Distributed resource negotiation and space-time conflict resolution method based on market mechanism

Abstract

This invention provides a market-based method for distributed resource negotiation and spatiotemporal conflict resolution. The method includes: a management node constructs a spatiotemporal demand Bloom filter based on task spatiotemporal constraints and broadcasts task announcements; simultaneously, it maintains a reputation model containing historical scheduling deviations and conflict contributions of nodes; contractor nodes match the filter and submit bids containing resource scheduling schemes and associated confidence levels; the management node calibrates the scheduling schemes based on reputation information and confidence levels to obtain a resource scheduling probability distribution, and calculates the global load entropy reduction accordingly; when conflicts exist, a loss function is constructed based on task migration, resource idleness, and opportunity cost to calculate the preemption compensation cost; an evaluation value is obtained by comprehensively considering the bidding benefits, entropy reduction, and compensation cost; the best contract is awarded; after the contract is executed, the contractor node's reputation record is updated based on the actual execution and conflict results, thereby achieving distributed resource negotiation and conflict optimization resolution.

Description

Technical Field

[0001] This application belongs to the field of conflict resolution, and in particular relates to a method for distributed resource negotiation and spatiotemporal conflict resolution based on market mechanisms. Background Technology

[0002] In modern large-scale distributed systems, such as cluster collaboration, network resource scheduling, autonomous vehicle fleet management, and cloud computing data center task allocation, these systems typically consist of a large number of autonomous or semi-autonomous agents that need to compete for limited spatiotemporal resources in a changing environment. Tasks are highly spatiotemporally coupled; the execution of one task not only consumes resources but also occupies specific time and space dimensions, thus constraining the planning of other potential tasks. Therefore, how to plan a resource negotiation mechanism within a decentralized architecture that can respond quickly, make fair decisions, and anticipate and resolve potential spatiotemporal conflicts is a key technical challenge. The Contract Network Protocol (CNP) is widely researched and applied due to its excellent scalability and flexibility. CNP realizes the allocation of tasks among distributed nodes by simulating the task posting, bidding, and awarding process in a market.

[0003] The contract network protocol relies on explicit bid prices and lacks in-depth consideration of contractor nodes' historical performance capabilities and the inherent uncertainties of bidding proposals. The decision-making process is often locally optimal, selecting only the best contractor for the currently assigned task. When high-priority tasks arise and require preemption of already occupied resources, existing mechanisms typically employ simple preemption strategies. They lack a sophisticated conflict resolution model to represent the comprehensive costs of preemption, including the migration costs of the preempted task, the idle costs resulting from resource reallocation, and the costs of forgoing potential opportunities. Furthermore, they fail to incorporate the historical cooperation or conflict contribution of the preempted party into compensation decisions, leading to unfairness and inefficiency in conflict resolution. Therefore, an improved distributed negotiation method is urgently needed, integrating node reputation, global system stability, and sophisticated conflict cost assessment to achieve more intelligent spatiotemporal resource scheduling. Summary of the Invention

[0004] This invention proposes a market-based distributed resource negotiation and spatiotemporal conflict resolution method to address the problem that existing methods cannot provide a distributed negotiation mechanism that integrates node reputation, global system stability, and refined conflict cost assessment. The method includes: The management node generates a spatiotemporal demand Bloom filter based on the task's spatiotemporal constraints and broadcasts the task announcement. At the same time, it maintains a reputation model that records the historical scheduling deviations and conflict contributions of contractor nodes. After matching the spatiotemporal demand Bloom filter, the contractor node submits a bid that includes the expected resource scheduling scheme and the associated confidence level. The management node evaluates the received bids and, combining the historical scheduling deviations in the reputation model with the correlation confidence of the bids, calibrates the expected resource scheduling scheme to generate a calibrated resource scheduling probability distribution; based on the probability distribution, it calculates the entropy reduction of the global spatiotemporal load distribution caused by granting the bid; and when the bid conflicts with a planned contract, it calculates the contract preemption compensation cost through a loss function that combines task migration cost, resource idle cost, and opportunity cost. The management node calculates a comprehensive evaluation value based on the bidding execution benefits, the entropy reduction, and the contract preemption compensation cost, and awards the contract to the contractor node with the best comprehensive evaluation value; and after the contract is executed, updates the record of the winning contractor node in the reputation model based on the actual execution data and the resulting spatiotemporal conflicts.

[0005] Furthermore, the step of combining the historical scheduling deviations in the reputation model with the association confidence of the bids to calibrate the expected resource scheduling scheme and generate a calibrated resource scheduling probability distribution includes: The mean and standard deviation of the time deviation of contractor nodes in the start time and duration of the scheduling scheme, as well as the mean and standard deviation of the resource usage deviation in the allocation of various core resources, are extracted from the reputation model. The time standard deviation and the capacity standard deviation for each resource are adjusted according to the correlation confidence level declared in the tender, where a higher confidence level corresponds to a smaller adjusted standard deviation; The statistical characteristics of the deviation are respectively used as the extracted mean time deviation and the adjusted time standard deviation to define the time Gaussian noise model, and the extracted mean occupancy deviation and the adjusted standard deviation of each resource dimension to define the capacity Gaussian noise model, which are applied to the time attribute and resource capacity attribute of each spatiotemporal anchor point of the expected resource scheduling scheme in the bidding. The scheduling scheme with added noise was simulated multiple times using the Monte Carlo sampling method. The frequency of resource occupancy in each discrete spatiotemporal cell was counted, and the frequency was normalized to form a calibrated resource scheduling probability distribution.

[0006] Further, the calculation of the entropy reduction in the global spatiotemporal occupancy load distribution resulting from the granting of a bid based on the probability distribution includes: Discretize the four-dimensional spatiotemporal domain of task execution into a spatiotemporal volumetric raster; For each spatiotemporal volume element i, the expected load of the volume element is obtained by summing the probabilities of the volume element being occupied by all planned tasks. The expected occupancy load of each element is normalized into an occupancy probability. ; Calculate the global load entropy before awarding a new bid. The summation iterates through all Volume elements >0; The calibrated resource scheduling probability distribution of the new bids is added to the global occupancy model, and the expected occupancy load of each element due to all planned tasks and the current bids to be evaluated is recalculated. And normalized to occupancy probability Calculate the global load entropy after the new bid is awarded. ; Calculate the and The difference yields the entropy reduction of the global spatiotemporal load distribution.

