Bait palatability dynamic optimization method and system based on termite pheromone release rule

Abstract

The application discloses a bait palatability dynamic optimization method and system based on termite pheromone release rules, acquires pheromone release state, bait state, termite feeding behavior and environmental parameter data, constructs a time sequence hypergraph structure and performs representation learning to obtain a time sequence representation of a bait station operation state; a world model of DRAMA model type reinforcement learning is constructed based on the state representation to predict bait station state evolution under the action of pheromone controlled release and bait adjustment, and generate a synergistic regulation strategy; the pheromone release process and bait state are dynamically adjusted and fed back according to the strategy. The application realizes synergistic optimization of pheromone controlled release and bait adjustment, and improves the continuity of bait station operation and application reliability.

Description

Technical Field

[0001] This invention relates to the field of intelligent pest control and agricultural informatization, and in particular to a method and system for dynamic optimization of bait palatability based on the release patterns of termite pheromones. Background Technology

[0002] With the increasing density of urban buildings and the widespread use of wooden structures, termite infestations are characterized by their high degree of concealment, wide spread, and long treatment cycles. Termite bait stations, as an important control technology, have been applied in fields such as landscaping, building construction, municipal facilities, and cultural relic protection. Current termite bait station technology mainly relies on bait attraction and pheromone guidance mechanisms. By releasing specific pheromones within the bait station, termites are guided to continuously feed on bait containing regulatory components, thereby achieving intervention and control of the termite colony. This type of technology typically focuses on pheromone ratio design, bait composition improvement, and optimized placement, and maintains the bait station's operational status through regular manual inspections and empirical adjustments.

[0003] In practice, some existing technologies incorporate sensors to collect environmental parameters and feeding behavior data within the bait station, adjusting pheromone release levels or changing bait forms based on historical data to extend the feeding cycle. However, these methods often employ fixed rules or single-timescale data analysis, failing to adequately characterize the temporal changes in termite feeding behavior and making it difficult to accurately reflect the multidimensional relationships between pheromone release status, bait status, and environmental changes. During bait station operation, when the pheromone release rhythm deviates from the actual termite feeding rhythm, it can easily lead to decreased feeding frequency, interrupted feeding paths, or colony migration, making it difficult to maintain bait palatability in the long term.

[0004] Existing technologies for bait station status modeling mostly employ simple state descriptions or static feature combinations, lacking the ability to systematically model the overall evolution of bait station operation over time. This makes it difficult to effectively predict the operational status under different combinations of pheromone release parameters and bait adjustment parameters before adjustments are made. Without refined state prediction and decision-making mechanisms, bait station adjustment behavior often lags behind actual operational changes, further exacerbating the risk of feeding interruptions and affecting the stability and continuity of termite control effectiveness.

[0005] Therefore, how to provide a dynamic optimization method and system for pheromone-driven bait palatability based on DRAMA model-based reinforcement learning combined with temporal hypergraph representation learning is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] One objective of this invention is to propose a dynamic optimization method and system for bait palatability based on termite pheromone release patterns. Addressing the problem of feeding interruptions that easily occur in existing technologies when maintaining stable bait palatability based on pheromone release patterns, this invention proposes a technical solution that constructs a temporal hypergraph using multi-source temporal data to obtain a temporal representation of the bait station's operating state. It then constructs a bait station state prediction model and decision-making mechanism based on DRAMA model-based reinforcement learning, collaboratively optimizing pheromone release control parameters and bait adjustment parameters. By introducing a closed loop of state prediction and decision feedback, this invention achieves continuous modeling, strategy generation, and dynamic updating of the bait station's operating state, effectively reducing the risk of feeding interruptions and improving the continuity, stability, and adaptability of termite bait station operation regulation.

[0007] The method and system for dynamic optimization of bait palatability based on termite pheromone release patterns according to embodiments of the present invention include the following steps:

[0008] S1: During the operation of the termite bait station, the operation data of the bait station is collected periodically. The operation data includes pheromone release status data, bait status data, termite feeding behavior data and environmental parameter data. The operation data is then processed for time synchronization to obtain multi-source time-series data.

[0009] S2: Construct a time-series hypergraph structure based on multi-source time-series data, map pheromone release state, bait state, termite feeding behavior and environmental parameters as hypergraph nodes, and construct hyperedges for the association relationships between multiple nodes to form time-series hypergraph data corresponding to the bait station operation status;

[0010] S3: Perform temporal hypergraph representation learning on temporal hypergraph data to obtain state representation results that reflect the changes in node state and hyperedge relationship over time;

[0011] S4: Using the state representation results as input, construct a world model based on DRAMA model-based reinforcement learning, model the evolution process of bait station state under the action of pheromone release parameters and bait regulation parameters, and obtain a bait station state prediction model.

[0012] S5: Based on the bait station state prediction model, perform reinforcement learning decision calculations to generate pheromone controlled release strategies and bait adjustment strategies;

[0013] S6: Adjust the pheromone release process and bait state of the bait station according to the pheromone controlled release strategy and bait adjustment strategy to obtain the adjusted bait station operation state;

[0014] S7: Monitor the adjusted operating status of the bait station and use the monitored data to update the time-series hypergraph data and the bait station status prediction model, forming a dynamic optimization process for bait palatability driven by pheromone controlled release.

[0015] Optionally, S1 specifically includes:

[0016] S11: During the operation of the termite bait station, pheromone release status data is acquired through collection devices set inside and around the bait station. The pheromone release status data is used to characterize the release time, release frequency and release intensity of pheromones within a preset collection cycle.

[0017] S12: During the collection period, acquire bait state data. The bait state data is used to characterize the physical and chemical state of the bait in the bait station and form bait state time series data at a preset sampling interval.

