual back pressure control mechanism, and smoothly integrates the recovered processing equipment.

[0054] The history update module records the entire process of each pick-up and place operation performed by the overhead rail robot and updates the estimated processing completion time in the running status information.

[0055] The beneficial effects of this invention are as follows: This invention quantifies the service priority of each processing device through a service urgency index, comprehensively considering three dimensions: outbound cache urgency, inbound cache urgency, and processing completion time urgency. This enables the overhead rail robot to prioritize serving the most urgent processing devices, effectively avoiding outbound backlog and inbound material shortages, and improving the overall operating efficiency of the production system. By employing an ant colony algorithm to globally optimize the service path of the overhead rail robot, and through the collaborative search of multiple virtual ants and the positive feedback mechanism of pheromones, high-quality action sequences are found in complex combinatorial optimization problems. Compared with traditional greedy scheduling strategies, this invention has stronger global optimization capabilities and can maximize service efficiency while satisfying cache capacity constraints and interlocking safety constraints.

[0056] This invention achieves dynamic rerouting capabilities through a dual-sequence mechanism of execution sequence and shadow sequence. When a processing equipment malfunction is detected, it can quickly switch to a pre-calculated shadow sequence, completing path adjustment without interrupting robot operation, significantly improving the system's response speed to sudden faults and production continuity. By calculating the anomaly probability of processing equipment based on multi-source signal fusion, and comprehensively considering multiple dimensions of information such as processing equipment fault codes, abnormal temperature signals, abnormal vibration signals, sudden changes in processing cycle time, and the number of interlock rejections, fault determination is performed. Compared with single-signal judgment, it has a higher fault detection rate and a lower false alarm rate, achieving accurate identification of abnormal states of processing equipment. This invention uses a virtual backpressure control mechanism and a progressive quota mechanism to automatically adjust the upstream feeding cycle time to prevent downstream accumulation when a fault occurs, and uses a ramp function to smoothly incorporate the recovered processing equipment during fault recovery, avoiding sudden changes in production cycle time and ensuring the stable operation of the production system. Attached Figure Description

[0057] Figure 1 is a flowchart illustrating the timing control method for the pick-up and place-down of the overhead rail robot and multiple processing devices according to the present invention.

[0058] Figure 2 is an example diagram of the generation and execution of action sequences for the pick-up and place-down timing control method of the overhead rail robot and multiple processing equipment of the present invention;

[0059] Figure 3 is an example of the smooth integration of the pick-up and place-up timing control method of the overhead rail robot and multiple processing equipment according to the present invention;

[0060] Figure 4 is a module example diagram of the pick-and-place timing control system for the overhead rail robot and multiple processing equipment of the present invention. Detailed Implementation

[0061] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0062] A method and system for controlling the loading and unloading sequence of a ceiling-mounted robot and multiple processing devices includes the following embodiments:

[0063] Example 1:

[0064] A method for controlling the loading and unloading sequence of a ceiling-rail robot and multiple processing devices is applied to a production system comprising one ceiling-rail robot and N processing devices, wherein the ceiling-rail robot serves the N processing devices along a track. Each processing device is equipped with an inbound buffer area and an outbound buffer area, wherein the capacity of the inbound buffer area of ​​the i-th processing device is denoted as the inbound buffer capacity of that processing device, and the capacity of the outbound buffer area is denoted as the outbound buffer capacity of that processing device. The method is shown in Figure 1 and includes the following steps:

[0065] Step 100: Read the set of processing equipment and the cache capacity parameters of each processing equipment, establish hard constraints based on the cache capacity parameters, and calculate the service urgency index to quantify the service priority of the processing equipment.

[0066] The system reads the set of processing equipment, which includes all processing equipment from machine 1 to machine N. It also reads the buffer capacity parameters for each processing equipment, including inbound and outbound buffer capacities, both of which are positive integers greater than zero. The buffer capacity parameter is used as input because it directly determines the system's material buffering capacity and anti-interference capability. The inbound buffer capacity reflects the amount of material to be processed that the processing equipment can store, while the outbound buffer capacity reflects the amount of processed material that the processing equipment can temporarily store. These two parameters together constitute the material flow boundary conditions for the processing equipment, serving as the fundamental constraint for formulating the pick-and-place timing strategy. A preset rolling planning window length is set, representing the number of steps in the future action sequence generated for each planning iteration. The default value is 5 steps. This default value is determined based on the number of processing equipment and the robot's movement speed. When the number of processing equipment is large or the robot's movement speed is slow, this value can be appropriately increased to improve planning foresight; when the number of processing equipment is small or real-time requirements are high, this value can be appropriately decreased to improve response speed. The reason for using a rolling planning window is that the state of the production system changes dynamically over time. Fixed long-term planning is difficult to adapt to sudden changes in the state of processing equipment and fluctuations in material demand. Rolling planning, on the other hand, plans only a limited number of future actions in each control cycle and continuously updates the planning results during execution. This allows it to maintain a certain degree of foresight while responding to environmental changes in a timely manner, thereby achieving a balance between computational complexity and control effectiveness.

[0067] Establish hard constraints for system operation, including buffer capacity constraints and interlocking safety constraints. These hard constraints represent the physical limitations and safety boundaries of the system's operation. Violating these constraints will lead to system malfunction or safety risks; therefore, they must be strictly adhered to during the action planning phase to ensure that the generated action sequence is physically executable and safety-acceptable. Regarding buffer capacity constraints, for any given time and any processing equipment, the inbound buffer occupancy of the processing equipment must be greater than or equal to zero and not exceed its inbound buffer capacity. Similarly, the outbound buffer occupancy of the processing equipment must be greater than or equal to zero and not exceed its outbound buffer capacity. The purpose of buffer capacity constraints is to prevent material overflow and material shortages. Continuing to feed material when the inbound buffer occupancy reaches its capacity will result in material accumulation and nowhere to store it; conversely, continuing to retrieve material when the outbound buffer occupancy is zero will result in no material to retrieve. Both of these situations will cause system malfunctions. Regarding interlock safety constraints, an interlock permission signal is set to indicate whether the processing equipment allows the robot to perform pick-and-place operations. When the interlock permission signal is in the permitted state, it means that the processing equipment allows the robot to perform pick-and-place operations; when the interlock permission signal is in the prohibited state, it prohibits the robot from performing pick-and-place operations. The purpose of interlock safety constraints is to ensure the safety of human-machine collaboration and equipment operation. When the processing equipment is performing processing actions or is in maintenance mode, the robot entering the pick-and-place area may interfere with the moving parts of the equipment or affect the processing accuracy. The interlock permission signal enables state synchronization and mutual exclusion control between the processing equipment and the robot, avoiding conflicts and collisions.

[0068] A service urgency index is used to quantify the service priority of each processing device. The reason for introducing the service urgency index is that the overhead rail robot needs to allocate service resources among multiple processing devices. The demand for robot services varies among different processing devices at different times. A quantified urgency index can map multi-dimensional service demands into a single priority value, providing a clear objective function for subsequent path optimization algorithms. This ensures the robot prioritizes serving the most urgent processing device, thereby reducing the overall system's buffer overflow and material shortage risks. The service urgency index is calculated by multiplying a preset first weighting coefficient by the outbound buffer urgency, adding a preset second weighting coefficient by the inbound buffer urgency, and then adding a preset third weighting coefficient by the processing completion time urgency. The reason for using a weighted summation method to calculate the service urgency index is that outbound cache urgency, inbound cache urgency, and processing completion time urgency reflect different dimensions of service demand. Outbound cache urgency reflects the urgency of material retrieval, inbound cache urgency reflects the urgency of material release, and processing completion time urgency reflects the time pressure of timely unloading after the processing equipment is about to complete processing. These three dimensions are independent of each other but jointly affect service priority. Weighted summation can comprehensively consider multi-dimensional demands and adjust the relative importance of each dimension through weight coefficients. To ensure the comparability of various indicators and avoid dimensional mixing, the outbound cache urgency, inbound cache urgency, and processing completion time urgency are all normalized, with values ​​ranging from zero to one. The reason for using normalization is that the original dimensions and numerical ranges of the three urgency indicators are different. The original unit of the outbound buffer urgency and inbound buffer urgency is the number of material items, while the original unit of the processing completion time urgency is time in seconds. If a weighted sum is directly applied, the values ​​of different dimensions cannot be reasonably compared. Normalization maps all indicators to a dimensionless range of zero to one, making the indicators numerically comparable, and the adjustment effect of the weighting coefficients is more intuitive. The default value of the first weighting coefficient is 1.0, the default value of the second weighting coefficient is 0.8, and the default value of the third weighting coefficient is 1.2. The default values ​​of the above three weighting coefficients are determined according to the sensitivity of the production system to the three risks of outbound accumulation, inbound material shortage, and processing waiting. They can be adjusted by statistically analyzing the frequency and impact of various risk events through historical operating data. When the outbound accumulation risk is high, the first weighting coefficient should be increased; when the inbound material shortage risk is high, the second weighting coefficient should be increased; and when the capacity loss caused by processing waiting is large, the third weighting coefficient should be increased.

[0069] The calculation method for processing completion time urgency is as follows: the remaining processing time of the processing equipment is equal to the difference between the estimated processing completion time and the current time. If the difference is negative or zero, the remaining processing time is zero. The processing completion time urgency is calculated by dividing the preset processing completion time baseline by the remainder, where the remainder is the remaining processing time plus the preset processing completion time baseline. The default value of the preset processing completion time baseline is 5 seconds. This default value is determined based on the processing cycle and control period of the processing equipment, ensuring that the processing completion time urgency approaches one when the remaining processing time is close to zero and approaches zero when the remaining processing time is much greater than the baseline, thereby avoiding numerical divergence and reducing sensitivity to time estimation noise.