[0007] Furthermore, when a bid conflicts with a planned contract, the contract preemption compensation cost is calculated using a loss function that combines task migration cost, resource idle cost, and opportunity cost, including: Retrieve historical conflict contribution of preempted contracts from a reputation model. And based on the formula Calculate the weights, where k is a preset positive coefficient; Based on the topological distance between the current network node seizing resources and the backup task network node, a preset unit topological distance migration cost table is consulted to obtain the unit topological distance migration cost. The task migration cost is then obtained by multiplying the unit topological distance migration cost by the topological distance. ; Predict the idle time of the resource from being preempted to being reallocated to a standby task. Multiplying this by the unit time holding cost of the resource yields the resource idle cost. ; The opportunity cost is determined by assessing the direct profit loss resulting from the cancellation of the preemptive contract. ; Contract snatching compensation cost .

[0008] Furthermore, the calculation of the comprehensive evaluation value based on the bidding execution benefits, the entropy reduction, and the contract preemption compensation cost includes: Extracting the declared implementation benefits from the bid ; Obtain the calculated entropy reduction ΔH of the global spatiotemporal occupancy load distribution; Obtain the calculated contract preemption compensation cost If there is no seizure, then =0; Through formula Calculate the comprehensive evaluation value V, where α, β, and γ are preset weighting coefficients used to unify the dimensions and balance the importance of each component.

[0009] Furthermore, the management node generates a spatiotemporal requirement Bloom filter based on task spatiotemporal constraints, including: Extract the task's time window and the set of network topology logical boundaries; The time window is divided into a series of discrete time slice identifiers according to a preset time unit; The logical boundary is divided according to the preset server cluster partitioning rules to obtain a series of discrete spatial cell identifiers; For each time-slice-space cell combination required by the task, the identifier of the combination is used as input, multiple hash values ​​are calculated through multiple different hash functions, and the bit position corresponding to the index of the hash value in the Bloom filter is set to 1.

[0010] Furthermore, updating the record of the winning contractor node in the reputation model based on actual execution data and the resulting spatiotemporal conflicts includes: After the contract is completed, collect the actual resource allocation sequence and time log of the contractor's resources; The actual resource allocation sequence and time log are compared with the expected resource scheduling plan submitted at the time of winning the bid. The actual deviations in start time, duration, and resource capacity are calculated. The historical deviation mean and standard deviation records of the contractor in the reputation model are updated using the exponential weighted moving average method. Analyze the spatiotemporal conflict events caused by the contractor and confirmed during the actual execution process, assess the disturbances caused by the conflict events to the system, and calculate the realized conflict contribution of this task. The realized conflict contribution is incorporated into the contractor's historical conflict contribution record in the reputation model using an exponentially weighted moving average method.

[0011] Furthermore, this invention also relates to a market-based distributed resource negotiation and spatiotemporal conflict resolution system, comprising the following modules: The generation module is used by management nodes to generate a spatiotemporal demand Bloom filter based on task spatiotemporal constraints and broadcast task announcements. At the same time, it maintains a reputation model that records the historical scheduling deviations and conflict contributions of contractor nodes. After matching the spatiotemporal demand Bloom filter, contractor nodes submit bids that include expected resource scheduling schemes and associated confidence levels. The calculation module is used to manage the decision-making evaluation of received bids by the nodes, combine the historical scheduling deviations in the reputation model with the correlation confidence of the bids, calibrate the expected resource scheduling scheme to generate a calibrated resource scheduling probability distribution; calculate the entropy reduction of the global spatiotemporal load distribution caused by granting the bid based on the probability distribution; and calculate the contract preemption compensation cost when the bid conflicts with the planned contract through a loss function that combines task migration cost, resource idle cost and opportunity cost. The update module is used to manage nodes to calculate a comprehensive evaluation value based on the bidding execution benefits, the entropy reduction, and the contract preemption compensation cost, and to award the contract to the contractor node with the best comprehensive evaluation value; and after the contract is executed, to update the record of the winning contractor node in the reputation model based on the actual execution data and the resulting spatiotemporal conflicts.

[0012] Preferably, the step of combining the historical scheduling deviations in the reputation model with the correlation confidence of the bids to calibrate the expected resource scheduling scheme and generate a calibrated resource scheduling probability distribution includes: The mean and standard deviation of the time deviation of contractor nodes in the start time and duration of the scheduling scheme, as well as the mean and standard deviation of the resource usage deviation in the allocation of various core resources, are extracted from the reputation model. The time standard deviation and the capacity standard deviation for each resource are adjusted according to the correlation confidence level declared in the tender, where a higher confidence level corresponds to a smaller adjusted standard deviation; The statistical characteristics of the deviation are respectively used as the extracted mean time deviation and the adjusted time standard deviation to define the time Gaussian noise model, and the extracted mean occupancy deviation and the adjusted standard deviation of each resource dimension to define the capacity Gaussian noise model, which are applied to the time attribute and resource capacity attribute of each spatiotemporal anchor point of the expected resource scheduling scheme in the bidding. The scheduling scheme with added noise was simulated multiple times using the Monte Carlo sampling method. The frequency of resource occupancy in each discrete spatiotemporal cell was counted, and the frequency was normalized to form a calibrated resource scheduling probability distribution.

[0013] Preferably, the step of calculating the entropy reduction in the global spatiotemporal occupancy load distribution resulting from the granting of a bid based on the probability distribution includes: Discretize the four-dimensional spatiotemporal domain of task execution into a spatiotemporal volumetric raster; For each spatiotemporal volume element i, the expected load of the volume element is obtained by summing the probabilities of the volume element being occupied by all planned tasks. The expected occupancy load of each element is normalized into an occupancy probability. ; Calculate the global load entropy before awarding a new bid. The summation iterates through all Volume elements >0; The calibrated resource scheduling probability distribution of the new bids is added to the global occupancy model, and the expected occupancy load of each element due to all planned tasks and the current bids to be evaluated is recalculated. And normalized to occupancy probability Calculate the global load entropy after the new bid is awarded. ; Calculate the and The difference yields the entropy reduction of the global spatiotemporal load distribution.

[0014] Preferably, when a bid conflicts with a planned contract again, the contract preemption compensation cost is calculated using a loss function that combines task migration cost, resource idle cost, and opportunity cost, including: Retrieve historical conflict contribution of preempted contracts from a reputation model. And based on the formula Calculate the weights, where k is a preset positive coefficient; Based on the topological distance between the current network node seizing resources and the backup task network node, a preset unit topological distance migration cost table is consulted to obtain the unit topological distance migration cost. The task migration cost is then obtained by multiplying the unit topological distance migration cost by the topological distance. ; Predict the idle time of the resource from being preempted to being reallocated to a standby task. Multiplying this by the unit time holding cost of the resource yields the resource idle cost. ; The opportunity cost is determined by assessing the direct profit loss resulting from the cancellation of the preemptive contract. ; Contract snatching compensation cost .