[0018] S13: During the collection period, acquire termite feeding behavior data. The termite feeding behavior data is used to characterize the number of times termites feed, the duration of feeding, and the distribution of feeding paths in the bait station, and form termite feeding behavior time series data in chronological order.

[0019] S14: During the collection period, acquire environmental parameter data. The environmental parameter data is used to characterize the temperature, humidity and air flow parameters in the environment where the bait station is located, and form environmental parameter time series data.

[0020] S15: Perform unified time reference alignment processing on pheromone release status data, bait status data, termite feeding behavior data, and environmental parameter data. The time reference alignment processing uses a preset timestamp as the reference and resamples data from different collection frequencies to the same time scale to obtain multi-source time-series data after time synchronization.

[0021] Optionally, S2 specifically includes:

[0022] S21: Based on multi-source time-series data, pheromone release status data, bait status data, termite feeding behavior data, and environmental parameter data are classified and labeled according to the data source type to form a multi-source time-series data set with data type labels.

[0023] S22: Extract the state variables corresponding to pheromone release state data, bait state data, termite feeding behavior data, and environmental parameter data from the multi-source time series data set, and define them as node sets in the time series hypergraph to obtain the node initialization results.

[0024] S23: Under the same time index, based on the synchronous change relationship between different state variables in multi-source time series data, construct a set of association relationships connecting multiple nodes. Each set of association relationships corresponds to a hyperedge, thus obtaining a set of hyperedges.

[0025] S24: Organize the node set and the hyperedge set into a time series, arrange the node states and hyperedge connections under different time indices in chronological order, and form the initial structure of the temporal hypergraph with time evolution characteristics;

[0026] S25: Perform consistency verification on the node state data and hyperedge connection relationships in the initial structure of the temporal hypergraph, remove node states and hyperedge connection relationships with missing time indexes or incomplete data, and obtain the verified temporal hypergraph structure.

[0027] S26: Output the verified time-series hypergraph structure as the time-series hypergraph data corresponding to the bait station's operating status.

[0028] Optionally, S3 specifically includes:

[0029] S31: Based on the time-series hypergraph data, the node state data corresponding to each time index in the time-series hypergraph data is numerically normalized to obtain the normalized node state sequence data.

[0030] S32: Based on the normalized node state sequence data, perform joint feature encoding on multiple node states connected by hyperedges under the same time index to generate hyperedge feature data that characterizes the association relationship of multiple nodes within the hyperedge;

[0031] S33: Arrange the node state sequence data and the corresponding hyperedge feature data in a temporal sequence according to the time index order to construct the node-hyperedge joint temporal feature sequence;

[0032] S34: Perform time-series modeling on the node-hyperedge joint time-series feature sequence, calculate the correlation mapping between node state changes and hyperedge relationship changes under different time indices, and obtain intermediate time-series relationship feature data;

[0033] S35: Based on intermediate temporal relationship feature data, update the node state and hyperedge relationship corresponding to each time index to generate a temporal hypergraph state representation result that reflects the evolution of node state and hyperedge relationship over time.

[0034] Optionally, S4 specifically includes:

[0035] S41: Based on the temporal hypergraph state representation results, construct the bait station operation state sequence according to the time index order. The bait station operation state sequence is used to describe the state change process of the bait station under continuous time index.

[0036] S42: Based on the bait station operation state sequence, construct the world model input data structure based on DRAMA model reinforcement learning. The world model input data structure includes the bait station operation state corresponding to the current time index and the bait station operation state corresponding to the historical time index.

[0037] S43: Introduce pheromone controlled release parameters and bait adjustment parameters as state transition conditions in the world model, simulate the evolution path of bait station operation under different parameter values, and obtain multiple sets of candidate state evolution sequences.

[0038] S44: Expand multiple candidate state evolution sequences according to time index, calculate the changes in the operating state of the bait station under multiple future time indices, and form a state evolution prediction sequence;

[0039] S45: Based on the state evolution prediction sequence, a bait station state prediction model is constructed. The bait station state prediction model is used to describe the temporal evolution relationship of the bait station's operating state under the influence of pheromone controlled release parameters and bait adjustment parameters.

[0040] Optionally, S5 specifically includes:

[0041] S51: Based on the bait station status prediction model, and based on the bait station status prediction model, extract the bait station operation status prediction results under multiple future time indices to form a status prediction sequence.

[0042] S52: Based on the state prediction sequence, construct a set of reinforcement learning decision states. The set of reinforcement learning decision states is used to characterize the decision input conditions corresponding to the operation state of the bait station under different time indices.

[0043] S53: In the reinforcement learning decision-making process, the pheromone release control parameter and the bait adjustment parameter are defined as optional decision variables, and a set of parameter combinations corresponding to the set of reinforcement learning decision states is generated under each time index;

[0044] S54: Based on the set of decision states and parameter combinations in reinforcement learning, perform strategy evaluation calculations on each set of pheromone release parameters and bait regulation parameters to obtain a sequence of decision results corresponding to different parameter combinations;

[0045] S55: Based on the decision result sequence, screen and determine the pheromone controlled release parameters and the bait adjustment parameters to generate the target pheromone controlled release strategy and the target bait adjustment strategy.

[0046] Optionally, S6 specifically includes:

[0047] S61: Based on the target pheromone controlled release strategy, read the pheromone controlled release parameter values ​​corresponding to different time indices in the target pheromone controlled release strategy to form a pheromone release control parameter sequence;

[0048] S62: According to the pheromone release control parameter sequence, the release time, release frequency and release intensity of the pheromone release device inside the bait station are set under the corresponding time index to complete the adjustment of the pheromone release process and obtain the adjusted pheromone release state.

