A rule representation and reasoning method with preferences for task planning

Abstract

This application discloses a method for representing and reasoning rules with preferences for task planning, including the following steps: (1) classifying and representing task-level rules from three aspects: rule preconditions, rule preferences, and rule post-commands; (2) implementing a rule conflict detection, interpretation, and resolution method under a minimum correction strategy from the perspective of logical unsatisfiability types for possible conflicts between rules; and (3) implementing multi-rule deduction in a typical task scenario based on a balanced strategy of multi-rule collaborative reasoning and a task deduction platform. The method of the embodiment of the present invention solves the problem of representing and reasoning rules with preferences in task planning for the first time, providing a new and practical method for this application scenario.

Description

Technical Field

[0001] The present invention relates to a task planning-oriented rule representation and reasoning method with preference, belonging to the technical field of artificial intelligence basic knowledge representation and reasoning. Background Art

[0002] Currently, in the field of basic artificial intelligence research centered on knowledge representation and reasoning, the main rule representation technologies are divided into symbolic rule representation and reasoning represented by logic programs, production systems, and semantic web languages; rule representation and reasoning methods based on numerical metric spaces represented by neural networks; and rule representation and reasoning methods based on probabilistic graphical models represented by Bayesian networks and Markov logic networks. Furthermore, there are also related rule induction techniques that efficiently and effectively obtain artificial prior knowledge. However, these representation technologies have difficulty directly representing task planning rules with preferences (hereinafter referred to as task-level rules). Such rules have the characteristic of nested rule consequences. Current major rule representation methods such as production rules and logic programs cannot intuitively express such rules, making it difficult to use their related knowledge engineering skeleton tools for efficient rule entry.

[0003] At the same time, existing mature knowledge processing technologies and knowledge engineering skeleton tools usually define conflicting rules as rules that can simultaneously satisfy execution conditions. However, rule conflicts based on the mutual exclusion and contradiction of rule consequents are situations that are not handled in the current major knowledge representation and reasoning technologies and knowledge engineering skeleton tools. As a result, it is difficult to identify and resolve a large number of consequent-conflicting rules in task-level rules using existing methods.

[0004] During the execution of multiple task planning rules, resources such as work units need to be reasonably sorted and screened. The current major knowledge representation and reasoning methods, such as production rules and logic programs, do not provide corresponding processing methods for the priority relationship of resources, resulting in the existing mature rule processing technology and knowledge engineering skeleton tools being difficult to use for the execution of such task planning rules.

[0005] To implement a rule representation and reasoning method with preferences for task planning, the following three key issues need to be addressed:

[0006] (1) How to establish a representation of preference rules for task planning;

[0007] (2) How to identify and resolve potential conflicts between rules;

[0008] (3) How to perform multi-rule collaborative task planning reasoning in task planning scenarios. Summary of the Invention

[0009] The present invention aims to solve the problem of preference representation and reasoning methods for task planning-oriented rules. To this end, the purpose of the present invention is to provide a method for representation and reasoning of preference rules for task planning-oriented rules. The method can represent preference rules for task planning in combination with the preconditions, conclusions, preferences, logical operators and other contents of specific rules; by analyzing the representation of the rules, identify and resolve conflicts between different rules; and finally, complete collaborative reasoning of multiple rules based on a balance strategy in a typical task planning scenario.

[0010] To achieve the above objectives, an embodiment of the present invention proposes a preference-based rule representation and reasoning method for task planning, comprising the following steps: Step S1, constructing a task-level rule representation, normalizing the task-level rule information input by the user, establishing a unified representation and storing it, constructing an ontology of task-level rules, classifying the task-level rules from three aspects: rule preconditions, rule preferences, and rule post-commands, and forming a representation method for task-level rules;

[0011] Step S2, conflict detection and resolution: Different rules may make different decisions when faced with the same workspace state, thus causing conflicts. This step examines possible conflicts between different rules and proposes resolution solutions. It determines whether there are conflicts between rules from three perspectives: logical unsatisfiability, mutual exclusivity of actions between different rules, and whether the workspace state changes caused by rule execution, which in turn causes conflicts with other rules. It also provides resolution methods and explanations.

[0012] Step S3, multi-rule collaborative reasoning, issues instructions to the local work unit that align with the user's intent based on the multiple task-level rules entered by the user. A balanced strategy for multi-rule collaborative reasoning is constructed, along with a prototype and evaluation method for the multi-rule collaborative reasoning engine. Based on the task reasoning platform, multi-rule collaborative reasoning is implemented in typical task planning scenarios. Workspace status information is read in real time, and based on a balanced reasoning model, the triggered rules are determined, thereby instructing the work unit to execute the relevant commands.

[0013] Among them, in step S1, the rule classification is formed from the three aspects of rule premise conditions, rule preferences and rule post-commands from the aspects of preconditions, conclusions, preferences, logical operators, workspace status, work unit types, association relationships, and reasoning task types, and the ontology of task-level rules including rule patterns, precondition entities, conclusion entities, preference entities, mutually exclusive tasks, etc. is constructed.