[0015] Preferably, the step of calculating the comprehensive evaluation value based on the bidding execution benefits, the entropy reduction, and the contract preemption compensation cost includes: Extracting the declared implementation benefits from the bid ; Obtain the calculated entropy reduction ΔH of the global spatiotemporal occupancy load distribution; Obtain the calculated contract preemption compensation cost If there is no seizure, then =0; Through formula Calculate the comprehensive evaluation value V, where α, β, and γ are preset weighting coefficients used to unify the dimensions and balance the importance of each component.

[0016] Preferably, the management node generates a spatiotemporal requirement Bloom filter based on task spatiotemporal constraints, including: Extract the task's time window and the set of network topology logical boundaries; The time window is divided into a series of discrete time slice identifiers according to a preset time unit; The logical boundary is divided according to the preset server cluster partitioning rules to obtain a series of discrete spatial cell identifiers; For each time-slice-space cell combination required by the task, the identifier of the combination is used as input, multiple hash values ​​are calculated through multiple different hash functions, and the bit position corresponding to the index of the hash value in the Bloom filter is set to 1.

[0017] Preferably, updating the record of the winning contractor node in the reputation model based on actual execution data and the resulting spatiotemporal conflicts includes: After the contract is completed, collect the actual resource allocation sequence and time log of the contractor's resources; The actual resource allocation sequence and time log are compared with the expected resource scheduling plan submitted at the time of winning the bid. The actual deviations in start time, duration, and resource capacity are calculated. The historical deviation mean and standard deviation records of the contractor in the reputation model are updated using the exponential weighted moving average method. Analyze the spatiotemporal conflict events caused by the contractor and confirmed during the actual execution process, assess the disturbances caused by the conflict events to the system, and calculate the realized conflict contribution of this task. The realized conflict contribution is incorporated into the contractor's historical conflict contribution record in the reputation model using an exponentially weighted moving average method.

[0018] This invention reduces communication overhead and node matching computation burden during the task release phase by using a spatiotemporal demand Bloom filter. By constructing a reputation model incorporating historical scheduling deviations and conflict contributions, and combining it with a bid confidence calibration scheme, it achieves an accurate representation of contractor node reliability and prediction of future resource occupancy. Utilizing global spatiotemporal load entropy reduction as a decision factor helps optimize the overall resource load balance of the system, fundamentally reducing potential conflict risks. Simultaneously, a contract preemption compensation cost function based on multidimensional costs ensures the fairness and economic rationality of resource scheduling. A comprehensive evaluation combining execution efficiency, entropy reduction, and preemption costs makes resource negotiation decisions more comprehensive and optimal, while post-execution reputation updates form a closed-loop feedback, continuously improving the collaborative scheduling efficiency and stability of the entire distributed system. Attached Figure Description

[0019] Figure 1 A flowchart of a market-based distributed resource negotiation and spatiotemporal conflict resolution method; Figure 2 A line graph illustrating the comparison of task success rates; Figure 3 A line graph showing the comparison of the number of spatiotemporal conflicts; Figure 4 This is a diagram showing the comparison of average idle time of resources. Detailed Implementation

[0020] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this specification. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this specification as detailed in the appended claims.

[0021] It should be understood that the terms “comprising” and “having”, and any variations thereof, in the embodiments of this specification are intended to cover but not exclude inclusion. For example, a product or device that includes a series of components is not necessarily limited to those components that are explicitly listed, but may include other components that are not explicitly listed or that are inherent to such product or device.

[0022] Firstly, this invention proposes a market-based method for distributed resource negotiation and spatiotemporal conflict resolution, see reference [link to relevant documentation]. Figure 1 As shown, it includes the following steps: S1, the management node generates a spatiotemporal requirement Bloom filter based on the task's spatiotemporal constraints and broadcasts the task announcement. At the same time, it maintains a reputation model that records the historical scheduling deviations and conflict contributions of contractor nodes.

[0023] The management node discretizes the continuous spatiotemporal range required for the task into a set of spatiotemporal cells. Spatially, it uses a data center network topology coding algorithm to convert physical or logical locations, such as Region-Zone-Rack-Node, into string codes. Temporally, it divides the space into fixed-length time slices, with each spatiotemporal cell uniquely identified by a network topology code and a time slice index. The management node initializes a bit array as a Bloom filter. For each spatiotemporal cell identifier, it calculates multiple hash values ​​using various hash algorithms such as MurmurHash3 and FNV-1a, and sets the bit value at the corresponding index position in the bit array to 1. The task announcement includes the task's unique identifier, task reward, and the generated Bloom filter bit array. It is broadcast to all contractor nodes via the Advanced Message Queuing Protocol (AMQP) using message brokers such as RabbitMQ in a publish-subscribe pattern. The reputation model is maintained on the management node using a key-value database such as Redis. The key is the contractor node ID, and the value is an object containing the exponential moving average of historical scheduling deviations and the accumulated conflict contribution value.

[0024] In one implementation, the management node generates a spatiotemporal requirement Bloom filter based on task spatiotemporal constraints, including: Extract the task's time window and the set of network topology logical boundaries; The time window is divided into a series of discrete time slice identifiers according to a preset time unit; The logical boundary is divided according to the preset server cluster partitioning rules to obtain a series of discrete spatial cell identifiers; For each time-slice-space cell combination required by the task, the identifier of the combination is used as input, multiple hash values ​​are calculated through multiple different hash functions, and the bit position corresponding to the index of the hash value in the Bloom filter is set to 1.

[0025] Set key parameters for the Bloom filter: bit array size m, e.g., m=65536; number of hash functions k, e.g., k=7. Also set the precision of the spatiotemporal discretization: time slice units. =5 minutes, spatial grid size For a standard compute node (such as a specific server instance), when a new task is received, a time window is extracted, for example [14:00, 16:00], and this time window is divided into 24 five-minute time slices. Each time slice is assigned a unique identifier, such as one based on a Unix timestamp. Simultaneously, the logical boundaries of the network topology are extracted, such as a resource pool defined by a specific subnet or data center availability zone. A resource mapping algorithm is used to determine all standard compute node units belonging to that availability zone, and a unique identifier, such as one based on an IP address or server MAC address, is generated for each unit.