[0049] S63: Based on the target bait adjustment strategy, read the bait adjustment parameter values ​​corresponding to different time indices in the target bait adjustment strategy to form a bait state adjustment parameter sequence;

[0050] S64: According to the bait state adjustment parameter sequence, adjust the physical and chemical state parameters of the bait in the bait station under the corresponding time index to complete the bait state adjustment and obtain the adjusted bait state;

[0051] S65: Combine and update the pheromone release state after the pheromone release process is regulated with the bait state after the bait state is regulated to generate the regulated bait station operation status data.

[0052] Optionally, S7 specifically includes:

[0053] S71: After the operation status of the bait station is adjusted, the operation status data of the adjusted bait station is collected according to the preset monitoring cycle. The operation status data includes the adjusted pheromone release status data, the adjusted bait status data, the termite feeding behavior data, and the environmental parameter data.

[0054] S72: Perform time-marking processing on the collected bait station operation status data, associate the operation status data with the corresponding time index, and form operation status time-series data with time index identifier;

[0055] S73: Based on the time-series data of the operating status with time index identifiers, the changes in pheromone release status, bait status, and termite feeding behavior are jointly processed to obtain the feedback operating status sequence;

[0056] S74: Reorganize the feedback running status sequence into a multi-source time series data structure according to the data type, keeping the data structure of the multi-source time series data consistent, and obtain the feedback multi-source time series data;

[0057] S75: Input multi-source time-series data to update the time-series hypergraph data, perform node state update and hyperedge relationship update, and generate updated time-series hypergraph data;

[0058] S76: Based on the updated time-series hypergraph data, the bait station state prediction model is updated to obtain the updated bait station state prediction model.

[0059] Optional, including the following modules:

[0060] The data acquisition module is used to periodically collect pheromone release status data, bait status data, termite feeding behavior data, and environmental parameter data during the operation of the termite bait station. It also performs unified time reference alignment processing on the collected data and outputs multi-source time series data.

[0061] The temporal hypergraph construction module is used to identify data types, extract state variables, and generate nodes based on multi-source time-series data. It also constructs hyperedge relationships connecting multiple nodes under the same time index, organizes node states and hyperedge relationships in a time series, and forms temporal hypergraph data corresponding to the bait station's operating status.

[0062] The temporal hypergraph representation learning module is used to perform node state normalization, hyperedge joint feature encoding, node-hyperedge joint temporal feature sequence construction, and temporal modeling processing on temporal hypergraph data, and outputs temporal hypergraph state representation results that reflect the changes of node state and hyperedge relationship over time.

[0063] The bait station state prediction modeling module is used to construct the bait station operation state sequence based on the temporal hypergraph state representation results, and introduce pheromone controlled release parameters and bait adjustment parameters to model the evolution process of the bait station operation state and generate a bait station state prediction model.

[0064] The reinforcement learning decision generation module is used to form a state prediction sequence based on the bait station state prediction model, construct a set of reinforcement learning decision states, evaluate and screen the pheromone release control parameters and bait regulation parameters, and output the target pheromone release control strategy and the target bait regulation strategy.

[0065] The bait station adjustment execution module is used to set parameters for the release time, release frequency and release intensity of the pheromone release device inside the bait station according to the target pheromone controlled release strategy, and to adjust the physical state parameters and chemical state parameters of the bait in the bait station according to the target bait adjustment strategy, and generate adjusted bait station operation status data.

[0066] The operation status feedback and update module is used to continuously monitor the adjusted bait station operation status data and send the monitored data back to the operation data acquisition module and the time series hypergraph construction module to update the time series hypergraph data and the bait station status prediction model, forming a closed-loop operation structure.

[0067] The beneficial effects of this invention are:

[0068] This invention collects pheromone release state data, bait state data, termite feeding behavior data, and environmental parameter data. Based on multi-source time-series data, it constructs a temporal hypergraph structure and combines this with temporal hypergraph representation learning to obtain a temporal representation of the bait station's operational state. This achieves continuous modeling and unified expression of the termite bait station's operational state, improving the completeness and temporal consistency of the bait station's state representation and enhancing the ability to characterize the relationship between termite feeding behavior and pheromone release. By introducing a world model based on DRAMA model-style reinforcement learning, this invention models and predicts the evolution of the bait station's operational state under the influence of pheromone control parameters and bait regulation parameters. This improves the state prediction capability during strategy generation and demonstrates strong stability and adaptability in long-term termite bait station operation scenarios. In terms of the coordinated control of pheromone control and bait regulation, this invention solves the problem of feeding interruptions during bait palatability maintenance in existing technologies through reinforcement learning decision-making mechanisms and operational state feedback update mechanisms. This achieves continuous optimization of the bait station's operational state, improving the operational continuity and application reliability of termite bait stations in pest control applications. Attached Figure Description

[0069] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0070] Figure 1 This is a schematic diagram of the dynamic optimization method and system flow for bait palatability based on the release pattern of termite pheromones proposed in this invention.

[0071] Figure 2 This is a schematic diagram of the collaborative decision-making mechanism of temporal hypergraph representation learning and reinforcement learning world model proposed in this invention. Detailed Implementation

[0072] Combination Figures 1-2 The present invention will be described in further detail below. These accompanying drawings are simplified schematic diagrams, illustrating only the basic structure of the invention and showing the main components relevant to the invention. Figure 1 and Figure 2 The present invention provides a method and system for dynamic optimization of bait palatability based on termite pheromone release patterns, comprising the following steps:

[0073] S1: During the operation of the termite bait station, the operation data of the bait station is collected periodically. The operation data includes pheromone release status data, bait status data, termite feeding behavior data and environmental parameter data. The operation data is then processed for time synchronization to obtain multi-source time-series data.