[0070] Outbound buffer urgency indicates the degree to which the outbound buffer exceeds the warning threshold. It is calculated by subtracting the preset outbound buffer warning threshold from the current outbound buffer occupancy of the processing equipment. If the difference is positive, it is divided by the outbound buffer capacity of the processing equipment to obtain a normalized result. If the difference is negative or zero, the outbound buffer urgency is zero, and any part of the normalized result greater than one is truncated to one. Since the outbound buffer capacity is a positive integer greater than zero, this division operation will not result in a denominator of zero. The default value of the preset outbound buffer warning threshold is 80% of the outbound buffer capacity of the processing equipment. This default value is determined based on the processing cycle time of the equipment and the robot service cycle. When the robot service cycle is long, this threshold should be appropriately reduced to reserve more buffer space.

[0071] The inbound buffer urgency indicates the degree to which the inbound buffer is below the warning threshold. It is calculated by subtracting the current inbound buffer occupancy of the processing equipment from the preset inbound buffer warning threshold. If the difference is positive, it is divided by the inbound buffer capacity of the processing equipment to obtain a normalized result. If the difference is negative or zero, the inbound buffer urgency is zero, and any part of the normalized result greater than one is truncated to one. Since the inbound buffer capacity is a positive integer greater than zero, this division operation will not result in a denominator of zero. The default value of the preset inbound buffer warning threshold is 20% of the processing equipment's inbound buffer capacity. This default value is determined based on the stability of upstream material supply and the processing speed of the equipment. When upstream material supply fluctuates significantly or the processing speed of the equipment is high, this threshold can be appropriately increased to ensure the continuity of material supply.

[0072] Step 200: Collect the operating status information of each processing device at a preset sampling period, calculate the abnormal probability of each processing device based on multi-source signals, divide the processing devices into a fault isolation set and a serviceable set according to the abnormal probability and operating status information, and perform a dynamic switching operation when any processing device is transferred from the serviceable set to the fault isolation set.

[0073] Operating status information of each processing device is collected at a preset sampling period. The default value of the preset sampling period is 100 milliseconds. This default value is determined based on the rate of change of the processing device's status and computing power. When the processing device's status changes rapidly, the sampling period should be shortened to improve the real-time performance of status perception. When computing resources are limited, the sampling period can be appropriately extended to reduce the computational load. The reason for using periodic sampling is that the operating status of the processing device is a continuously changing physical process, while the control system requires discrete digital signals for decision-making. By sampling at a fixed period, continuous states can be converted into discrete state sequences. At the same time, the selection of the sampling period needs to balance the real-time performance of status perception and the consumption of computing resources. Too short a sampling period will generate a large amount of redundant data and increase the computational burden, while too long a sampling period may miss key status changes, leading to response delays. The scheduling and solution period is set to control the execution frequency of the ant colony algorithm path optimization. The scheduling and solution period is an integer multiple of the preset sampling period, with a default value of 500 milliseconds. This default value is determined based on the computational complexity and real-time requirements of the ant colony algorithm. By setting the scheduling and solution period to an integer multiple of the sampling period, state sampling and path optimization are decoupled. State sampling continuously updates the system state at a higher frequency, while path optimization performs computationally intensive solutions at a lower frequency. This ensures real-time state awareness while reserving sufficient computation time for the ant colony algorithm. The operating status information includes the processing equipment status, inbound buffer occupancy, outbound buffer occupancy, and estimated processing completion time. The processing equipment status includes states such as processing, waiting to be loaded, waiting to be unloaded, fault, maintenance, and idle. Simultaneously, the robot's current position is collected, and the movement time required for the robot to move from its current position to each processing equipment is calculated based on the track layout.

[0074] The anomaly probability of each processing device is calculated based on multi-source signal fusion. The anomaly probability ranges from zero to one. The fused multi-source signals include processing device fault codes, abnormal temperature signals, abnormal vibration signals, sudden changes in processing cycle time, and the number of interlock rejections. Multi-source signal fusion is used because single signals often lead to false alarms and false negatives. Fault symptoms in processing devices typically manifest as coordinated anomalies in multiple physical parameters. By fusing information from multiple dimensions, such as processing device fault codes, abnormal temperature signals, abnormal vibration signals, sudden changes in processing cycle time, and the number of interlock rejections, the device status can be cross-verified from different angles, improving the accuracy and robustness of anomaly detection and reducing the probability of misjudgments caused by single sensor failures or environmental noise. The reason for choosing these five types of signals as input is that they represent different dimensions of abnormality in processing equipment. The fault code of the processing equipment is the direct output of the equipment's self-diagnosis system. The abnormal temperature signal reflects the deviation of the equipment's thermodynamic state. The abnormal vibration signal reflects the abnormality of the equipment's mechanical motion. The sudden change in processing cycle time reflects the fluctuation of the equipment's process performance. The number of interlock rejections reflects the abnormal frequency of interaction between the equipment and the robot. These five types of signals cover multiple levels such as electrical, thermal, mechanical, process, and interaction, and can comprehensively characterize the health status of the equipment. The specific calculation method for the anomaly probability is as follows: First, normalize each multi-source signal. The normalized value of the processing equipment fault code signal is one when the processing equipment reports a fault code, and zero otherwise. The normalized value of the temperature anomaly signal is the difference between the current temperature and the upper limit of the normal temperature range divided by the preset temperature anomaly range, and then the cutoff value between zero and one. The normalized value of the vibration anomaly signal is the difference between the current vibration amplitude and the normal vibration threshold divided by the preset vibration anomaly range, and then the cutoff value between zero and one. The normalized value of the processing cycle change signal is the absolute value of the deviation between the current cycle and the historical average cycle divided by the historical average cycle, and then the cutoff value between zero and one. The normalized value of the interlock rejection signal is the number of interlock rejections in the most recent preset statistical window divided by the preset maximum rejection number, and then the cutoff value between zero and one. The upper limit of the normal temperature range, the normal vibration threshold, the preset abnormal temperature range, and the preset abnormal vibration range are derived from the factory parameters of the processing equipment, the process specifications, or statistical quantiles based on historical operating data. Specifically, the rated operating temperature upper limit is read from the technical specifications of the processing equipment as the upper limit of the normal temperature range, and the rated vibration amplitude upper limit is read as the normal vibration threshold. The preset abnormal temperature range is set to 20% of the upper limit of the normal temperature range as the default value, and the preset abnormal vibration range is set to 30% of the normal vibration threshold as the default value. The above threshold and range parameters can be dynamically updated according to the preset parameter update cycle during system operation. The default value of the preset parameter update cycle is 24 hours. The update method is to calculate the 95th quantile based on the temperature and vibration data of the processing equipment under normal operating conditions in the most recent 24 hours, and use this quantile as the updated upper limit of the normal temperature range and the normal vibration threshold.Then, the normalized multi-source signals are multiplied by their respective preset weighting coefficients and summed. The default values ​​for the weighting coefficients of the processing equipment fault codes, temperature anomalies, vibration anomalies, clockwise abrupt changes, and interlock rejections are 0.15 and 0.15, respectively. The sum of these weighting coefficients is one. The default values ​​of each weighting coefficient are determined based on the indication intensity of the various abnormal signals for processing equipment faults and can be adjusted by statistically analyzing the correlation between each signal and the actual fault using historical fault data.

[0075] Fault determination is based on anomaly probability and processing equipment status. Processing equipment is divided into a fault isolation set and a serviceable set. The determination rule is: if the anomaly probability of a processing equipment is greater than or equal to a preset anomaly probability threshold, or if the processing equipment status is faulty, then the processing equipment is added to the fault isolation set; otherwise, it is added to the serviceable set. The default value of the preset anomaly probability threshold is 0.7. This default value is determined based on the production system's tolerance for false positives and false negatives. When the system's tolerance for false negatives is low, this threshold should be lowered to improve the fault detection rate; when the system's tolerance for false positives of normal processing equipment is low, this threshold should be raised to reduce the false alarm rate. The serviceable set and the fault isolation set are complementary sets; that is, the serviceable set includes all processing equipment out of the N processing equipment that does not belong to the fault isolation set. To avoid scheduling policy jitter caused by frequent transitions between the fault isolation set and the serviceable set of processing equipment, a preset minimum hold time for set switching is set, with a default value of 10 seconds. When a processing equipment is transferred from the serviceable set to the fault isolation set, even if the abnormal probability of the processing equipment decreases to below the preset abnormal probability threshold within the preset minimum hold time for set switching, it will still maintain its state in the fault isolation set until the preset minimum hold time for set switching has elapsed before its set affiliation is allowed to be reassessed. This introduces a time hysteresis mechanism to suppress policy jitter caused by state transitions.

[0076] Simultaneously, two action sequences are maintained: an execution sequence and a shadow sequence. The reason for employing this dual-sequence mechanism is that equipment failures in the production system are random and sudden. If path replanning is initiated only after a failure occurs, the time-consuming replanning calculations will cause the robot to pause and wait, affecting the system's continuity and efficiency. By maintaining two parallel action sequences—the execution sequence and the shadow sequence—the shadow sequence for the next round is pre-calculated in the background while the execution sequence is running. When a failure occurs, the shadow sequence can be used immediately for a quick switch, and the full replanning process is only triggered when the shadow sequence becomes unavailable. This significantly reduces failure response latency and improves the system's dynamic adaptability. The execution sequence is the action sequence to be issued to the overhead rail robot for execution. Its length is the number of steps specified by the preset rolling planning window length, and it includes specific material picking, material placement, movement, or waiting actions and the corresponding target processing equipment. The shadow sequence is the pre-calculated optimal action sequence for the next round, assuming the current fault isolation set remains unchanged. The shadow sequence is generated immediately after the execution sequence begins execution, with background calculations initiated to ensure a rapid response when a switch is needed. During the execution of the sequence, changes in the serviceable set are continuously monitored. When any processing equipment is detected to have moved from the serviceable set to the fault isolation set, a dynamic switching operation is performed. The specific steps of the dynamic switching operation are as follows: First, determine the number of actions that have been completed in the execution sequence and obtain the current state information accordingly. The state information includes the actual state of the overhead robot, the robot carrier occupancy, and the inbound and outbound buffer occupancy of each processing equipment. Then, starting from the first action in the shadow sequence, perform action feasibility verification and state prediction sequentially. The action feasibility verification includes at least hard constraints and removes actions involving newly faulty processing equipment. Find the longest consecutive executable prefix subsequence in the shadow sequence that meets the above conditions. If the length of the prefix subsequence is greater than zero, replace the unexecuted actions in the execution sequence with the prefix subsequence; if the length of the prefix subsequence is zero, immediately trigger the ant colony algorithm to replan and generate a new execution sequence. Through the above mechanism, dynamic rerouting without interrupting the operation is achieved.