[0014] In step S2, three types of rule mutual exclusion are constructed from logical unsatisfiability types and resources: preference contradiction, action mutual exclusion, and action contradiction. Based on this, a conflict detection algorithm is designed. By checking and comparing preconditions, rule preferences, and postconditions, it infers the conflict type and generates a resolution plan. This algorithm implements rule conflict determination during rule editing and entry, and provides conflict explanations through interface prompts.

[0015] In step S2, conflict resolution is divided into two stages: static conflict resolution and runtime conflict resolution. The former is designed as a human-machine combination of machine prompts and manual resolution based on conflict detection during the rule editing and entry process. The latter is achieved during the rule running stage by designing a strategy including conflict rule set generation and conflict rule selection in the inference engine based on the actual rule content and test results.

[0016] In step S3, the preference-based equilibrium model constructed by considering the priority preference relationship in the work unit selection also includes: (1) a single-rule preference processing model, which takes the optimal preference processing strategy on the complete order; (2) a multi-rule preference processing model, which adopts the three classical social choice theory methods of Pareto optimality, worst optimality, and best optimality to form the optimal conclusion selection strategy under the preference conclusion.

[0017] According to an embodiment of the present invention, a method for representing and reasoning rules with preferences for task planning breaks down input rules into three parts: preconditions, rule preferences, and post-commands. These components include various attributes such as the workspace, command type, and target party. This method generates a rule representation and stores it in a JSON-formatted file. The preconditions, preferences, and post-commands of multiple preference-based rules are then used to determine whether any conflicts exist, such as logical unsatisfiability, mutually exclusive actions between different rules, or a change in the workspace state caused by rule execution, leading to conflicts with other rules. If so, an explanation and correction plan are provided. Otherwise, the multiple rules are input into an inference engine for multi-rule collaborative deduction. The inference engine reads the workspace state in real time to determine whether it satisfies any rule's preconditions. When a rule's preconditions match the current workspace state, the rule is considered triggered. If a rule is triggered, candidate rules are ranked according to social choice theory. For the rule with the highest priority, the engine infers the execution unit according to its preference, and interacts with the deduction platform in the form of Lua scripts according to the post-commands of the rule to issue commands to these execution units. This process is repeated until the end of the deduction, and the multi-rule collaborative deduction can be completed. The conflict detection and resolution between rules involved in this method is to ensure the correctness of the rule combination input by the rule reasoning engine, that is, contradictory rules cannot be applied to the same deduction; the social choice theory method involved is to ensure the consistency of the operation, that is, at a certain moment, when a unit is commanded by multiple rules, it is restricted to being commanded only by the rule with the highest priority.

[0018] Compared with existing technologies, this solution has the following advantages:

[0019] 1. This solution proposes a method for modeling preferences that can uniformly represent a preference rule that includes attribute preferences, value preferences, and attribute preferences and value preferences with constraints. Based on this method, it is possible to achieve a unified representation of rules with preferences.

[0020] 2. This solution addresses potential conflicts between rules during task planning by proposing a conflict detection and resolution algorithm for task planning rules with preferences, based on the concept of logical unsatisfiability. By defining relationships between preferences, preconditions, and postconditions across different rules, this algorithm enables more precise conflict detection, provides explanations for conflicts, and proposes resolution methods.

[0021] 3. This solution can automatically and periodically request the task deduction system to update the deduction progress in the form of Lua scripts through the interactive module, obtain the real-time workspace status, and process and analyze it to achieve real-time rule reasoning and instruction issuance. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0023] Figure 1 A schematic diagram of the overall process of a rule representation and reasoning method with preferences for task planning according to an embodiment of the present invention;

[0024] Figure 2 Schematic diagram of a rule representation construction method in Example 1 of the present invention;

[0025] Figure 3 Schematic diagram of a conflict detection and resolution method in Example 1 of the present invention;

[0026] Figure 4 This is the logic for determining the type of conflict that may exist between two rules in Example 1 of the present invention;

[0027] Figure 5 Schematic diagram of a multi-rule collaborative deduction method in Example 1 of the present invention;

[0028] Figure 6 This is a process diagram of the workspace status perception step in Example 1 of the present invention;

[0029] Figure 7 This is a process diagram of the automatic reasoning step in Example 1 of the present invention. DETAILED DESCRIPTION

[0030] The following describes embodiments of the present invention in detail. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended only to explain the present invention and are not to be construed as limiting the present invention.