[0026] Generate all possible time slice identifier-spatial cell identifier combinations. For each combination, create a unique string as input to a hash function, for example, "t_1678881600_node_192_168_1_15". Input this string sequentially into seven independent hash functions, such as Murmur3, FNV-1, xxHash, etc. Each hash function calculates a hash value and takes the modulo of m (65536) to obtain an index position in the range [0, 65535]. For example, the seven hash functions might produce indices: {123, 4567, 8910, 23456, 31415, 54321, 65000}. Set all the bits at these seven index positions in the Bloom filter bit array to 1. Iterate through all spatiotemporal combination cells and repeat this process. The resulting bit array is the spatiotemporal requirement Bloom filter for this task, which can be used for initial contractor screening.

[0027] S2, after the contractor node matches the spatiotemporal demand Bloom filter, it submits a bid that includes the expected resource scheduling scheme and the associated confidence level.

[0028] After receiving the task announcement, the contractor node generates a set of feasible spatiotemporal cells based on its available resources. For each spatiotemporal cell in this set, a hash value is calculated using the same multiple hash algorithms as the management node, and the bit value at the corresponding index position in the Bloom filter bit array is checked. If all positions are 1, the task requirements are considered to be initially matched with its own resources. After a successful match, the contractor node uses a heuristic graph allocation algorithm or a virtual machine placement planning algorithm to plan a specific resource allocation evolution sequence as the expected resource scheduling scheme, taking into account its own resource topology and constraints. The association confidence is calculated based on the historical execution stability data of the resources used in the scheduling scheme, and its value is 1 minus the ratio of the expected execution time standard deviation to the expected execution time mean.

[0029] S3, the management node evaluates the received bids and, combining the historical scheduling deviations in the reputation model with the correlation confidence of the bids, calibrates the expected resource scheduling scheme to generate a calibrated resource scheduling probability distribution.

[0030] The management node reads the historical scheduling deviation values ​​from the bidding contractor nodes in the Redis database, using these values ​​as prior information. The management node treats the occurrence time of each spatiotemporal point in the expected resource scheduling scheme in the bidding process as a random variable, with the mean being the expected time given in the scheme. Using a Bayesian update method, combining the historical scheduling deviation as prior information and the bidding association confidence score as new evidence, the posterior variance of this random variable is calculated. This variance is obtained by weighting the exponential moving average of the historical scheduling deviation with 1 minus the association confidence score. A normal or gamma distribution with the expected time as the mean and the posterior variance as the standard deviation is generated for each key spatiotemporal point in the scheduling scheme. The set of these probability distributions constitutes the calibrated resource scheduling probability distribution.

[0031] In one implementation, the step of combining the historical scheduling deviations in the reputation model with the association confidence of the bids to calibrate the expected resource scheduling scheme and generate a calibrated resource scheduling probability distribution includes: The mean and standard deviation of the time deviation of contractor nodes in the start time and duration of the scheduling scheme, as well as the mean and standard deviation of the resource usage deviation in the allocation of various core resources, are extracted from the reputation model. The time standard deviation and the capacity standard deviation for each resource are adjusted according to the correlation confidence level declared in the tender, where a higher confidence level corresponds to a smaller adjusted standard deviation; The statistical characteristics of the deviation are respectively used as the extracted mean time deviation and the adjusted time standard deviation to define the time Gaussian noise model, and the extracted mean occupancy deviation and the adjusted standard deviation of each resource dimension to define the capacity Gaussian noise model, which are applied to the time attribute and resource capacity attribute of each spatiotemporal anchor point of the expected resource scheduling scheme in the bidding. The scheduling scheme with added noise was simulated multiple times using the Monte Carlo sampling method. The frequency of resource occupancy in each discrete spatiotemporal cell was counted, and the frequency was normalized to form a calibrated resource scheduling probability distribution.

[0032] The calibration process retrieves historical scheduling deviation statistics for the contractor node in the current bid from the reputation model. This data includes the mean start time deviation over a time dimension. For example, +5 seconds and standard deviation For example, 15 seconds, and the mean of duration deviation. For example, -120 seconds and standard deviation For example, 300 seconds; simultaneously, the average allocation deviation of each resource item along the resource dimension (CPU, memory, network bandwidth) is retrieved. , , For example =+2 cores and standard deviation , , For example =5. Based on the association confidence level Conf declared in the tender, a value between 0 and 1, for example 0.9, is used with the formula Adjust for historical standard deviation. The confidence level influence factor is preferably [0.7, 0.95], for example, 0.8; and These represent the minimum and maximum possible values ​​for the confidence level, typically 0 and 1. A confidence level of 0.9 will reduce the standard deviation, indicating that the contractor has a high degree of confidence in the accuracy of this scheduling.

[0033] Statistical parameters are applied to the Gaussian noise model. For each key spatiotemporal anchor point in the bidding scheme, such as resource request point, expansion / contraction point, and release point, a random noise is superimposed on the timestamp T and resource capacity parameters (X, Y, Z). A Monte Carlo method is used to perform N independent simulations, for example, N=1000. In each simulation, a complete four-dimensional resource allocation evolution sequence with random noise is generated for the entire scheduling scheme. Simultaneously, the entire task spatiotemporal domain is divided into discrete four-dimensional cells, for example, 1 CPU core × 2GB RAM × 10GB storage × 30 seconds. After 1000 simulations, the number of times each cell is occupied by the simulation sequence is counted. Divide this number by the total number of simulations N, that is... This allows us to obtain a calibrated resource scheduling probability distribution and identify the uncertainties of resources in the future time and space.

[0034] S4, Calculate the entropy reduction of the global spatiotemporal occupancy load distribution caused by the granting of the bid based on the probability distribution.

[0035] The management node maintains a global spatiotemporal four-dimensional grid, with each cell storing an occupancy probability value. Based on the scheduling scheme of all currently confirmed contracts, the management node calculates the occupancy probability of each cell and uses the Shannon entropy formula to calculate the total entropy value of the current global spatiotemporal occupancy load distribution. For bids to be evaluated, the management node overlays the calibrated resource scheduling probability distribution onto the global spatiotemporal grid. It updates the global occupancy probability distribution by synthesizing the original occupancy probability of each affected cell with the occupancy probability of the new task. Finally, it uses the Shannon entropy formula to calculate the new total entropy value after granting the bid. The entropy reduction is... .