[0077] Step 300: Based on the service urgency index, the ant colony algorithm is used to globally optimize the service path of the overhead rail robot. Each processing device in the service set is mapped to a node in the ant colony algorithm, and the robot's movement path between processing devices is mapped to an edge between nodes. An action sequence is generated and executed through iterative search and pheromone update.

[0078] This step employs the ant colony optimization algorithm to globally optimize the service path of the overhead rail robot, generating the optimal action sequence. The ant colony algorithm is chosen because the service path optimization problem for the overhead rail robot is essentially a constrained multi-objective combinatorial optimization problem. It requires finding the optimal access order and action sequence for processing equipment while satisfying cache capacity constraints and interlocking security constraints, comprehensively considering multiple objectives such as service urgency index, movement cost, and waiting cost. The solution space of this problem grows exponentially with the number of processing equipment, and traditional exact algorithms struggle to solve it within the time required for real-time control. The ant colony algorithm, as a heuristic optimization algorithm, utilizes the parallel search of multiple virtual ants and the positive feedback mechanism of pheromones to quickly converge to a high-quality near-optimal solution within a finite number of iterations. Furthermore, the algorithm's randomness and diversity give it strong global search capabilities, preventing it from getting trapped in local optima, making it suitable for handling such complex dynamic scheduling problems. The ant colony algorithm is a swarm intelligence optimization algorithm that simulates the foraging behavior of ants in nature. Through the cooperative search of multiple virtual ants and the positive feedback mechanism of pheromones, it can find high-quality solutions to complex combinatorial optimization problems. The working principle of the ant colony algorithm is that each ant constructs candidate solutions based on pheromone concentration and heuristic information. Pheromone concentration reflects the quality of the path in historical search experience, while heuristic information reflects the immediate attractiveness of the path in the current state. After constructing the path, the ant releases pheromones on the edges it passes through based on the path quality. High-quality paths receive more pheromone deposition, and paths with high pheromone concentration are more likely to be selected in subsequent iterations, thus forming a positive feedback loop that guides the search toward the optimal solution. At the same time, the natural evaporation mechanism of pheromones prevents the algorithm from converging too early and maintains a certain level of exploration capability.

[0079] First, the ant colony algorithm is initialized and configured. A preset number of ants is set, which determines the number of paths searched in parallel during each iteration. The default value is 20 ants, determined based on the number of processing devices that can be served. Typically, this number is 2-3 times the number of processing devices to ensure search diversity. When the number of processing devices is large, the number of ants can be increased to expand the search range. A preset maximum number of iterations is set to control the algorithm's termination condition. The default value is 100 iterations, determined based on the solution space size and computation time constraints. When the solution space is large or the solution quality requirement is high, the number of iterations can be increased; when real-time performance is high, the number of iterations can be decreased. A preset pheromone importance coefficient and a preset heuristic information importance coefficient are set. These two coefficients are used to adjust the ants' dependence on pheromone concentration and heuristic information when choosing a path, respectively. The default value of the preset pheromone importance coefficient is 1.0, and the default value of the preset heuristic information importance coefficient is 2.0. The default values ​​of these two coefficients are determined based on the balance requirements of algorithm convergence speed and solution diversity. When it is necessary to speed up convergence, the preset pheromone importance coefficient can be increased; when it is necessary to enhance search diversity, the preset heuristic information importance coefficient can be increased. A preset pheromone volatilization coefficient is set to simulate the natural dissipation of pheromones over time. Its value ranges from zero to one, with a default value of 0.1. This default value is determined based on the algorithm's requirement for the rate of forgetting historical experience. When it is necessary to quickly adapt to environmental changes, this coefficient can be increased to accelerate pheromone decay; when it is necessary to maintain search stability, this coefficient can be decreased to prolong pheromone memory. Set a preset pheromone increment constant to calculate the amount of pheromone released after the ant completes the path. Its default value is 100. This default value is determined based on the numerical range of pheromone concentration. The amount of pheromone released should be on the same order of magnitude as the initial pheromone concentration to ensure the effectiveness of pheromone renewal.

[0080] A problem model for the ant colony algorithm is constructed, mapping each processing device in the serviceable set to a node in the ant colony algorithm, and mapping the robot's movement path between processing devices to edges between nodes. The pheromone concentration on each edge is initialized to a preset initial pheromone concentration value, with a default value of 1.0. The principle behind setting this default value is to choose a moderate positive number so that each edge has a similar probability of being selected in the early stages of the algorithm, avoiding initial bias from affecting the search results. The heuristic pheromone value of each edge is calculated, taking into account the service urgency index of the target processing device, movement cost, and waiting cost.

[0081] To ensure that the service sequence output by the ant colony algorithm can be implemented as an executable sequence of pick-and-place actions, a mechanism for action feasibility constraints and state prediction is further introduced into the problem model. The reason for introducing these mechanisms is that the standard ant colony algorithm only addresses the optimization of the node access order. However, the actual control of the overhead rail robot requires not only determining which processing equipment to access, but also specifying what actions to perform on each equipment and whether those actions are feasible in the current state. The robot's carrier capacity limitations, the buffer state of the processing equipment, and the interlock signal state all affect the executability of the actions. If these constraints are not considered during the path construction phase, the generated service sequence may fail to be implemented due to constraint violations. Therefore, a state prediction mechanism needs to be embedded in the path construction process of the ant colony algorithm to perform feasibility checks on each action step, ensuring that the generated action sequence is physically executable. The robot carrier capacity is set to represent the number of workpieces the robot can carry simultaneously, and the robot carrier occupancy is set to represent the number of workpieces currently being carried. The robot carrier occupancy must be greater than or equal to zero and not exceed the robot carrier capacity. The default value for the robot carrier capacity is 1. The reason for introducing robot carrier capacity and robot carrier occupancy is that the robot's material carrying capacity is limited. The robot can only perform material picking actions when there is spare capacity on the carrier, and can only perform material unloading actions when there is material on the carrier. By tracking the dynamic changes in robot carrier occupancy, the feasibility of material picking and unloading actions can be judged during path construction, avoiding the generation of action sequences that violate carrier capacity constraints. The feasibility conditions for material picking are set as follows: the outbound buffer occupancy of the target processing equipment is greater than zero and the interlock permission signal is in the allowed state, and the robot carrier occupancy is less than the robot carrier capacity; the feasibility conditions for unloading are set as follows: the inbound buffer occupancy of the target processing equipment is less than the inbound buffer capacity and the interlock permission signal is in the allowed state, and the robot carrier occupancy is greater than zero; the feasibility conditions for combined picking and unloading are set as follows: the same target processing equipment simultaneously meets the feasibility conditions for unloading and picking at the same arrival time, and the execution order of the combined picking and unloading actions is unloading first and then picking.

[0082] State prediction is used to perform forward calculations of the state at the arrival time when constructing candidate solutions. The reason for using state prediction is that the robot needs a certain amount of time to move from its current position to the target processing equipment. During this time, the state of the processing equipment may change. For example, the equipment currently processing may complete processing, the buffer occupancy may change due to other operations, and the interlock permission signal may change due to equipment state switching. If the feasibility of the action is judged only based on the current state, the actual state when the robot arrives may not meet the execution conditions, resulting in action failure or waiting. Therefore, it is necessary to predict the state at the future arrival time based on the movement time and waiting time, and to perform action feasibility verification under the predicted state to ensure that the generated action sequence is feasible at the actual execution time. The calculation method for state prediction is as follows: based on the current position of the overhead rail robot and the robot's vehicle occupancy, the estimated arrival time for each step is obtained by accumulating the movement cost and waiting cost according to the candidate action sequence. At each step's arrival time, the feasibility of the action is verified by combining the interlock permission signal, the inbound buffer occupancy, the outbound buffer occupancy, and the estimated processing completion time. If a step is a material picking action, the outbound buffer of the corresponding processing equipment is reduced by a preset single picking quantity, and the robot carrier's occupancy is increased by a preset single picking quantity. If a step is a material unloading action, the inbound buffer of the corresponding processing equipment is increased by a preset single unloading quantity, and the robot carrier's occupancy is reduced by a preset single unloading quantity. If a step is a combined picking and unloading action, the unloading status update is performed first, followed by the picking status update. The default values ​​for both the preset single picking quantity and the preset single unloading quantity are 1.

[0083] The movement cost represents the time required for the robot to move from its current position to the processing equipment, and its value is equal to the movement time calculated based on the robot's current position. The waiting cost represents the time the robot needs to wait for the processing equipment to complete after arriving at the equipment. It is calculated by subtracting the sum of the current time and the movement cost from the estimated completion time of the processing equipment. If the difference is positive, it is taken as the waiting cost; if the difference is negative or zero, the waiting cost is zero.