[0074] S2: Construct a time-series hypergraph structure based on multi-source time-series data, map pheromone release state, bait state, termite feeding behavior and environmental parameters as hypergraph nodes, and construct hyperedges for the association relationships between multiple nodes to form time-series hypergraph data corresponding to the bait station operation status;

[0075] S3: Perform temporal hypergraph representation learning on temporal hypergraph data to obtain state representation results that reflect the changes in node state and hyperedge relationship over time;

[0076] S4: Using the state representation results as input, construct a world model based on DRAMA model-based reinforcement learning, model the evolution process of bait station state under the action of pheromone release parameters and bait regulation parameters, and obtain a bait station state prediction model.

[0077] S5: Based on the bait station state prediction model, perform reinforcement learning decision calculations to generate pheromone controlled release strategies and bait adjustment strategies;

[0078] S6: Adjust the pheromone release process and bait state of the bait station according to the pheromone controlled release strategy and bait adjustment strategy to obtain the adjusted bait station operation state;

[0079] S7: Monitor the adjusted operating status of the bait station and use the monitored data to update the time-series hypergraph data and the bait station status prediction model, forming a dynamic optimization process for bait palatability driven by pheromone controlled release.

[0080] In this embodiment, S1 specifically includes:

[0081] S11: During the operation of the termite bait station, pheromone release status data is acquired through collection devices set inside and around the bait station. The pheromone release status data is used to characterize the release time, release frequency and release intensity of pheromones within a preset collection cycle.

[0082] S12: During the collection period, acquire bait state data. The bait state data is used to characterize the physical and chemical state of the bait in the bait station and form bait state time series data at a preset sampling interval.

[0083] S13: During the collection period, acquire termite feeding behavior data. The termite feeding behavior data is used to characterize the number of times termites feed, the duration of feeding, and the distribution of feeding paths in the bait station, and form termite feeding behavior time series data in chronological order.

[0084] S14: During the collection period, acquire environmental parameter data. The environmental parameter data is used to characterize the temperature, humidity and air flow parameters in the environment where the bait station is located, and form environmental parameter time series data.

[0085] S15: Perform unified time reference alignment processing on pheromone release status data, bait status data, termite feeding behavior data, and environmental parameter data. The time reference alignment processing uses a preset timestamp as the reference and resamples data from different collection frequencies to the same time scale to obtain multi-source time-series data after time synchronization.

[0086] In this embodiment, S2 specifically includes:

[0087] S21: Based on multi-source time-series data, pheromone release status data, bait status data, termite feeding behavior data, and environmental parameter data are classified and labeled according to the data source type to form a multi-source time-series data set with data type labels.

[0088] S22: Extract the state variables corresponding to pheromone release state data, bait state data, termite feeding behavior data, and environmental parameter data from the multi-source time series data set, and define them as node sets in the time series hypergraph to obtain the node initialization results.

[0089] S23: Under the same time index, based on the synchronous change relationship between different state variables in multi-source time series data, construct a set of association relationships connecting multiple nodes. Each set of association relationships corresponds to a hyperedge, thus obtaining a set of hyperedges.

[0090] S24: Organize the node set and the hyperedge set into a time series, arrange the node states and hyperedge connections under different time indices in chronological order, and form the initial structure of the temporal hypergraph with time evolution characteristics;

[0091] S25: Perform consistency verification on the node state data and hyperedge connection relationships in the initial structure of the temporal hypergraph, remove node states and hyperedge connection relationships with missing time indexes or incomplete data, and obtain the verified temporal hypergraph structure.

[0092] S26: Output the verified time-series hypergraph structure as the time-series hypergraph data corresponding to the bait station's operating status.

[0093] In this embodiment, S3 specifically includes:

[0094] S31: Based on the time-series hypergraph data, the node state data corresponding to each time index in the time-series hypergraph data is numerically normalized to obtain the normalized node state sequence data.

[0095] S32: Based on the normalized node state sequence data, perform joint feature encoding on multiple node states connected by hyperedges under the same time index to generate hyperedge feature data that characterizes the association relationship of multiple nodes within the hyperedge;

[0096] S33: Arrange the node state sequence data and the corresponding hyperedge feature data in a temporal sequence according to the time index order to construct the node-hyperedge joint temporal feature sequence;

[0097] S34: Perform time-series modeling on the node-hyperedge joint time-series feature sequence, calculate the correlation mapping between node state changes and hyperedge relationship changes under different time indices, and obtain intermediate time-series relationship feature data;

[0098] S35: Based on intermediate temporal relationship feature data, update the node state and hyperedge relationship corresponding to each time index to generate a temporal hypergraph state representation result that reflects the evolution of node state and hyperedge relationship over time.

[0099] In this embodiment, S4 specifically includes:

[0100] S41: Based on the temporal hypergraph state representation results, construct the bait station operation state sequence according to the time index order. The bait station operation state sequence is used to describe the state change process of the bait station under continuous time index.

[0101] S42: Based on the bait station operation state sequence, construct the world model input data structure based on DRAMA model reinforcement learning. The world model input data structure includes the bait station operation state corresponding to the current time index and the bait station operation state corresponding to the historical time index.

[0102] S43: Introduce pheromone controlled release parameters and bait adjustment parameters as state transition conditions in the world model, simulate the evolution path of bait station operation under different parameter values, and obtain multiple sets of candidate state evolution sequences.

[0103] S44: Expand multiple candidate state evolution sequences according to time index, calculate the changes in the operating state of the bait station under multiple future time indices, and form a state evolution prediction sequence;

[0104] S45: Based on the state evolution prediction sequence, a bait station state prediction model is constructed. The bait station state prediction model is used to describe the temporal evolution relationship of the bait station's operating state under the influence of pheromone controlled release parameters and bait adjustment parameters.