[0036] In one implementation, the calculation of the entropy reduction in the global spatiotemporal occupancy load distribution resulting from the granting of a bid based on the probability distribution includes: Discretize the four-dimensional spatiotemporal domain of task execution into a spatiotemporal volumetric raster; For each spatiotemporal volume element i, the expected load of the volume element is obtained by summing the probabilities of the volume element being occupied by all planned tasks. The expected occupancy load of each element is normalized into an occupancy probability. ; Calculate the global load entropy before awarding a new bid. The summation iterates through all Volume elements >0; The calibrated resource scheduling probability distribution of the new bids is added to the global occupancy model, and the expected occupancy load of each element due to all planned tasks and the current bids to be evaluated is recalculated. And normalized to occupancy probability Calculate the global load entropy after the new bid is awarded. ; Calculate the and The difference yields the entropy reduction of the global spatiotemporal load distribution.

[0037] Set up a four-dimensional spatiotemporal grid system covering all task scopes. The system's resolution needs to be set according to the task's precision requirements. For example, the resource dimension might be divided into resource voxels of 1 CPU core × 2GB memory × 10GB storage, and the time dimension into 30-second time slices. For the i-th spatiotemporal voxel in the system, the current expected load is... It is the resource scheduling probability distribution of all awarded contracts j. The sum of probability values ​​for this spatiotemporal element represents the expected intensity of its occupancy under the current plan. To calculate the information entropy, the load needs to be normalized to a probability distribution, making the global total load... Then the occupancy probability of each element is Therefore, the entropy of the current global spatiotemporal occupancy load is calculated using the Shannon entropy formula mentioned above. This value represents the degree of disorder or uncertainty in the distribution of system resources.

[0038] When evaluating a new bid, the calibrated resource scheduling probability distribution will be used. Add it to the global model. At this point, the expected load of each volume element i is updated to... Update global total load. And calculate the new occupancy probability. Based on this new probability distribution, calculate the global load entropy of the system after the bid is awarded. Entropy reduction. pass Calculations show that a positive value is obtained. The value indicates that awarding this bid will make the overall resource distribution of the system more centralized and orderly, and reduce the potential risk of spatiotemporal conflicts, making it a useful indicator in decision evaluation. For example, if =150.7 bits =145.2 bits, then the entropy reduction It is 5.5 bits.

[0039] S5, again, when a bid conflicts with a planned contract, calculate the contract preemption compensation cost through a loss function that combines task migration cost, resource idle cost, and opportunity cost.

[0040] The management node detects conflicts by checking whether there are overlapping units in the spatiotemporal grid with occupancy probabilities greater than a preset threshold between the resource scheduling probability distribution of newly bids and the scheduling schemes of planned contracts. If a conflict is detected, the preempted contract is identified; the preemption compensation cost is then determined. Calculated using a weighted loss function, i.e. Among them, task migration cost Resource idle cost is derived by simulating the rescheduling of preempted tasks and calculating the cost difference between the new and old network topology paths using either Dijkstra's algorithm or the Floyd-Warshall algorithm. Opportunity cost equals the estimated time between the release of resources, the occurrence of preemption, and the receipt of new tasks, multiplied by the unit time maintenance cost. The contract value of the preempted task or the penalty for delayed delivery. Weighting , , It is calculated by dividing the constant k by 1 and adding the historical conflict contribution of the contractor node corresponding to the preempted contract.

[0041] In one implementation, when a bid conflicts with a planned contract again, the contract preemption compensation cost is calculated through a loss function that combines task migration cost, resource idle cost, and opportunity cost, including: Retrieve historical conflict contribution of preempted contracts from a reputation model. And based on the formula Calculate the weights, where k is a preset positive coefficient; Based on the topological distance between the current network node seizing resources and the backup task network node, a preset unit topological distance migration cost table is consulted to obtain the unit topological distance migration cost. The task migration cost is then obtained by multiplying the unit topological distance migration cost by the topological distance. ; Predict the idle time of the resource from being preempted to being reallocated to a standby task. Multiplying this by the unit time holding cost of the resource yields the resource idle cost. ; The opportunity cost is determined by assessing the direct profit loss resulting from the cancellation of the preemptive contract. ; Contract snatching compensation cost .

[0042] When a bid conflicts with a planned contract in terms of time and space and needs to be secured, the historical conflict contribution of the contractor to which the contract is being secured is retrieved from the reputation model. A normalized value, such as 0.6, indicates that the contractor has historically tended to instigate conflict. The preemption penalty weight is calculated using a formula, where k is an adjustment coefficient; for example, if k=5, then W=0.25. This weighting mechanism reduces the compensation cost of preempting a poorly performing contractor. Calculate various costs: task migration costs. Determine the current network location of the preempted resource, such as a computing instance, and plan an optimal data migration path to the backup task node for that resource, calculating the topology distance D, for example, 80 network hops. Based on the computing instance model, retrieve the unit topology distance migration cost (including network energy consumption and bandwidth usage) from a preset cost table, for example, 0.5 yuan / hop. =40 yuan. Resource idle cost. Estimate the total time required from the moment resources are preempted until the execution of backup tasks begins. This includes data migration time, environment initialization time, etc. =1.5 hours. The unit time holding cost of this calculation example, including server energy consumption and equipment depreciation, is 60 yuan / hour. =90 yuan. Opportunity cost : Directly take the expected profit value recorded in the contract that was seized. For example, if the contract is expected to bring in a profit of 200 yuan, then... =200 yuan. The weighted sum of all cost items yields the contract preemption compensation cost. =82.5 yuan.

[0043] S6, the management node calculates a comprehensive evaluation value based on the bidding execution benefits, the entropy reduction, and the contract preemption compensation cost, and awards the contract to the contractor node with the best comprehensive evaluation value.

[0044] In one implementation, the step of calculating the comprehensive evaluation value based on the bidding execution benefits, the entropy reduction, and the contract preemption compensation cost includes: Extracting the declared implementation benefits from the bid ; Obtain the calculated entropy reduction ΔH of the global spatiotemporal occupancy load distribution; Obtain the calculated contract preemption compensation cost If there is no seizure, then =0; Through formula Calculate the comprehensive evaluation value V, where α, β, and γ are preset weighting coefficients used to unify the dimensions and balance the importance of each component.