[0084] The specific calculation method for heuristic information values ​​employs a ratio form to ensure that the values ​​are always positive and have good discriminative power. The reason for using a ratio form is that the ant colony algorithm requires heuristic information values ​​to be positive in probability calculations. However, the original numerical ranges of the service urgency index and time cost differ significantly and may even contain zero values. Using simple difference or linear combination forms could result in negative or zero values, causing the algorithm to fail. The ratio form naturally guarantees a positive result, and the relative change in the numerator and denominator generates good numerical discriminative power, widening the gap between the heuristic information values ​​of different candidate processing equipment and enhancing the algorithm's selection bias. The time cost item is calculated as the sum of movement cost and waiting cost plus a preset time cost baseline value. The default value of the preset time cost baseline value is 1.0 second. The principle for setting this default value is to take a moderately positive number to avoid the time cost item being zero, which would cause division anomalies, while also keeping the numerical range of the time cost item reasonable. The service urgency factor is calculated by adding a preset urgency benchmark value to the service urgency index. Dividing the service urgency factor by the time cost factor yields the heuristic information value. The default value for the preset urgency benchmark is 0.1. This default value is set to ensure that the heuristic information value is still positive even when the service urgency index is zero, giving all candidate processing equipment a chance of being selected. Through this ratio calculation method, processing equipment with higher service urgency indices or lower time costs will receive larger heuristic information values, thus having a higher probability of being selected in ant path selection.

[0085] The ant colony algorithm performs an iterative search process. In each iteration, each ant starts from the robot's current position and sequentially selects the next target processing device node until a complete service path containing the number of steps specified by the preset rolling planning window length is constructed. The ants employ a probabilistic selection mechanism when selecting the next target processing device node. This mechanism is used because deterministic selection would cause all ants to choose the currently optimal edge, quickly converging to a local optimum. The probabilistic selection mechanism allows each ant to calculate the probability of selecting each candidate edge based on pheromone concentration and heuristic information values. Even suboptimal edges have a certain probability of being selected, thus maintaining search diversity and enabling the algorithm to explore a wider solution space, avoiding premature entrapment in local optima. For each candidate processing device of the current ant, the algorithm first calculates whether there is at least one executable action at the expected arrival time based on state prediction. Executable actions include material handling, material placement, or a combination of both. Candidate processing devices that do not meet the action feasibility constraints are eliminated. Then, the probability of selection for the remaining candidate processing devices is calculated. The pheromone concentration along the edge from the current node to the candidate processing device is increased to a power of a preset pheromone importance coefficient to obtain an enhanced pheromone concentration. The heuristic pheromone value of this edge is then increased to a power of a preset heuristic pheromone importance coefficient to obtain an enhanced heuristic pheromone value. The enhanced pheromone concentration multiplied by the enhanced heuristic pheromone value is used as the selection tendency value for the candidate processing device. This selection tendency value is then divided by the sum of the selection tendency values ​​of all candidate processing devices; the quotient is the probability that the candidate processing device will be selected. Ants select the next target processing device node using a roulette wheel approach based on the above probability distribution. If no candidate processing device satisfies the action feasibility constraint at a certain step, a waiting action is inserted to advance the current time until an executable action exists or a replanning condition is triggered.

[0086] After each ant completes a full path, the overall fitness value of that path is calculated. The path fitness value is calculated by first calculating the total service urgency of the path, summing the service urgency indices of the target processing equipment corresponding to each step in the path. Then, the total time cost of the path is calculated, summing the movement cost and waiting cost of each step in the path. The path fitness value is the total service urgency divided by the total time cost, which is the sum of the total time cost and a preset fitness time baseline value. The default value of the preset fitness time baseline value is 1.0 second. This default value is set to avoid division anomalies when the total time cost is zero, while also keeping the fitness value range reasonable. Paths with higher total service urgency or lower total time cost will obtain a higher fitness value through this ratio calculation method. The path with the highest fitness value in the current iteration is recorded as the optimal path for this round and compared with the historical globally optimal path. If the fitness value of the optimal path in this round is higher than that of the globally optimal path, then the globally optimal path is updated as the optimal path for this round.

[0087] After one iteration, a pheromone update operation is performed. First, pheromone evaporation is executed, multiplying the pheromone concentration on all edges by a decay term. The decay term is the difference between one and a preset pheromone evaporation coefficient, causing the pheromone concentration to decrease proportionally. The reason for pheromone evaporation is to simulate the natural dissipation of pheromones over time, preventing excessive accumulation of historical pheromones that could lead the algorithm to early suboptimal solutions. The pheromone evaporation mechanism gradually dilutes the pheromones deposited in earlier iterations, leaving pheromone concentration space for better paths discovered in subsequent iterations, thus maintaining the algorithm's exploratory ability and adaptability to environmental changes. Next, pheromone release is performed. For each path traversed by an ant, pheromones are added to each edge contained in the path. The amount of pheromone added is the preset pheromone increment constant multiplied by the fitness value of the ant's path; the higher the path fitness value, the more pheromones are released. The reason for using a pheromone release amount proportional to the fitness value is to achieve a positive feedback mechanism. High-quality paths receive more pheromone deposition, making them more likely to be selected in subsequent iterations, thus guiding the search towards higher-quality solutions. Low-quality paths receive less pheromone and are gradually eliminated from the search range. Furthermore, additional pheromone is added to the globally optimal path. The amount added is calculated by multiplying a preset pheromone increment constant by a preset elite ant weight coefficient by the fitness value of the globally optimal path. The default value of the preset elite ant weight coefficient is 2.0, determined based on the required strengthening of the globally optimal solution. This coefficient can be increased to accelerate algorithm convergence and decreased to maintain search diversity. This elite strategy strengthens the pheromone concentration on the optimal path, guiding subsequent ants to search towards the optimal solution. The reason for using an elite strategy is that the globally optimal path represents the best solution found in the current search process. Adding extra pheromone to it accelerates the algorithm's convergence towards the region near the optimal solution. Simultaneously, adjusting the elite ant weight coefficient balances convergence speed and search diversity.

[0088] Repeat the iterative search and pheromone update process described above until the preset maximum number of iterations is reached or the convergence condition is met. The convergence condition is that the globally optimal path remains unchanged for a certain number of consecutive iterations. This certain number of consecutive iterations is the preset number of convergence judgment rounds, with a default value of 10 rounds. This default value is determined based on the algorithm's stability requirements. When the solution space is large or a more thorough search is needed, the number of rounds can be increased; when real-time requirements are high, the number of rounds can be decreased. After the algorithm terminates, the globally optimal path is output as the optimal service sequence for the sky-rail robot.

[0089] Based on the optimal service sequence output by the ant colony algorithm, the target processing equipment for each step is determined sequentially. At each step, based on the expected arrival time, the feasibility of the action is verified and the action type is determined. The action type includes material handling, material release, or a combination of both. The generated actions are added to the action sequence. If no executable action type exists at the expected arrival time for a certain step, that step is marked as infeasible, and the ant colony algorithm is triggered to replan and generate a new action sequence. Executing an action sequence updates the overhead rail robot, the robot carrier's occupancy, and the inbound and outbound buffer occupancy of the relevant processing equipment. When all actions in the execution sequence have been completed or the environmental state has changed significantly, the ant colony algorithm is invoked again for path optimization, generating a new action sequence to achieve rolling optimization control.

[0090] Step 400: Perform isolation operations on the processing equipment within the fault isolation set, monitor the downstream accumulation risk caused by the fault, activate the virtual back pressure control mechanism, and smoothly integrate the recovered processing equipment.