[0105] In this embodiment, S5 specifically includes:

[0106] S51: Based on the bait station status prediction model, and based on the bait station status prediction model, extract the bait station operation status prediction results under multiple future time indices to form a status prediction sequence.

[0107] S52: Based on the state prediction sequence, construct a set of reinforcement learning decision states. The set of reinforcement learning decision states is used to characterize the decision input conditions corresponding to the operation state of the bait station under different time indices.

[0108] S53: In the reinforcement learning decision-making process, the pheromone release control parameter and the bait adjustment parameter are defined as optional decision variables, and a set of parameter combinations corresponding to the set of reinforcement learning decision states is generated under each time index;

[0109] S54: Based on the set of decision states and parameter combinations in reinforcement learning, perform strategy evaluation calculations on each set of pheromone release parameters and bait regulation parameters to obtain a sequence of decision results corresponding to different parameter combinations;

[0110] S55: Based on the decision result sequence, screen and determine the pheromone controlled release parameters and the bait adjustment parameters to generate the target pheromone controlled release strategy and the target bait adjustment strategy.

[0111] In this embodiment, S6 specifically includes:

[0112] S61: Based on the target pheromone controlled release strategy, read the pheromone controlled release parameter values ​​corresponding to different time indices in the target pheromone controlled release strategy to form a pheromone release control parameter sequence;

[0113] S62: According to the pheromone release control parameter sequence, the release time, release frequency and release intensity of the pheromone release device inside the bait station are set under the corresponding time index to complete the adjustment of the pheromone release process and obtain the adjusted pheromone release state.

[0114] S63: Based on the target bait adjustment strategy, read the bait adjustment parameter values ​​corresponding to different time indices in the target bait adjustment strategy to form a bait state adjustment parameter sequence;

[0115] S64: According to the bait state adjustment parameter sequence, adjust the physical and chemical state parameters of the bait in the bait station under the corresponding time index to complete the bait state adjustment and obtain the adjusted bait state;

[0116] S65: Combine and update the pheromone release state after the pheromone release process is regulated with the bait state after the bait state is regulated to generate the regulated bait station operation status data.

[0117] In this embodiment, S7 specifically includes:

[0118] S71: After the operation status of the bait station is adjusted, the operation status data of the adjusted bait station is collected according to the preset monitoring cycle. The operation status data includes the adjusted pheromone release status data, the adjusted bait status data, the termite feeding behavior data, and the environmental parameter data.

[0119] S72: Perform time-marking processing on the collected bait station operation status data, associate the operation status data with the corresponding time index, and form operation status time-series data with time index identifier;

[0120] S73: Based on the time-series data of the operating status with time index identifiers, the changes in pheromone release status, bait status, and termite feeding behavior are jointly processed to obtain the feedback operating status sequence;

[0121] S74: Reorganize the feedback running status sequence into a multi-source time series data structure according to the data type, keeping the data structure of the multi-source time series data consistent, and obtain the feedback multi-source time series data;

[0122] S75: Input multi-source time-series data to update the time-series hypergraph data, perform node state update and hyperedge relationship update, and generate updated time-series hypergraph data;

[0123] S76: Based on the updated time-series hypergraph data, the bait station state prediction model is updated to obtain the updated bait station state prediction model.

[0124] This embodiment includes the following modules:

[0125] The data acquisition module is used to periodically collect pheromone release status data, bait status data, termite feeding behavior data, and environmental parameter data during the operation of the termite bait station. It also performs unified time reference alignment processing on the collected data and outputs multi-source time series data.

[0126] The temporal hypergraph construction module is used to identify data types, extract state variables, and generate nodes based on multi-source time-series data. It also constructs hyperedge relationships connecting multiple nodes under the same time index, organizes node states and hyperedge relationships in a time series, and forms temporal hypergraph data corresponding to the bait station's operating status.

[0127] The temporal hypergraph representation learning module is used to perform node state normalization, hyperedge joint feature encoding, node-hyperedge joint temporal feature sequence construction, and temporal modeling processing on temporal hypergraph data, and outputs temporal hypergraph state representation results that reflect the changes of node state and hyperedge relationship over time.

[0128] The bait station state prediction modeling module is used to construct the bait station operation state sequence based on the temporal hypergraph state representation results, and introduce pheromone controlled release parameters and bait adjustment parameters to model the evolution process of the bait station operation state and generate a bait station state prediction model.

[0129] The reinforcement learning decision generation module is used to form a state prediction sequence based on the bait station state prediction model, construct a set of reinforcement learning decision states, evaluate and screen the pheromone release control parameters and bait regulation parameters, and output the target pheromone release control strategy and the target bait regulation strategy.

[0130] The bait station adjustment execution module is used to set parameters for the release time, release frequency and release intensity of the pheromone release device inside the bait station according to the target pheromone controlled release strategy, and to adjust the physical state parameters and chemical state parameters of the bait in the bait station according to the target bait adjustment strategy, and generate adjusted bait station operation status data.

[0131] The operation status feedback and update module is used to continuously monitor the adjusted bait station operation status data and send the monitored data back to the operation data acquisition module and the time series hypergraph construction module to update the time series hypergraph data and the bait station status prediction model, forming a closed-loop operation structure.

[0132] Example 1: To verify the feasibility of this invention in practice, this example selects a continuously operating termite bait station as a practical application scenario to verify the coordinated control process of pheromone release regulation and bait state regulation during long-term operation of the termite bait station. In this scenario, the bait station needs to maintain the continuous feeding behavior of termites over a long period to ensure the stability of the pheromone induction pathway and promote continuous bait consumption, thereby meeting the actual needs of termite behavior regulation in pest control applications.