[0045] A comprehensive evaluation value V is calculated for each bid to be evaluated. This process integrates three core evaluation dimensions and extracts the direct economic benefits that the bid proposal brings to the system. ,For example =500 yuan. Obtain the entropy reduction of the global spatiotemporal load distribution calculated in the previous steps. ,For example =5.5 bits, this value represents the contribution of the bid to the overall stability of the system. Obtain the calculated compensation cost for potential contract preemption caused by this bid. If the bid does not conflict with any existing contracts, then =0; if preemption is required, substitute the calculated value, for example =82.5 yuan. The comprehensive evaluation value is calculated using a linear weighted formula. The weighting coefficients are... , , This is preset by the system administrator based on operational strategies. For example, setting... =1.0, used to directly represent economic benefits; =1.0, used to deduct preemption costs. For Its function is to convert the entropy reduction, measured in information theory, into a value with the same dimensions as economic benefits, a positive value. Values ​​for example =20 yuan / bit. The negative sign in the formula indicates that the greater the entropy reduction, the more ordered the system is, and the smaller the penalty to the evaluation value.

[0046] S7, and after the contract is executed, update the record of the winning contractor node in the reputation model based on the actual execution data and the resulting spatiotemporal conflicts.

[0047] After the contract is completed, the winning contractor node reports the actual spatiotemporal trajectory data. The management node calculates the sum of the absolute values ​​of the deviations between the actual start and end times and the planned times to obtain the scheduling deviation for this task. The exponential moving average algorithm is then used to update the historical scheduling deviation record of the contractor node in the reputation model. ,in This is the old average deviation. This is due to a scheduling deviation in this task. The learning rate is used. Simultaneously, the monitoring system of the management node records actual conflict events with other resources during the task execution process. A conflict score is calculated based on the number and severity of these events, and this score is added to the conflict contribution record of the winning contractor node.

[0048] In one implementation, updating the record of the winning contractor node in the reputation model based on actual execution data and resulting spatiotemporal conflicts includes: After the contract is completed, collect the actual resource allocation sequence and time log of the contractor's resources; The actual resource allocation sequence and time log are compared with the expected resource scheduling plan submitted at the time of winning the bid. The actual deviations in start time, duration, and resource capacity are calculated. The historical deviation mean and standard deviation records of the contractor in the reputation model are updated using the exponential weighted moving average method. Analyze the spatiotemporal conflict events caused by the contractor and confirmed during the actual execution process, assess the disturbances caused by the conflict events to the system, and calculate the realized conflict contribution of this task. The realized conflict contribution is incorporated into the contractor's historical conflict contribution record in the reputation model using an exponentially weighted moving average method.

[0049] After the mission is completed, the winning contractor's execution data is collected and processed, the planned and actual timestamps are compared, and the start time deviation of this mission is calculated. For example, +35 seconds and duration deviation. For example, -180 seconds. For resource capacity allocation, the overall resource capacity deviation value is obtained by calculating the discrete Friesian distance between the planned allocation sequence and the actual resource occupancy sequence. For example, 25.3 resource units. The contractor's long-term statistics in the credit model are updated using the Exponentially Weighted Moving Average (EWMA) method. Variance is updated in a similar manner, adjusting for standard deviation.

[0050] Analyze the conflict logs to identify verified conflict events caused by the contractor's actions. Represent these events according to their severity to determine the realized conflict contribution for this mission. For example, a resource contention with another task causing a response delay is worth 5 points, a delay in another task is worth 1 point per minute of delay, and a system memory overflow or crash is worth 50 points. If, in this task, the contractor caused one resource contention and one 3-minute delay, then... =8. The historical conflict contribution record is also updated using the EWMA method: A reputation model can consistently represent each contractor's true performance capability and reliability.

[0051] Furthermore, a set of control experiments were conducted to verify the scheme of the present invention. The control experiments were carried out on a distributed computing system simulation platform based on CloudSim secondary development. The experimental hardware infrastructure layer was set to include 3 regions, each region with 4 availability zones, and each availability zone containing several racks and nodes, with a total of 2000 standard physical servers deployed as a distributed resource pool; the physical configuration of a single server was randomly distributed between [32 cores, 128 cores] CPU, [128GB, 512GB] memory, and [10Gbps, 40Gbps] network bandwidth. The task load data came from an open-source trajectory dataset from a real cloud data center, and after secondary sampling, 10,000 computing task request sequences with multi-dimensional resource constraints were generated. The task arrival rate followed a Poisson distribution, and the execution duration of a single task ranged from 10 minutes to 12 hours. The network topology was configured with one management node and 50 contractor nodes representing different stakeholders. To simulate the varying service quality in a real distributed resource market, the initial historical scheduling bias and conflict contribution of each contractor node were randomly initialized within a reasonable range. The simulation period was set to run continuously for 168 hours. All comparison schemes were run independently under identical task arrival sequences and initial resource topology configurations, and the average value of multiple experiments was taken to ensure statistical significance.

[0052] The experiment set up four schemes for comparison: Scheme A serves as the baseline, adopting a deterministic scheduling plan with the sole objective of maximizing the direct benefits of the task and using a fixed preemption cost; Scheme B, based on Scheme A, inputs a resource scheduling probability distribution based on reputation calibration; Scheme C, based on Scheme B, further inputs the reduction of global spatiotemporal occupancy load entropy as a decision factor; Scheme D is the complete scheme proposed in this paper, which, based on Scheme C, inputs a preemption compensation cost associated with historical conflict contribution.

[0053] The specific data collected in the experiment are as follows. Scheme A achieved an average task success rate of 85.2%, with 112 confirmed spatiotemporal conflicts occurring within the period, an average load entropy of 165.3 bits, and an average resource idle time of 2.5 hours. Scheme B improved the average task success rate to 91.5%, reduced the number of spatiotemporal conflicts to 65, lowered the average load entropy to 158.1 bits, and achieved an average resource idle time of 2.3 hours. Scheme C achieved an average task success rate of 94.3%, further reduced the number of spatiotemporal conflicts to 48, optimized the average load entropy to 147.2 bits, and achieved an average resource idle time of 2.1 hours. The complete Scheme D performed best, with an average task success rate as high as 96.1%, only 31 spatiotemporal conflicts, a stable average load entropy of 142.5 bits, and a shortened average resource idle time to 2.0 hours. (See reference...) Figure 2 As shown, by Figure 2 It can be seen that with the gradual input of reputation model, probabilistic scheduling, entropy optimization and preemption cost mechanism, the task success rate shows a step-by-step upward trend. Scheme D improves by nearly 11 percentage points compared with the baseline scheme A, indicating that the proposed multi-level optimization strategy can improve the reliability of task completion.