[0091] Isolate the processing equipment within the fault isolation set, prohibiting the overhead rail robot from entering the pick-and-place interlock area of ​​the faulty processing equipment, and automatically bypass the faulty processing equipment at the motion planning level. The reason for performing the isolation operation is that the faulty processing equipment cannot complete the processing task normally. Continuing to feed or pick up materials from it is an ineffective operation, which will only occupy the robot's service time and consume buffer resources. By isolating the faulty processing equipment from the serviceable set, the robot can concentrate its limited service resources on the normally operating processing equipment, thereby maintaining the overall capacity of the system. The prohibition of the overhead track robot from entering the pick-up / placement interlock zone of the faulty processing equipment is implemented through a two-layer mechanism. The first layer is scheduling layer isolation. During the ant colony algorithm path construction process in step 300, the processing equipment in the fault isolation set is removed from the candidate node set, preventing it from participating in path planning. This avoids access to the faulty processing equipment during the action sequence generation stage. The second layer is interlock layer hard prohibition. When the overhead track robot attempts to initiate a pick-up / placement request to the processing equipment in the fault isolation set for any reason, the interlock permission signal of that processing equipment is forcibly set to the prohibited state, physically preventing the robot from entering the pick-up / placement interlock zone. Through the dual insurance mechanism of scheduling layer soft isolation and interlock layer hard prohibition, the isolation operation of the faulty processing equipment is strictly executed both logically and physically, ensuring the safety of system operation. The system monitors the downstream backlog risk caused by the fault. When the ratio of the outbound buffer occupancy of any processing equipment to the outbound buffer capacity of that processing equipment is detected to be greater than or equal to a preset backpressure trigger threshold, a virtual backpressure control mechanism is activated. The reason for monitoring downstream backpressure risk is that when a processing device malfunctions, its upstream processing devices continue to operate normally. If the upstream feeding cycle is not adjusted in time, the outbound buffer of the upstream processing device will accumulate rapidly until it overflows, leading to a chain reaction of failures. Therefore, it is necessary to detect backpressure risk in a timely manner and activate the flow control mechanism by monitoring the outbound buffer occupancy. The default value of the preset backpressure trigger threshold is 0.9. This default value is determined based on the early warning amount for buffer overflow risk. When it is necessary to intervene in flow control earlier to avoid buffer overflow, this threshold should be lowered. When the system buffer margin is large, this threshold can be appropriately increased. The virtual backpressure control mechanism includes calculating the backpressure coefficient. The backpressure coefficient is calculated by subtracting the buffer occupancy item from the calculated value for each processing device in the serviceable set. The buffer occupancy item is the ratio of the outbound buffer occupancy of the processing device to its outbound buffer capacity. If the result is negative, it is taken as zero. Then, the minimum value is taken among the above calculation results of all processing devices in the serviceable set. This minimum value is the backpressure coefficient. The value range of the backpressure coefficient is between zero and one.The reason for adopting the virtual backpressure control mechanism is that it draws on the concept of backpressure in fluid mechanics. By adjusting the upstream feed rate in reverse according to the downstream buffer status, it achieves global flow balance. The backpressure coefficient reflects the remaining capacity ratio of the processing equipment with the most strained buffer in the system. When the backpressure coefficient is close to zero, it means that at least one processing equipment's outbound buffer is close to full load. At this time, it is necessary to significantly reduce the upstream feed rate or even stop feeding. When the backpressure coefficient is close to one, it means that all processing equipment's outbound buffers have sufficient margin. At this time, the normal feed rate can be maintained. This dynamic adjustment mechanism can avoid buffer overflow during faults while maintaining the system's maximum possible capacity. To avoid excessive impact of instantaneous buffer fluctuations of a single processing device on the back pressure coefficient, the outbound buffer occupancy of each processing device is calculated using a sliding average value within a preset back pressure smoothing window. The default value of this preset back pressure smoothing window is the last 5 sampling periods. This default value is determined based on the buffer occupancy fluctuation characteristics and back pressure response sensitivity requirements. The sliding average process filters out high-frequency noise, making the change in the back pressure coefficient more stable and avoiding unnecessary material supply adjustments triggered by single-point abnormal data. At the same time, processing devices in the fault isolation set are automatically excluded when calculating the back pressure coefficient, and the calculation is only based on the normally operating processing devices in the serviceable set, avoiding interference from the abnormal buffer state of faulty devices with back pressure control decisions. The back pressure coefficient achieves flow control by adjusting the upstream feeding cycle time. Specifically, when the back pressure coefficient is greater than or equal to the preset back pressure stop threshold, the standard feeding interval time of the upstream feeding processing equipment is divided by the back pressure coefficient to obtain the adjusted feeding interval time. When the back pressure coefficient is less than the preset back pressure stop threshold, the upstream feeding is suspended until the back pressure coefficient recovers. The default value of the preset back pressure stop threshold is 0.1. This default value is determined based on the system's tolerance for buffer overflow risk. When buffer overflow would lead to serious consequences, this threshold should be increased to trigger the feeding suspension earlier.

[0092] When a processing device in the fault isolation set reports a recovery signal, or its anomaly probability is less than a preset anomaly probability threshold, the processing device is transferred from the fault isolation set to the observation recovery set. The reason for setting up the observation recovery set is that after a processing device recovers from a fault state, its operational stability requires a period of observation and verification. If it is immediately included in the serviceable set for normal scheduling, secondary failures may occur because the device is not yet fully stable, affecting the continuity of system operation. By setting up the observation recovery set as a transitional state from the fault isolation set to the serviceable set, the stability of the recovered processing device can be observed, reducing the risk of secondary failures and improving system reliability. A stability window determination is performed on the processing devices in the observation recovery set. If a processing device simultaneously meets two conditions within a continuous preset stability observation window: the processing device state is not in a fault state, and the interlock allowance rate is greater than or equal to a preset allowance rate threshold, then the processing device is allowed to be included in the serviceable set. The reason for using the stability window determination is that by setting a continuous time window and requiring the stability conditions to be continuously met within that window, transient fluctuations in normal state can be filtered out, ensuring that the processing device has indeed recovered to a stable operating state, rather than being a random momentary normal operation, thereby improving the reliability of the recovery determination. The default value for the preset stability observation window duration is 30 seconds. This default value is determined based on the stability observation requirements after the processing equipment has recovered from a fault. If secondary faults are likely to occur after the processing equipment recovers, this duration should be extended. If the processing equipment has good stability after recovery, this duration can be shortened to expedite the return to operation. The interlock tolerance rate is set as the ratio of the number of successful pick-and-place requests to the total number of requests within the preset stability observation window duration. The default value for the preset tolerance rate threshold is 0.95, and its range is greater than zero and less than or equal to one. This default value is determined based on the reliability requirements after the processing equipment has recovered. When the reliability requirements for the recovered processing equipment are high, this threshold should be increased. When rapid recovery of production capacity is required, this threshold can be appropriately decreased.

[0093] For processing equipment that recovers after passing the stability window criterion, a gradual quota mechanism is used for smooth inclusion. The reason for using this mechanism is that although processing equipment has passed the stability window criterion after recovering from a fault, its operating state may not yet be fully stable. If it is immediately put into full-load scheduling to receive the same service frequency as normal equipment, sudden load changes may lead to secondary failures or affect processing quality. The gradual quota mechanism increases the service weight of recovering processing equipment from zero to one, and its scheduling probability smoothly transitions from low to high, thus achieving gradual load recovery, reducing the risk of secondary failures, and giving the equipment sufficient warm-up time. In the ant colony algorithm's path selection process, the selection tendency value of each candidate processing equipment at each step can be regarded as its single-step score. The single-step score comprehensively reflects factors such as the service urgency index, movement cost, and waiting cost of the processing equipment. The larger the value, the stronger the tendency for the processing equipment to be prioritized for service in the current step. For normal processing equipment in the serviceable set, its scheduling score is directly equal to its single-step score, and it participates in path construction according to its single-step score in the ant colony algorithm's probabilistic selection mechanism. For recovery processing equipment that has just passed the stability window assessment, in order to avoid sudden changes in the cycle time caused by it immediately participating in scheduling at full load, it is necessary to adjust its scheduling score by weight.

[0094] To assign service weights to the processing equipment for recovery, a ramp function is used to calculate the service weights. The calculation method involves subtracting the time it took for the processing equipment to pass the stability window from the current time, then dividing by a preset ramp duration parameter. If the ratio is greater than one, it is set to one; otherwise, the ratio is used as the service weight. The reason for using a ramp function to calculate the service weight is that it achieves a linear and smooth transition from zero to one. Its mathematical form is simple and its physical meaning is clear. The linear increase in service weight over time reflects the gradual recovery process of the processing equipment's load capacity, avoiding the sudden load changes that might occur with a step function. Simultaneously, the linear relationship ensures a constant ramp rate, facilitating intuitive control of the recovery process's time span based on the preset ramp duration parameter. The default value for the preset ramp duration parameter is 60 seconds. This default value is determined based on the need for smooth integration of the processing equipment into the scheduling process. When a smoother recovery of the processing equipment's load is needed to avoid sudden changes in cycle time, this duration should be extended; when the demand for capacity recovery is urgent, this duration can be shortened. The service weight ranges from zero to one. When the recovery processing equipment just passes the stability window judgment, the service weight is zero. As time goes by, the service weight gradually increases. After the preset ramp-up time parameter, the service weight reaches one.

[0095] The scheduling score of the recovery processing equipment is updated to a weighted score. The weighted score is calculated by multiplying the service weight of the processing equipment by its single-step score. During the ant colony algorithm's path selection process, the recovery processing equipment uses the weighted score instead of the single-step score in probability calculations. Since the service weight is less than one, the weighted score will be less than the single-step score, thus reducing the probability of the recovery processing equipment being selected and causing it to participate in scheduling with a lower service frequency. As the service weight gradually increases from zero to one, the weighted score gradually approaches the single-step score, allowing the recovery processing equipment to gradually transition from limited participation in scheduling to normal competitive scheduling, avoiding abrupt changes in clock speed. When the service weight equals one, the weighted score equals the single-step score, and the recovery processing equipment's scheduling score reverts to a normal single-step score, entering a fully normal scheduling state, participating in path optimization in the same way as other normal processing equipment in the serviceable set.

[0096] Step 500: Record the entire process of each pick-up and place operation performed by the overhead rail robot and update the estimated processing completion time in the running status information.

[0097] A data acquisition mechanism is established to record the entire process of each pick-up and place operation performed by the overhead track robot. This mechanism is necessary because numerous uncertainties exist during actual system operation, including fluctuations in robot speed, changes in processing cycle time of the equipment, and delays in interlock responses. These factors cause discrepancies between the theoretical model and actual execution. By recording the actual execution data and comparing it with theoretical predictions, model deviations can be identified and dynamically corrected, thereby improving the accuracy of subsequent planning and the system's adaptability. The entire process record includes key information such as the operation start time, operation end time, target processing equipment identifier, action type, actual movement duration, actual waiting time, interlock response status, and buffer status changes. The reason for selecting this information as the recording content is that it covers the time, space, and state dimensions of the pick-and-place operation. The start and end times of the operation are used to calculate the actual execution time. The target processing equipment identifier and action type are used to associate the operation with the equipment. The actual movement time and actual waiting time are used to correct the movement time estimate and processing cycle time estimate. The interlock response status is used to evaluate the reliability of equipment interaction. The buffer status change is used to verify the correctness of material flow. This information together constitutes a complete operation profile, providing a data foundation for parameter self-calibration. The above records are stored in the operation history database, which adopts a rolling storage strategy, retaining all records within the most recent preset historical data retention period. The default value of the preset historical data retention period is 24 hours. This default value is determined based on the statistical sample size and storage resource constraints required for system parameter self-calibration. When longer-term trend analysis is required, the retention period can be extended; when storage resources are limited, the retention period can be shortened.