[0133] In this application scenario, existing technologies typically rely on fixed or semi-fixed pheromone release schemes, or periodically change and adjust the bait state based solely on empirical rules. In actual operation, these methods often struggle to cope with the effects of changes in termite feeding behavior over time, fluctuations in environmental parameters, and changes in pheromone diffusion states. This can easily lead to a mismatch between pheromone induction intensity and bait palatability, resulting in problems such as interrupted termite feeding behavior, decreased feeding frequency, or deviations in feeding paths. The core problem addressed in this embodiment is to introduce a technical solution based on DRAMA model-based reinforcement learning combined with temporal hypergraph representation learning to continuously model, predict, and make decisions regarding the operational state of termite bait stations, thereby maintaining dynamic coordination between pheromone release and bait state adjustment in actual operation.

[0134] In the specific implementation process, during the operation of the termite bait station, data on pheromone release status, bait status, termite feeding behavior, and environmental parameters are collected according to a fixed collection cycle. The pheromone release status data includes three dimensions: release time, release frequency, and release intensity. The bait status data includes physical and chemical state parameters of the bait. The termite feeding behavior data includes the number of feedings, feeding duration, and feeding path distribution. The environmental parameter data includes temperature, humidity, and airflow parameters. All of the above data undergoes unified time-base alignment processing to form multi-source time-series data for subsequent time-series modeling.

[0135] After obtaining multi-source time-series data, data from different sources are mapped to nodes in a time-series hypergraph. Hyperedges are constructed based on the synchronous changes between state variables under the same time index, thus forming time-series hypergraph data capable of simultaneously characterizing multi-variable relationships. Subsequently, time-series hypergraph representation learning is performed on the data. Node states are numerically normalized, and joint feature encoding is performed on the states of multiple nodes connected by hyperedges. The node states and hyperedge features are then arranged according to their time indices to obtain a node-hyperedge joint time-series feature sequence. Based on this, time-series modeling is used to obtain a time-series hypergraph state representation reflecting the evolution of node states and hyperedge relationships over time.

[0136] Building upon this foundation, a world model based on DRAMA-based reinforcement learning is introduced to further model the temporal hypergraph state representation results. Specifically, the operating states of bait stations under continuous time indices are organized into a sequence of operating states. The operating states under the current and historical time indices are used as inputs to the world model, while pheromone release parameters and bait adjustment parameters are used as state transition conditions. The evolution paths of the operating states of bait stations under different parameter combinations are simulated in the world model. By expanding multiple sets of candidate state evolution sequences, predicted state evolution sequences under multiple future time indices are obtained, and a bait station state prediction model is constructed based on these sequences.

[0137] Subsequently, a state prediction sequence is generated based on the bait station state prediction model, and a reinforcement learning decision state set is constructed. The pheromone controlled release parameters and bait adjustment parameters are defined as optional decision variables, and a corresponding parameter combination set is generated under each time index. By performing strategy evaluation calculations on different parameter combinations, a decision result sequence is obtained, and target pheromone controlled release strategies and target bait adjustment strategies are selected and generated accordingly. During the execution phase, parameters such as release time, release frequency, and release intensity of the pheromone release device inside the bait station are set according to the target pheromone controlled release strategy. Simultaneously, the physical and chemical state parameters of the bait are adjusted according to the target bait adjustment strategy, thereby generating adjusted bait station operating state data. The adjusted operating state data is continuously monitored and fed back to update the time-series hypergraph data and the bait station state prediction model, forming a closed-loop operation process.

[0138] In this embodiment, the main parameters of the DRAMA model-based reinforcement learning world model are set to a state embedding dimension of 128 and a time window length of 12. During policy evaluation, 100 parameter combinations are evaluated per round, and the policy is stably converged after 50 consecutive iterations. During the temporal hypergraph representation learning process, node state normalization uses min-max normalization, the hyperedge joint feature encoding dimension is set to 64, and the temporal modeling window length is set to 10. During model training and updates, a model parameter update is triggered after each complete running cycle.

[0139] To verify the actual improvement brought about by the present invention, the changes in key indicators of the bait station operation status before and after adjustment were compared and analyzed. The stability of pheromone release, consistency of bait status, number of termite feedings, duration of termite feeding, and coverage of feeding path were selected as evaluation indicators. The measured values ​​and model predictions were compared within five consecutive operating cycles. The results are shown in Table 1.

[0140] Table 1. Comparison of Measured and Predicted Values ​​of Key Indicators for Baiting Station Operation Status

[0141] Runtime Measured values ​​of pheromone release stability Predicted values ​​of pheromone release stability Measured number of times termites fed. Predicted number of termite feeding events Actual measured value of feeding path coverage Period 1 0.82 0.80 145 142 0.76 Period 2 0.84 0.83 152 150 0.78 Period 3 0.86 0.85 158 156 0.81 Period 4 0.87 0.86 162 160 0.83 Period 5 0.89 0.88 168 166 0.85

[0142] As shown in Table 1, the model predictions and measured values ​​remained relatively close throughout the continuous operating period. The bait station's operational status exhibited a continuous upward trend in key indicators such as pheromone release stability, termite feeding frequency, and foraging path coverage. This demonstrates that the present invention, through a combination of temporal hypergraph representation learning and reinforcement learning decision-making, effectively models and dynamically adjusts the termite bait station's operational status. Compared to the operation without this technology, the interruption of feeding behavior was significantly reduced, and the bait station's operational status remained stable throughout the continuous period, verifying the feasibility and technical advantages of the present invention in practical applications.