[0054] See Figure 3 As shown, by Figure 3 It can be seen that the number of spatiotemporal conflicts continued to decrease with the improvement of the scheme, from 112 times in scheme A to 31 times in scheme D, a decrease of more than 70%. This shows that probabilistic scheduling and global entropy optimization can predict and avoid spatiotemporal resource overlap, reducing the risk of system operation. Figure 4 The optimization effect on average resource idle time was demonstrated, reducing it from 2.5 hours to 2.0 hours. This indicates that under reasonable scheduling and conflict avoidance mechanisms, resource idling waiting time is reduced, and overall utilization is improved. The improvement from Scheme A to Scheme B proves that using a probabilistic model instead of a deterministic model can predict and avoid potential conflicts, which is the primary reason for reducing the number of conflicts. The performance leap from Scheme B to Scheme C reflects the value of entropy reduction as an evaluation indicator. It guides the system to optimize resource layout from a global perspective, avoids the formation of congestion hotspots, and thus improves overall scheduling efficiency while reducing conflicts.

[0055] The improvement from Option C to Option D demonstrates the long-term benefits of the reputation mechanism. By inputting a preemption cost linked to historical conflict contributions, underperforming contractors can be eliminated at a lower cost, creating a positive incentive and elimination mechanism. This not only further reduces conflicts caused by unreliable executors in the short term but also fundamentally optimizes the quality of the contractor group participating in the system, achieving the highest task success rate and system stability, proving the synergistic value of the various components of the solution.

[0056] Secondly, the present invention also provides a market-based distributed resource negotiation and spatiotemporal conflict resolution system, comprising the following modules: The generation module is used by management nodes to generate a spatiotemporal demand Bloom filter based on task spatiotemporal constraints and broadcast task announcements. At the same time, it maintains a reputation model that records the historical scheduling deviations and conflict contributions of contractor nodes. After matching the spatiotemporal demand Bloom filter, contractor nodes submit bids that include expected resource scheduling schemes and associated confidence levels. The calculation module is used to manage the decision-making evaluation of received bids by the nodes, combine the historical scheduling deviations in the reputation model with the correlation confidence of the bids, calibrate the expected resource scheduling scheme to generate a calibrated resource scheduling probability distribution; calculate the entropy reduction of the global spatiotemporal load distribution caused by granting the bid based on the probability distribution; and calculate the contract preemption compensation cost when the bid conflicts with the planned contract through a loss function that combines task migration cost, resource idle cost and opportunity cost. The update module is used to manage nodes to calculate a comprehensive evaluation value based on the bidding execution benefits, the entropy reduction, and the contract preemption compensation cost, and to award the contract to the contractor node with the best comprehensive evaluation value; and after the contract is executed, to update the record of the winning contractor node in the reputation model based on the actual execution data and the resulting spatiotemporal conflicts.

[0057] It should be understood that in the foregoing description of the embodiments in this specification, various features are combined in a single embodiment, drawing, or description for the purpose of simplifying the description and to aid in understanding a feature. However, this does not mean that the combination of these features is necessary, and those skilled in the art, upon reading this specification, may readily identify some of the devices as separate embodiments. That is, the embodiments in this specification can also be understood as an integration of multiple secondary embodiments. And the content of each secondary embodiment is valid even if it contains fewer than all the features of a single foregoing disclosed embodiment.

Claims

1. A market-based distributed resource negotiation and spatiotemporal conflict resolution method, characterized in that, Includes the following steps: The management node generates a spatiotemporal demand Bloom filter based on the task's spatiotemporal constraints and broadcasts the task announcement. At the same time, it maintains a reputation model that records the historical scheduling deviations and conflict contributions of contractor nodes. After matching the spatiotemporal demand Bloom filter, the contractor node submits a bid that includes the expected resource scheduling scheme and the associated confidence level. The management node evaluates the received bids and, combining the historical scheduling deviations in the reputation model with the correlation confidence of the bids, calibrates the expected resource scheduling scheme to generate a calibrated resource scheduling probability distribution; based on the probability distribution, it calculates the entropy reduction of the global spatiotemporal load distribution caused by granting the bid; and when the bid conflicts with a planned contract, it calculates the contract preemption compensation cost through a loss function that combines task migration cost, resource idle cost, and opportunity cost. The management node calculates a comprehensive evaluation value based on the bidding execution benefits, the entropy reduction, and the contract preemption compensation cost, and then awards the contract to the contractor node with the best comprehensive evaluation value. After the contract is executed, the record of the winning contractor node in the reputation model is updated based on the actual execution data and the resulting spatiotemporal conflicts.

2. The method according to claim 1, characterized in that, The step of combining historical scheduling deviations in the reputation model with the confidence level of the bids to calibrate the expected resource scheduling scheme and generate a calibrated resource scheduling probability distribution includes: The mean and standard deviation of the time deviation of contractor nodes in the start time and duration of the scheduling scheme, as well as the mean and standard deviation of the resource usage deviation in the allocation of various core resources, are extracted from the reputation model. The time standard deviation and the capacity standard deviation for each resource are adjusted according to the correlation confidence level declared in the tender, where a higher confidence level corresponds to a smaller adjusted standard deviation; The statistical characteristics of the deviation are respectively used as the extracted mean time deviation and the adjusted time standard deviation to define the time Gaussian noise model, and the extracted mean occupancy deviation and the adjusted standard deviation of each resource dimension to define the capacity Gaussian noise model, which are applied to the time attribute and resource capacity attribute of each spatiotemporal anchor point of the expected resource scheduling scheme in the bidding. The scheduling scheme with added noise was simulated multiple times using the Monte Carlo sampling method. The frequency of resource occupancy in each discrete spatiotemporal cell was counted, and the frequency was normalized to form a calibrated resource scheduling probability distribution.

3. The method according to claim 1, characterized in that, The calculation of the entropy reduction in the global spatiotemporal occupancy load distribution resulting from the granting of bids based on the probability distribution includes: Discretize the four-dimensional spatiotemporal domain of task execution into a spatiotemporal volumetric raster; For each spatiotemporal volume element i, the expected load of the volume element is obtained by summing the probabilities of the volume element being occupied by all planned tasks. The expected occupancy load of each element is normalized into an occupancy probability. ; Calculate the global load entropy before awarding a new bid. The summation iterates through all Volume elements >0; The calibrated resource scheduling probability distribution of the new bids is added to the global occupancy model, and the expected occupancy load of each element due to all planned tasks and the current bids to be evaluated is recalculated. And normalized to occupancy probability Calculate the global load entropy after the new bid is awarded. ; Calculate the and The difference yields the entropy reduction of the global spatiotemporal occupancy load distribution.