[0098] Dynamic correction of movement time estimation is performed based on actual execution data from the historical database. The reason for this dynamic correction is that theoretically calculated movement times are based on an ideal track layout and a constant speed. However, in actual operation, the robot's speed fluctuates due to factors such as load weight, track wear, and motor performance degradation. Continuously using theoretical values ​​for planning leads to inaccurate arrival time predictions, affecting feasibility assessments and waiting cost calculations. Dynamic correction based on historical measured data allows the movement time estimate to gradually approach the actual value, improving planning accuracy. For the movement time of the overhead track robot from any position to various processing equipment, the initial value is a theoretically calculated value based on the track layout, which is corrected during system operation based on the actual movement time. The correction method for movement time estimation is as follows: for each completed movement operation, extract the starting position, the destination processing equipment, and the actual movement time. Query historical movement records with the same starting and ending points in the operation history database and calculate the historical average movement time. The historical average movement time is the arithmetic mean of the actual movement times of the same path within the most recent preset movement time statistics window. The default value of the preset movement time statistics window is the most recent 50 movement operations. This default value is determined based on the movement time fluctuation characteristics and statistical stability requirements. When the movement time fluctuation is large, the statistics window should be increased to improve estimation stability; when a rapid response to movement time changes is required, the statistics window can be decreased to improve estimation sensitivity. The current estimated movement time is updated to a weighted average of the historical average movement time and the theoretical calculation. The weight of the historical average movement time is a preset measured weight coefficient, and the weight of the theoretical calculation is the difference between one and the preset measured weight coefficient. The default value of the preset measured weight coefficient is 0.7. This default value is determined based on the balance between the reliability of measured data and the accuracy of the theoretical model. When the quality of measured data is high, this coefficient can be increased to rely more on the measured values. When the theoretical model is more accurate, this coefficient can be decreased to maintain model constraints.

[0099] Dynamic correction of the estimated processing completion time is performed based on actual execution data from the historical database. This dynamic correction is necessary because the actual processing cycle time of the processing equipment fluctuates due to factors such as tool wear, material hardness differences, and ambient temperature changes. Furthermore, the processing cycle time may exhibit a trend change as the equipment runs. Continuously using the initial nominal cycle time for estimation would lead to inaccurate calculations of processing completion time urgency, affecting service priority judgment. Dynamic correction based on historical measured data allows the processing cycle time estimation to track actual trends, improving the accuracy of processing completion time prediction. For each processing piece of equipment, the initial value is the nominal cycle time or the historical average cycle time, which is corrected during system operation based on the actual processing completion time. The initial method for generating the estimated processing completion time is as follows: when the processing equipment starts for the first time, restarts after a model change, or begins processing for the first time after recovering from a fault, the initial estimated processing completion time is the sum of the equipment's starting time and its nominal cycle time or the historical average cycle time. The method for correcting the estimated processing completion time is as follows: Each time the processing equipment's status changes from processing to waiting for loading or processing completed, the actual processing completion time is recorded and compared with the last time processing started on that equipment, calculating the actual processing time. The historical processing records of that equipment are queried in the operation history database to calculate the historical average processing time. The historical average processing time is the arithmetic mean of the actual processing times of that equipment within the most recent preset processing time statistics window. The default value of the preset processing time statistics window is the last 30 processing operations. This default value is determined based on the processing cycle time fluctuation characteristics and statistical stability requirements. When the processing cycle time fluctuation is large, the statistics window should be increased to improve estimation stability; when a rapid response to changes in the processing cycle time is required, the statistics window can be decreased to improve estimation sensitivity. The estimated processing cycle time of that equipment is updated to the historical average processing time. When that equipment starts processing again, the estimated processing completion time is set to the current time plus the estimated processing cycle time.

[0100] An interlock rejection frequency monitoring mechanism is established to statistically analyze the interlock response status of each processing device. For each processing device in the service set, an interlock rejection counter and an interlock request counter are maintained. Whenever the overhead robot initiates a pick-up / place-out request to the processing device, the interlock request counter is incremented. If the interlock permission signal of the processing device is in a prohibited state, causing the request to be rejected, the interlock rejection counter is incremented. The interlock rejection rate is calculated using a preset interlock statistical period as a time window. When the value of the interlock request counter is greater than zero within the time window, the interlock rejection rate is the value of the interlock rejection counter divided by the value of the interlock request counter within the time window. When the value of the interlock request counter is equal to zero within the time window, the interlock rejection rate is zero, thus avoiding the case where the denominator is zero in the division operation. The default value of the preset interlock statistical period is 300 seconds. This default value is determined based on the frequency of interlock rejection events and the statistical significance requirements. When interlock rejection events are relatively sparse, the statistical period should be extended to accumulate sufficient samples. When it is necessary to quickly identify interlock anomalies, the statistical period can be shortened to improve the response speed.

[0101] A monitoring mechanism for excessively long waiting times is established to statistically analyze the actual waiting times of each processing device. For each waiting operation after the overhead robot arrives at the processing device, the actual waiting time is recorded and compared with the expected waiting time of that processing device. The expected waiting time is the waiting cost calculated in step 300. The waiting time deviation is calculated as the difference between the actual waiting time and the expected waiting time. The average waiting time deviation is calculated using a preset waiting statistical period as a time window. The average waiting time deviation is the arithmetic mean of the waiting time deviations of all waiting operations of that processing device within that time window. The default value of the preset waiting statistical period is 300 seconds. This default value is determined based on the characteristics of waiting time fluctuations and the requirements for statistical stability. When the waiting time fluctuations are large, the statistical period should be extended to improve statistical stability. When it is necessary to quickly identify waiting anomalies, the statistical period can be shortened to improve response speed.

[0102] An adaptive adjustment mechanism for anomaly identification weights is established, dynamically adjusting the weight coefficients in the anomaly probability calculation based on the interlock rejection rate and the deviation of average waiting time. The reason for establishing this adaptive adjustment mechanism is that different production systems and processing equipment exhibit different fault modes. For some equipment, the primary precursor to failure is frequent interlock rejection, while for others it is abnormal processing cycle time. Using fixed weight coefficients may not be sensitive enough to the fault symptoms of some equipment. By dynamically adjusting the weight coefficients based on the anomaly patterns observed during actual operation, the anomaly identification mechanism can adapt to the fault characteristics of different equipment, improving the accuracy and timeliness of fault warnings. For each processing equipment in the service set, if the interlock rejection rate of that equipment is greater than or equal to a preset interlock rejection rate threshold, it is determined that the equipment exhibits frequent interlock rejection. The default value of the preset interlock rejection rate threshold is 0.3. This default value is determined based on the baseline interlock rejection rate under normal operating conditions and the anomaly identification sensitivity requirements. When the interlock rejection rate is low under normal operating conditions, this threshold should be lowered to increase the anomaly detection rate; when interlock rejection is more sporadic, this threshold should be increased to reduce the false alarm rate. For processing equipment exhibiting frequent interlock rejections, the interlock rejection weight coefficient in the anomaly probability calculation is increased by a preset weight adjustment step size. The default value of this preset weight adjustment step size is 0.05, determined based on the balance between sensitivity and stability requirements of weight adjustment. This step size can be increased when rapid response to abnormal mode changes is needed, and decreased when maintaining weight adjustment stability is required. After weight coefficient adjustment, normalization processing is required. The sum of the weight coefficients of all multi-source signals is readjusted to one. The normalization method is to divide each weight coefficient by the sum of all weight coefficients.

[0103] For each processing machine in the service set, if the average waiting time deviation of that processing machine is greater than or equal to a preset waiting deviation threshold, it is determined that the processing machine has an excessively long waiting time. The default value of the preset waiting deviation threshold is 5 seconds. This default value is determined based on the processing cycle time estimation accuracy and the sensitivity requirements for waiting anomaly identification. When the processing cycle time estimation accuracy is high, this threshold should be lowered to improve the anomaly detection rate; when the processing cycle time fluctuates significantly, this threshold should be raised to reduce the false alarm rate. For processing machines exhibiting excessively long waiting times, the processing cycle time mutation weight coefficient in the anomaly probability calculation is increased by a preset weight adjustment step size and normalized. Through the above adaptive weight adjustment mechanism, the anomaly probability calculation can dynamically adjust the importance of various anomaly signals according to the actual operating characteristics of each processing machine, achieving adaptive enhancement of anomaly identification capabilities.

[0104] A service quality assessment mechanism is established to quantitatively evaluate the overall service performance of the overhead rail robot. The reason for establishing this mechanism is that the system's operational performance is influenced by multiple factors, including the weighting of the service urgency index, the parameter configuration of the ant colony algorithm, and the threshold setting for anomaly identification. While the initial default values ​​of these parameters are determined based on experience, the actual operational characteristics of different production systems vary. By establishing a service quality assessment mechanism to quantify the system's actual operational performance and adaptively adjusting key parameters based on the assessment results, the system can be gradually optimized to the best configuration suitable for the current production environment, achieving an evolution from experience-based parameters to data-driven parameters. Service quality indicators are calculated using a preset assessment period as the time window. These indicators include key performance metrics such as average response time, buffer overflow counts, material shortage counts, and equipment idle time percentage. The default value for the preset assessment period is 3600 seconds. This default value is determined based on the statistical significance requirements and real-time monitoring needs of the service quality assessment. The period can be extended when longer-term performance trend analysis is required, and shortened when real-time monitoring of service quality is needed. These four indicators were chosen as the basis for service quality assessment because they reflect different dimensions of system operation. Average response time reflects the timeliness of robot services; buffer overflows reflect the effectiveness of outbound material management; material shortages reflect the continuity of inbound material supply; and equipment idle time percentage reflects the utilization rate of processing equipment. These four indicators together constitute a comprehensive picture of system operation quality, enabling a complete evaluation of the strengths and weaknesses of control strategies. Average response time is the arithmetic mean of the time intervals between the issuance of a service request by all processing equipment and the arrival of the overhead robot within the given time window. Buffer overflows are the number of times the outbound buffer occupancy of any processing equipment reaches its capacity within the given time window. Material shortages are the number of times the inbound buffer occupancy of any processing equipment drops to zero within the given time window. Equipment idle time percentage is the sum of the idle time of all processing equipment within the given time window divided by the product of the time window length and the number of processing equipment.