Claims

1. A dynamic optimization method for bait palatability based on termite pheromone release patterns, characterized in that, Includes the following steps: S1: During the operation of the termite bait station, the operation data of the bait station is collected periodically. The operation data includes pheromone release status data, bait status data, termite feeding behavior data and environmental parameter data. The operation data is then processed for time synchronization to obtain multi-source time-series data. S2: Construct a time-series hypergraph structure based on multi-source time-series data, map pheromone release state, bait state, termite feeding behavior and environmental parameters as hypergraph nodes, and construct hyperedges for the association relationships between multiple nodes to form time-series hypergraph data corresponding to the bait station operation status; S3: Perform temporal hypergraph representation learning on temporal hypergraph data to obtain state representation results that reflect the changes in node state and hyperedge relationship over time; S4: Using the state representation results as input, construct a world model based on DRAMA model-based reinforcement learning, model the evolution process of bait station state under the action of pheromone release parameters and bait regulation parameters, and obtain a bait station state prediction model. S5: Based on the bait station state prediction model, perform reinforcement learning decision calculations to generate pheromone controlled release strategies and bait adjustment strategies; S6: Adjust the pheromone release process and bait state of the bait station according to the pheromone controlled release strategy and bait adjustment strategy to obtain the adjusted bait station operation state; S7: Monitor the adjusted operating status of the bait station and use the monitored data to update the time-series hypergraph data and the bait station status prediction model, forming a dynamic optimization process for bait palatability driven by pheromone controlled release.

2. The method for dynamic optimization of bait palatability based on termite pheromone release patterns according to claim 1, characterized in that, S1 specifically includes: S11: During the operation of the termite bait station, pheromone release status data is acquired through collection devices set inside and around the bait station. The pheromone release status data is used to characterize the release time, release frequency and release intensity of pheromones within a preset collection cycle. S12: During the collection period, acquire bait state data. The bait state data is used to characterize the physical and chemical state of the bait in the bait station and form bait state time series data at a preset sampling interval. S13: During the collection period, acquire termite feeding behavior data. The termite feeding behavior data is used to characterize the number of times termites feed, the duration of feeding, and the distribution of feeding paths in the bait station, and form termite feeding behavior time series data in chronological order. S14: During the collection period, acquire environmental parameter data. The environmental parameter data is used to characterize the temperature, humidity and air flow parameters in the environment where the bait station is located, and form environmental parameter time series data. S15: Perform unified time reference alignment processing on pheromone release status data, bait status data, termite feeding behavior data, and environmental parameter data. The time reference alignment processing uses a preset timestamp as the reference and resamples data from different collection frequencies to the same time scale to obtain multi-source time-series data after time synchronization.

3. The method for dynamic optimization of bait palatability based on termite pheromone release patterns according to claim 1, characterized in that, S2 specifically includes: S21: Based on multi-source time-series data, pheromone release status data, bait status data, termite feeding behavior data, and environmental parameter data are classified and labeled according to the data source type to form a multi-source time-series data set with data type labels. S22: Extract the state variables corresponding to pheromone release state data, bait state data, termite feeding behavior data, and environmental parameter data from the multi-source time series data set, and define them as node sets in the time series hypergraph to obtain the node initialization results. S23: Under the same time index, based on the synchronous change relationship between different state variables in multi-source time series data, construct a set of association relationships connecting multiple nodes. Each set of association relationships corresponds to a hyperedge, thus obtaining a set of hyperedges. S24: Organize the node set and the hyperedge set into a time series, arrange the node states and hyperedge connections under different time indices in chronological order, and form the initial structure of the temporal hypergraph with time evolution characteristics; S25: Perform consistency verification on the node state data and hyperedge connection relationships in the initial structure of the temporal hypergraph, remove node states and hyperedge connection relationships with missing time indexes or incomplete data, and obtain the verified temporal hypergraph structure. S26: Output the verified time-series hypergraph structure as the time-series hypergraph data corresponding to the bait station's operating status.

4. The method for dynamic optimization of bait palatability based on termite pheromone release patterns according to claim 1, characterized in that, S3 specifically includes: S31: Based on the time-series hypergraph data, the node state data corresponding to each time index in the time-series hypergraph data is numerically normalized to obtain the normalized node state sequence data. S32: Based on the normalized node state sequence data, perform joint feature encoding on multiple node states connected by hyperedges under the same time index to generate hyperedge feature data that characterizes the association relationship of multiple nodes within the hyperedge; S33: Arrange the node state sequence data and the corresponding hyperedge feature data in a temporal sequence according to the time index order to construct the node-hyperedge joint temporal feature sequence; S34: Perform time-series modeling on the node-hyperedge joint time-series feature sequence, calculate the correlation mapping between node state changes and hyperedge relationship changes under different time indices, and obtain intermediate time-series relationship feature data; S35: Based on intermediate temporal relationship feature data, update the node state and hyperedge relationship corresponding to each time index to generate a temporal hypergraph state representation result that reflects the evolution of node state and hyperedge relationship over time.

5. The method for dynamic optimization of bait palatability based on termite pheromone release patterns according to claim 1, characterized in that, S4 specifically includes: S41: Based on the temporal hypergraph state representation results, construct the bait station operation state sequence according to the time index order. The bait station operation state sequence is used to describe the state change process of the bait station under continuous time index. S42: Based on the bait station operation state sequence, construct the world model input data structure based on DRAMA model reinforcement learning. The world model input data structure includes the bait station operation state corresponding to the current time index and the bait station operation state corresponding to the historical time index. S43: Introduce pheromone controlled release parameters and bait adjustment parameters as state transition conditions in the world model, simulate the evolution path of bait station operation under different parameter values, and obtain multiple sets of candidate state evolution sequences. S44: Expand multiple candidate state evolution sequences according to time index, calculate the changes in the operating state of the bait station under multiple future time indices, and form a state evolution prediction sequence; S45: Based on the state evolution prediction sequence, a bait station state prediction model is constructed. The bait station state prediction model is used to describe the temporal evolution relationship of the bait station's operating state under the influence of pheromone controlled release parameters and bait adjustment parameters.