4. The method according to claim 1, characterized in that, When a bid conflicts with a planned contract, the contract preemption compensation cost is calculated using a loss function that combines task migration cost, resource idle cost, and opportunity cost. This includes: Retrieve historical conflict contribution of preempted contracts from a reputation model. And based on the formula Calculate the weights, where k is a preset positive coefficient; Based on the topological distance between the current network node seizing resources and the backup task network node, a preset unit topological distance migration cost table is consulted to obtain the unit topological distance migration cost. The task migration cost is then obtained by multiplying the unit topological distance migration cost by the topological distance. ; Predict the idle time of the resource from being preempted to being reallocated to a standby task. Multiplying this by the unit time holding cost of the resource yields the resource idle cost. ; The opportunity cost is determined by assessing the direct profit loss resulting from the cancellation of the preemptive contract. ; Contract snatching compensation cost .

5. The method according to claim 4, characterized in that, The calculation of the comprehensive evaluation value based on the bidding execution benefits, the entropy reduction, and the contract preemption compensation cost includes: Extracting the declared implementation benefits from the bid ; Obtain the calculated entropy reduction ΔH of the global spatiotemporal occupancy load distribution; Obtain the calculated contract preemption compensation cost If there is no seizure, then =0; Through formula Calculate the comprehensive evaluation value V, where α, β, and γ are preset weighting coefficients used to unify the dimensions and balance the importance of each component.

6. The method according to claim 1, characterized in that, The management node generates a spatiotemporal requirement Bloom filter based on task spatiotemporal constraints, including: Extract the task's time window and the set of network topology logical boundaries; The time window is divided into a series of discrete time slice identifiers according to a preset time unit; The logical boundary is divided according to the preset server cluster partitioning rules to obtain a series of discrete spatial cell identifiers; For each time-slice-space cell combination required by the task, the identifier of the combination is used as input, multiple hash values ​​are calculated through multiple different hash functions, and the bit position corresponding to the index of the hash value in the Bloom filter is set to 1.

7. The method according to claim 1, characterized in that, The step of updating the record of the winning contractor node in the reputation model based on actual execution data and the resulting spatiotemporal conflicts includes: After the contract is completed, collect the actual resource allocation sequence and time log of the contractor's resources; The actual resource allocation sequence and time log are compared with the expected resource scheduling plan submitted at the time of winning the bid. The actual deviations in start time, duration, and resource capacity are calculated. The historical deviation mean and standard deviation records of the contractor in the reputation model are updated using the exponential weighted moving average method. Analyze the spatiotemporal conflict events caused by the contractor and confirmed during the actual execution process, assess the disturbances caused by the conflict events to the system, and calculate the realized conflict contribution of this task. The realized conflict contribution is incorporated into the contractor's historical conflict contribution record in the reputation model using an exponentially weighted moving average method.

8. A market-based distributed resource negotiation and spatiotemporal conflict resolution system, characterized in that, Includes the following modules: The generation module is used by management nodes to generate a spatiotemporal demand Bloom filter based on task spatiotemporal constraints and broadcast task announcements. At the same time, it maintains a reputation model that records the historical scheduling deviations and conflict contributions of contractor nodes. After matching the spatiotemporal demand Bloom filter, contractor nodes submit bids that include expected resource scheduling schemes and associated confidence levels. The calculation module is used to manage the decision-making evaluation of received bids by the nodes, combine the historical scheduling deviations in the reputation model with the correlation confidence of the bids, calibrate the expected resource scheduling scheme to generate a calibrated resource scheduling probability distribution; calculate the entropy reduction of the global spatiotemporal load distribution caused by granting the bid based on the probability distribution; and calculate the contract preemption compensation cost when the bid conflicts with the planned contract through a loss function that combines task migration cost, resource idle cost and opportunity cost. The update module is used by the management node to calculate a comprehensive evaluation value based on the bidding execution benefits, the entropy reduction, and the contract preemption compensation cost, and to award the contract to the contractor node with the best comprehensive evaluation value. After the contract is executed, the record of the winning contractor node in the reputation model is updated based on the actual execution data and the resulting spatiotemporal conflicts.

9. The system according to claim 8, characterized in that, The step of combining historical scheduling deviations in the reputation model with the confidence level of the bids to calibrate the expected resource scheduling scheme and generate a calibrated resource scheduling probability distribution includes: The mean and standard deviation of the time deviation of contractor nodes in the start time and duration of the scheduling scheme, as well as the mean and standard deviation of the resource usage deviation in the allocation of various core resources, are extracted from the reputation model. The time standard deviation and the capacity standard deviation for each resource are adjusted according to the correlation confidence level declared in the tender, where a higher confidence level corresponds to a smaller adjusted standard deviation; The statistical characteristics of the deviation are respectively used as the extracted mean time deviation and the adjusted time standard deviation to define the time Gaussian noise model, and the extracted mean occupancy deviation and the adjusted standard deviation of each resource dimension to define the capacity Gaussian noise model, which are applied to the time attribute and resource capacity attribute of each spatiotemporal anchor point of the expected resource scheduling scheme in the bidding. The scheduling scheme with added noise was simulated multiple times using the Monte Carlo sampling method. The frequency of resource occupancy in each discrete spatiotemporal cell was counted, and the frequency was normalized to form a calibrated resource scheduling probability distribution.

10. The system according to claim 8, characterized in that, The calculation of the entropy reduction in the global spatiotemporal occupancy load distribution resulting from the granting of bids based on the probability distribution includes: Discretize the four-dimensional spatiotemporal domain of task execution into a spatiotemporal volumetric raster; For each spatiotemporal volume element i, the expected load of the volume element is obtained by summing the probabilities of the volume element being occupied by all planned tasks. The expected occupancy load of each element is normalized into an occupancy probability. ; Calculate the global load entropy before awarding a new bid. The summation iterates through all Volume elements >0; The calibrated resource scheduling probability distribution of the new bids is added to the global occupancy model, and the expected occupancy load of each element due to all planned tasks and the current bids to be evaluated is recalculated. And normalized to occupancy probability Calculate the global load entropy after the new bid is awarded. ; Calculate the and The difference yields the entropy reduction of the global spatiotemporal occupancy load distribution.

Images