[0105] The weighting coefficients set in step 100 are adaptively adjusted based on service quality indicators. The reason for this adaptive adjustment is that the three weighting coefficients in the service urgency index directly affect the robot's service priority judgment. Different production systems have different sensitivities to the three risks of outbound backlog, inbound material shortage, and processing wait. The initial default weighting coefficients may not fully adapt to the actual needs of the current system. By monitoring the number of buffer overflows, material shortages, and equipment idle time percentages in the service quality indicators, we can identify which dimension of the system has performance shortcomings and increase the weighting coefficients of the corresponding dimensions accordingly. This makes the calculation of service priority more focused on solving the main contradictions of the current system, thereby achieving adaptive optimization of the control strategy. If the average number of cache overflows is greater than or equal to the preset overflow threshold within several consecutive evaluation periods of the preset adjustment judgment window, the system is deemed to have a high risk of outbound backlog. The preset first weight coefficient is increased by the preset weight adjustment step size. The default value of the preset adjustment judgment window is 3 evaluation periods. This default value is determined based on the stability requirements of weight adjustment. When it is necessary to avoid triggering weight adjustment due to occasional events, the number of windows should be increased. When it is necessary to respond quickly to changes in system performance, the number of windows can be decreased. The default value of the preset overflow threshold is 2 times. This default value is determined based on the system's tolerance for cache overflow risk. When cache overflow will lead to serious consequences, this threshold should be reduced to trigger weight adjustment earlier. If, within several consecutive evaluation periods of the preset adjustment judgment window, the average number of material outages is greater than or equal to the preset material outage threshold, the system is deemed to have a high risk of material outages at the inlet. The preset second weighting coefficient is then increased by the preset weighting adjustment step size. The default value for the preset material outage threshold is 2 times, determined based on the system's tolerance for material outage risk. When material outages would lead to severe capacity loss, this threshold should be lowered to trigger weighting adjustments earlier. Similarly, if, within several consecutive evaluation periods of the preset adjustment judgment window, the average percentage of equipment idle time is greater than or equal to the preset idle percentage threshold, the system is deemed to have a high capacity loss due to processing waiting. The preset third weighting coefficient is then increased by the preset weighting adjustment step size. The default value for the preset idle percentage threshold is 0.15, determined based on the system's requirements for equipment utilization. When equipment utilization requirements are high, this threshold should be lowered to trigger weighting adjustments earlier. After the weight coefficients are adjusted, they need to be normalized. The sum of the preset first weight coefficient, the preset second weight coefficient, and the preset third weight coefficient is readjusted to the sum of the preset weight coefficients. The default value of the sum of the preset weight coefficients is 3.0, which is the sum of the initial default values ​​of the three weight coefficients.

[0106] A parameter self-calibration log recording mechanism is established to record all parameter adjustment operations. The record includes the adjustment time, parameter name, value before adjustment, value after adjustment, and reason for triggering the adjustment. The parameter self-calibration log is stored in the system configuration database for system maintenance personnel to query and analyze, enabling manual intervention or parameter rollback when necessary. Upper and lower limit constraints for parameter adjustment are set. For all adaptively adjustable weight coefficients, preset lower and upper weight limits are set. The default value for the preset lower weight limit is 0.05, and the default value for the preset upper weight limit is 5.0. These default values ​​are determined based on the reasonable range of weight coefficient values ​​to avoid adaptive adjustment leading to weight coefficients that are too small (ineffective) or too large (leading to system imbalance). After each weight coefficient adjustment, if the adjusted weight coefficient is less than the preset lower weight limit, it is set to the preset lower weight limit; if the adjusted weight coefficient is greater than the preset upper weight limit, it is set to the preset upper weight limit.

[0107] Return to step 200 and continuously execute the above steps to achieve dynamic control of the loading and unloading sequence of the overhead rail robot and multiple processing equipment.

[0108] Example 2

[0109] Referring to Figure 4, a timing control system for the pick-and-place of a ceiling-mounted robot and multiple processing devices is provided. This system stores computer-readable instructions, which, when read, execute the aforementioned timing control method for the pick-and-place of the ceiling-mounted robot and multiple processing devices. The system includes:

[0110] Priority determination module 101 reads the set of processing equipment and the cache capacity parameters of each processing equipment, establishes hard constraints based on the cache capacity parameters and calculates the service urgency index to quantify the service priority of the processing equipment.

[0111] The status perception module 102 collects the operating status information of each processing device at a preset sampling period, calculates the abnormal probability of each processing device based on multi-source signals, and divides the processing devices into a fault isolation set and a serviceable set according to the abnormal probability and operating status information. When any processing device is transferred from the serviceable set to the fault isolation set, a dynamic switching operation is performed.

[0112] The path optimization module 103, based on the service urgency index, uses the ant colony algorithm to globally optimize the service path of the overhead rail robot, maps each processing device in the service set to a node in the ant colony algorithm, maps the robot's movement path between processing devices to an edge between nodes, and generates and executes an action sequence through iterative search and pheromone update.

[0113] The fault control module 104 performs isolation operations on the processing equipment within the fault isolation set, monitors the downstream accumulation risk caused by the fault, activates the virtual back pressure control mechanism, and smoothly integrates the recovered processing equipment.

[0114] The history update module 105 records the entire process of each pick-up and place operation performed by the overhead rail robot and updates the estimated processing completion time in the running status information.

[0115] The embodiments of the present invention have been described above. However, the embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments under the guidance of the present embodiments, and all of them are within the protection scope of the present embodiments.

Claims

1. A method for controlling the loading and unloading sequence of a ceiling-mounted robot and multiple processing devices, characterized in that, Includes the following steps: The system reads the set of processing equipment and the cache capacity parameters of each equipment. Based on these cache capacity parameters, it establishes hard constraints and calculates a service urgency index to quantify the service priority of the processing equipment. It collects the operating status information of each processing equipment at a preset sampling period, calculates the anomaly probability of each equipment based on multi-source signals, and divides the processing equipment into a fault isolation set and a serviceable set according to the anomaly probability and operating status information. When any processing equipment moves from the serviceable set to the fault isolation set, a dynamic switching operation is performed. Based on the service urgency index, it uses an ant colony algorithm to globally optimize the service path of the overhead rail robot, mapping each processing equipment in the serviceable set to a node in the ant colony algorithm, and mapping the robot's movement path between processing equipment to edges between nodes. An action sequence is generated and executed through iterative search and pheromone updates. It performs isolation operations on the processing equipment in the fault isolation set, monitors the downstream accumulation risk caused by the fault, initiates a virtual backpressure control mechanism, and smoothly incorporates recovered processing equipment. It records the entire process of each pick-and-place operation performed by the overhead rail robot and updates the estimated processing completion time in the operating status information.

2. The method for controlling the loading and unloading sequence of the overhead track robot and multiple processing devices according to claim 1, characterized in that, The cache capacity parameters include inbound cache capacity and outbound cache capacity; hard constraints include cache capacity constraints and interlock safety constraints. The cache capacity constraint requires that for any given time and any processing equipment, the inbound cache occupancy of the processing equipment must be greater than or equal to zero and not exceed the inbound cache capacity of the processing equipment, and the outbound cache occupancy of the processing equipment must be greater than or equal to zero and not exceed the outbound cache capacity of the processing equipment. The interlock safety constraint indicates whether the processing equipment allows the robot to perform pick-and-place operations through an interlock permission signal. The service urgency index is calculated by multiplying a preset first weighting coefficient by the outbound cache urgency, adding a preset second weighting coefficient by the inbound cache urgency, and adding a preset third weighting coefficient by the processing completion time urgency. The outbound cache urgency indicates the degree to which the outbound cache exceeds the warning threshold, the inbound cache urgency indicates the degree to which the inbound cache is below the warning threshold, and the processing completion time urgency is the preset processing completion urgency baseline duration divided by the remainder, where the remainder is the remaining processing time plus the preset processing completion urgency baseline duration.

3. The method for controlling the loading and unloading sequence of the overhead track robot and multiple processing devices according to claim 1, characterized in that, The dynamic switching operation includes: running status information including processing equipment status, inbound buffer occupancy, outbound buffer occupancy, and estimated processing completion time; the anomaly probability calculation method involves normalizing each multi-source signal, multiplying each normalized multi-source signal by its corresponding preset weight coefficient, and then summing the results; if the anomaly probability of a processing equipment is greater than or equal to a preset anomaly probability threshold, or if the processing equipment status is faulty, then the processing equipment is added to the fault isolation set; otherwise, the processing equipment is added to the serviceable set; maintaining the execution sequence and shadow sequence; determining the number of completed actions in the execution sequence and obtaining the current status information; finding the longest consecutively executable prefix subsequence in the shadow sequence that meets the conditions, including performing action feasibility verification and status prediction sequentially starting from the first action in the shadow sequence and removing actions involving newly faulty processing equipment; if the length of the prefix subsequence is greater than zero, replacing the unexecuted actions in the execution sequence with the prefix subsequence; if the length of the prefix subsequence is zero, immediately re-planning and generating a new execution sequence.

4. The method for controlling the loading and unloading sequence of the overhead track robot and multiple processing devices according to claim 3, characterized in that, Multi-source signals include processing equipment fault codes, temperature anomaly signals, vibration anomaly signals, processing cycle change signals, and interlock rejection counts; the execution sequence is the action sequence to be issued to the overhead rail robot for execution, and its length is the number of steps specified by the preset rolling planning window length, including material picking actions, material placement actions, movement actions or waiting actions and the corresponding target processing equipment; the shadow sequence is the next round of optimal action sequence pre-calculated under the premise that the current fault isolation set remains unchanged, and the shadow sequence is generated immediately after the execution sequence starts execution and the background calculation is started; the status information includes the actual status of the overhead rail robot, the robot carrier occupancy, and the inbound and outbound buffer occupancy of each processing equipment; the action feasibility verification includes at least hard constraints.