6. The method for dynamic optimization of bait palatability based on termite pheromone release patterns according to claim 1, characterized in that, S5 specifically includes: S51: Based on the bait station status prediction model, and based on the bait station status prediction model, extract the bait station operation status prediction results under multiple future time indices to form a status prediction sequence. S52: Based on the state prediction sequence, construct a set of reinforcement learning decision states. The set of reinforcement learning decision states is used to characterize the decision input conditions corresponding to the operation state of the bait station under different time indices. S53: In the reinforcement learning decision-making process, the pheromone release control parameter and the bait adjustment parameter are defined as optional decision variables, and a set of parameter combinations corresponding to the set of reinforcement learning decision states is generated under each time index; S54: Based on the set of decision states and parameter combinations in reinforcement learning, perform strategy evaluation calculations on each set of pheromone release parameters and bait regulation parameters to obtain a sequence of decision results corresponding to different parameter combinations; S55: Based on the decision result sequence, screen and determine the pheromone controlled release parameters and the bait adjustment parameters to generate the target pheromone controlled release strategy and the target bait adjustment strategy.

7. The method for dynamic optimization of bait palatability based on termite pheromone release patterns according to claim 1, characterized in that, S6 specifically includes: S61: Based on the target pheromone controlled release strategy, read the pheromone controlled release parameter values ​​corresponding to different time indices in the target pheromone controlled release strategy to form a pheromone release control parameter sequence; S62: According to the pheromone release control parameter sequence, the release time, release frequency and release intensity of the pheromone release device inside the bait station are set under the corresponding time index to complete the adjustment of the pheromone release process and obtain the adjusted pheromone release state. S63: Based on the target bait adjustment strategy, read the bait adjustment parameter values ​​corresponding to different time indices in the target bait adjustment strategy to form a bait state adjustment parameter sequence; S64: According to the bait state adjustment parameter sequence, adjust the physical and chemical state parameters of the bait in the bait station under the corresponding time index to complete the bait state adjustment and obtain the adjusted bait state; S65: Combine and update the pheromone release state after the pheromone release process is regulated with the bait state after the bait state is regulated to generate the regulated bait station operation status data.

8. The method for dynamic optimization of bait palatability based on termite pheromone release patterns according to claim 1, characterized in that, Specifically, S7 includes: S71: After the operation status of the bait station is adjusted, the operation status data of the adjusted bait station is collected according to the preset monitoring cycle. The operation status data includes the adjusted pheromone release status data, the adjusted bait status data, the termite feeding behavior data, and the environmental parameter data. S72: Perform time-marking processing on the collected bait station operation status data, associate the operation status data with the corresponding time index, and form operation status time-series data with time index identifier; S73: Based on the time-series data of the operating status with time index identifiers, the changes in pheromone release status, bait status, and termite feeding behavior are jointly processed to obtain the feedback operating status sequence; S74: Reorganize the feedback running status sequence into a multi-source time series data structure according to the data type, keeping the data structure of the multi-source time series data consistent, and obtain the feedback multi-source time series data; S75: Input multi-source time-series data to update the time-series hypergraph data, perform node state update and hyperedge relationship update, and generate updated time-series hypergraph data; S76: Based on the updated time-series hypergraph data, the bait station state prediction model is updated to obtain the updated bait station state prediction model.

9. A dynamic optimization system for bait palatability based on termite pheromone release patterns, characterized in that, Includes the following modules: The data acquisition module is used to periodically collect pheromone release status data, bait status data, termite feeding behavior data, and environmental parameter data during the operation of the termite bait station. It also performs unified time reference alignment processing on the collected data and outputs multi-source time series data. The temporal hypergraph construction module is used to identify data types, extract state variables, and generate nodes based on multi-source time-series data. It also constructs hyperedge relationships connecting multiple nodes under the same time index, organizes node states and hyperedge relationships in a time series, and forms temporal hypergraph data corresponding to the bait station's operating status. The temporal hypergraph representation learning module is used to perform node state normalization, hyperedge joint feature encoding, node-hyperedge joint temporal feature sequence construction, and temporal modeling processing on temporal hypergraph data, and outputs temporal hypergraph state representation results that reflect the changes of node state and hyperedge relationship over time. The bait station state prediction modeling module is used to construct the bait station operation state sequence based on the temporal hypergraph state representation results, and introduce pheromone controlled release parameters and bait adjustment parameters to model the evolution process of the bait station operation state and generate a bait station state prediction model. The reinforcement learning decision generation module is used to form a state prediction sequence based on the bait station state prediction model, construct a set of reinforcement learning decision states, evaluate and screen the pheromone release control parameters and bait regulation parameters, and output the target pheromone release control strategy and the target bait regulation strategy. The bait station adjustment execution module is used to set parameters for the release time, release frequency and release intensity of the pheromone release device inside the bait station according to the target pheromone controlled release strategy, and to adjust the physical state parameters and chemical state parameters of the bait in the bait station according to the target bait adjustment strategy, and generate adjusted bait station operation status data. The operation status feedback and update module is used to continuously monitor the adjusted bait station operation status data and send the monitored data back to the operation data acquisition module and the time series hypergraph construction module to update the time series hypergraph data and the bait station status prediction model, forming a closed-loop operation structure.

Images