5. The method for controlling the loading and unloading sequence of the overhead track robot and multiple processing devices according to claim 1, characterized in that, The action sequence generated and executed through iterative search and pheromone updates includes: initializing the ant colony algorithm by setting the preset number of ants, the preset maximum number of iterations, the preset pheromone importance coefficient, the preset heuristic pheromone importance coefficient, the preset pheromone evaporation coefficient, and the preset pheromone increment constant; constructing the problem model of the ant colony algorithm by mapping each processing device in the service set to a node in the ant colony algorithm, mapping the robot's movement path between processing devices to edges between nodes, initializing the pheromone concentration on each edge to a preset initial pheromone concentration value, and calculating the heuristic pheromone value of each edge; introducing action feasibility constraints and state prediction mechanisms into the problem model by setting the robot's carrying capacity and robot carrying occupancy, and setting material handling feasibility conditions, material placement feasibility conditions, and combined material handling and placement feasibility conditions; and executing the iterative search process of the ant colony algorithm, in which each ant starts from the robot's current position and sequentially selects the next... The algorithm continues until a complete service path is constructed, including the number of steps specified by the preset rolling planning window length. When selecting the next target processing device node, the ants use a probabilistic selection mechanism. After each ant completes a complete path, the overall fitness value of the path is calculated, and the path with the highest fitness value in the current iteration is recorded as the optimal path for this round. After completing one round of iteration, a pheromone update operation is performed, including pheromone evaporation and pheromone release. The iterative search and pheromone update process is repeated until the preset maximum number of iterations is reached or the convergence condition is met. After the algorithm terminates, the globally optimal path is output as the optimal service sequence of the overhead rail robot. Based on the optimal service sequence, the target processing device for each step is determined in sequence, and at each step, the feasibility of the action and the action type are checked based on the expected arrival time. The action type includes material picking, material releasing, or a combination of picking and releasing. The generated actions are added to the action sequence.

6. The method for controlling the loading and unloading sequence of the overhead track robot and multiple processing devices according to claim 5, characterized in that, The heuristic information value is the service urgency item divided by the time cost item. The time cost item is the sum of the movement cost and the waiting cost plus a preset time cost benchmark value. The service urgency item is the service urgency index plus a preset urgency benchmark value. Here, the movement cost represents the time required for the robot to move from its current position to the processing equipment, and the waiting cost represents the time the robot needs to wait for processing to complete after arriving at the processing equipment. The waiting cost is calculated by subtracting the sum of the current time and the movement cost from the estimated processing completion time of the processing equipment. If the difference is positive, it is taken as the waiting cost; if the difference is negative or zero, the waiting cost is zero. The feasibility conditions for material retrieval are that the outbound buffer occupancy of the target processing equipment is greater than zero, the interlock permission signal is in the allowed state, and the robot carrier occupancy is less than the robot's occupancy. The feasibility conditions for loading and unloading are that the inbound buffer occupancy of the target processing equipment is less than the inbound buffer capacity, the interlock permission signal is in the allowed state, and the robot vehicle occupancy is greater than zero. The feasibility conditions for composite loading and unloading are that the same target processing equipment simultaneously meets the loading and unloading feasibility conditions at the same arrival time, and the execution order of composite loading and unloading actions is loading first and then unloading. State prediction is used to perform forward calculation of the state at the arrival time when constructing candidate solutions. The calculation method of state prediction is to accumulate the movement cost and waiting cost of each step based on the current time of the overhead rail robot and the robot vehicle occupancy according to the candidate action sequence to obtain the expected arrival time of each step, and to perform action feasibility verification at each step arrival time by combining the interlock permission signal, the inbound buffer occupancy, the outbound buffer occupancy, and the estimated processing completion time.

7. The method for controlling the loading and unloading sequence of a ceiling-mounted robot and multiple processing devices according to claim 6, characterized in that, The probabilistic selection mechanism involves first calculating, based on state prediction, whether each candidate processing device has at least one executable action at its expected arrival time for the current ant, and eliminating candidate processing devices that do not meet the action feasibility constraints. The executable actions include material picking, material releasing, or a combination of picking and releasing. Then, the probability of selecting the remaining candidate processing devices is calculated. Specifically, this includes taking the pheromone concentration on the edge from the current node to the candidate processing device, increasing it to a power of a preset pheromone importance coefficient to obtain an enhanced pheromone concentration, increasing the heuristic information value of the edge to a power of a preset heuristic information importance coefficient to obtain an enhanced heuristic information value, and multiplying the enhanced pheromone concentration by the enhanced heuristic information value as the candidate processing device selection probability. The selection tendency value of a candidate processing device is calculated by dividing the selection tendency value of the candidate processing device by the sum of the selection tendency values ​​of all candidate processing devices. The quotient is the probability that the candidate processing device is selected. Ants select the next target processing device node using a roulette wheel method based on the above probability distribution. The fitness value of a path is calculated by summing the service urgency indices of the target processing devices at each step of the path to obtain the total service urgency, and summing the movement cost and waiting cost at each step of the path to obtain the total time cost. The fitness value of a path is the total service urgency divided by the total time cost, which is the sum of the total time cost and the preset fitness time benchmark value. The convergence condition is that the globally optimal path remains unchanged in multiple consecutive iterations, and the number of consecutive iterations is the preset convergence judgment round.

8. The method for controlling the loading and unloading sequence of a ceiling-mounted robot and multiple processing devices according to claim 1, characterized in that, Activating the virtual backpressure control mechanism and smoothly incorporating recovered processing equipment includes: performing isolation operations on processing equipment within the fault isolation set, prohibiting the overhead rail robot from entering the pick-up and drop-off interlock area of ​​the faulty processing equipment; monitoring the downstream accumulation risk caused by the fault, and activating the virtual backpressure control mechanism when the ratio of the outbound buffer occupancy of any processing equipment to the outbound buffer capacity of the processing equipment is greater than or equal to a preset backpressure trigger threshold. The virtual backpressure control mechanism includes calculating the backpressure coefficient; the backpressure coefficient achieves flow control by adjusting the upstream feeding cycle time, and when the backpressure coefficient is greater than or equal to a preset backpressure stop threshold, the standard feeding interval time of the upstream feeding processing equipment is divided by the backpressure. The adjusted feeding interval time is used to suspend upstream feeding until the back pressure coefficient recovers when the back pressure coefficient is less than the preset back pressure stop threshold. When a processing device in the fault isolation set reports a recovery signal or the abnormal probability is less than the preset abnormal probability threshold, the processing device is transferred from the fault isolation set to the observation recovery set. A stability window determination is performed on the processing devices in the observation recovery set. If a processing device simultaneously meets the conditions of not being in a fault state and having an interlock allowance rate greater than or equal to the preset allowance rate threshold within a continuous preset stability observation window duration, the processing device is allowed to be included in the serviceable set. A gradual quota mechanism is used to smoothly include the recovered processing devices that pass the stability window determination.

9. The method for controlling the loading and unloading sequence of a ceiling-mounted robot and multiple processing devices according to claim 1, characterized in that, The back pressure coefficient is calculated as follows: for each processing device in the serviceable set, subtract the cache occupancy from 1. The cache occupancy is the ratio of the outbound cache occupancy to the outbound cache capacity of the processing device. If the result is negative, take zero. Then, take the minimum value among the above calculation results of all processing devices in the serviceable set. The minimum value is the back pressure coefficient. The interlock allowance rate is the ratio of the number of successful pick-up and release requests to the total number of requests within the preset stable observation window. The progressive quota mechanism allocates service weights to the recovery processing equipment. The service weights are calculated using a ramp function. The calculation method is to take the current time and subtract the time when the processing equipment passes the stability window determination, and then divide the difference by a preset ramp duration parameter. If the ratio is greater than one, it is taken as one; otherwise, the ratio is taken as the service weight. By gradually increasing the service weight from zero to one, the recovery processing equipment gradually transitions from participating in a small number of scheduling tasks to normal competitive scheduling. When the service weight equals one, the recovery processing equipment enters a fully normal scheduling state.

10. A timing control system for the pick-and-place of a ceiling-mounted robot and multiple processing devices, characterized in that, It is used to implement the timing control method for picking up and placing multiple processing devices of the overhead track robot as described in any one of claims 1-9; the system includes: a priority determination module, which reads the set of processing devices and the cache capacity parameters of each processing device, establishes hard constraints based on the cache capacity parameters and calculates a service urgency index to quantify the service priority of the processing devices; a status perception module, which collects the operating status information of each processing device at a preset sampling period, calculates the abnormal probability of each processing device based on multi-source signals, divides the processing devices into a fault isolation set and a serviceable set according to the abnormal probability and operating status information, and performs a dynamic switching operation when any processing device is transferred from the serviceable set to the fault isolation set; The path optimization module, based on the service urgency index, uses an ant colony algorithm to globally optimize the service path of the overhead rail robot. It maps each processing device in the serviceable set to a node in the ant colony algorithm, and maps the robot's movement path between processing devices to edges between nodes. It generates and executes action sequences through iterative search and pheromone updates. The fault control module performs isolation operations on processing devices within the fault isolation set, monitors the downstream accumulation risk caused by faults, activates a virtual backpressure control mechanism, and smoothly incorporates recovered processing devices. The history update module records each pick-and-place operation performed by the overhead rail robot throughout the entire process and updates the estimated processing completion time in the operating status information.

Images