[0031] Example 1: The following describes a preference-based rule representation and reasoning method for task planning according to an embodiment of the present invention in conjunction with the accompanying drawings. Figure 1 FIG. 1 is a flowchart of a task planning-oriented rule representation and reasoning method with preference according to an embodiment of the present invention. Figure 1 As shown, a rule representation and reasoning method with preference for task planning according to an embodiment of the present invention includes the following steps:

[0032] Step S1, “constructing task-level rule representation”, normalizes the task-level rule information input by the user, establishes a unified representation and stores it. The specific process is as follows: Figure 2 As shown, the following steps are included:

[0033] Step S11, "Obtaining Rule Precondition Information," describes a set of workspace conditions by constraining the workspace state. The rule is triggered only when the current workspace meets these conditions. The rule representation technology designed in this invention categorizes preconditions into seven subconditions: "Workspace," "Work Unit Number Constraint," "Target Type," "Target Owner," "Target Constraint," "Current Target State," and "Rule Effective Time Period." Obtaining information about these seven subconditions is equivalent to obtaining the rule's prerequisite information.

[0034] Step S12, "obtain rule preference information". Rule preference is to constrain the attributes and characteristics of the work unit of the party, and describe the characteristics of the specific execution unit that needs to be selected when the rule is triggered and the priority relationship between different units. By induction, the rule representation technology designed by the present invention regards rule preference information as consisting of four parts, namely "attribute preference", "attribute value preference", "attribute preference with constraints" and "attribute value preference with constraints" (for the definition of the four preferences, see Appendix A.). The system guides the user to input and obtain information of these four components, which is equivalent to obtaining the preference information of the rule.

[0035] Step S13, "Acquire Rule Post-Command Information," specifies the execution portion of a rule. When a rule is triggered and the priority of a candidate unit is determined, the post-command determines the number of candidate units to be selected and the operations to be performed on the selected units. In summary, the rule representation technology designed in this invention divides commands into two sub-conditions: "action type" and "number of units." Acquiring information about these two sub-conditions is equivalent to obtaining the post-command information for the rule.

[0036] In step S14, "Generate JSON file to store rules," the rule representation technology designed in this invention incorporates a JSON-based rule representation specification. By using key-value pairs to represent rule preconditions and postconditions, and using a graph data structure to represent rule preference information, each rule is mapped to a JSON-formatted file for storage.

[0037] Step S2, “Conflict detection and resolution”, in the face of the same workspace state, different rules may make different decisions, thus causing conflicts. This step checks the possible conflicts between different rules and proposes resolution solutions. In one embodiment of the present invention, this step is as follows: Figure 3 As shown:

[0038] Step S21, "Precondition Similarity Verification," determines whether the preconditions of two rules are identical or similar. For any two rules, if the seven attributes in their preconditions overlap and the remaining six attributes are identical, the preconditions are considered identical. If the rule time periods overlap but the "Work Unit Number Constraint" differs and the remaining four attributes are identical, the preconditions are considered similar. Otherwise, the preconditions are considered different.

[0039] Step S22, "Rule Preference Inclusion Determination," determines whether the preferences of two rules are included. For any two rules, if their preferences can select the same work unit, it is determined that the rule preference is included; otherwise, it is determined that the rule preference is not included.

[0040] Step S23, "Post-command type determination," determines the relationship between the post-commands of two rules. For any two rules, if their post-command actions are the same, the post-commands are considered identical. If their post-commands are "depart" and "return," the post-commands are considered mutually exclusive. If their post-commands are "additional dispatch" and "return," the post-commands are considered contradictory. Otherwise, the post-commands are considered unrelated.

[0041] Step S24, "determine the conflict type", this step is based on the judgment of S21-S23, to determine the possible conflict type between the two rules. The judgment logic is as follows: Figure 4 shown.

[0042] Step S25, "Generate a Resolution Plan," builds on the information from S24. When two rules conflict, the user is informed of the cause and location of the conflict in plain, natural language, and is prompted with optional modifications to facilitate conflict resolution. If conflicting rules still exist in S3, the conflict is automatically resolved based on the user-defined rule priority.

[0043] Step S3, "multi-rule collaborative deduction", is a step that issues instructions that meet the user's intentions to the local work unit based on multiple task-level rules input by the user. In one embodiment of the present invention, this step is as follows: Figure 5 As shown, the specific implementation process includes:

[0044] Step S31, “obtaining real-time information”, automatically and periodically requests the mission simulation system to update the simulation progress and obtain the real-time workspace status.

[0045] Step S32, "work area status perception", obtains the information required for reasoning by further analyzing the real-time work area information. The information required for reasoning is divided into two types, namely explicit information and implicit information. Explicit information can be obtained directly from the work area information, including but not limited to: information about the work unit of this party, such as the work unit's own attributes, geographic location, speed, direction, mounting, current status, task status, target, etc.; information about other work units detected, such as other work units' own attributes, geographic location, speed, direction, etc.; information about non-work units of this party (such as fixed buildings, etc.), such as geographic location, carried work units, current status, etc.; work area deduction information, such as deduction time, deduction speed, deduction score, etc. Implicit information needs to be based on explicit information and combined with common sense or pre-given knowledge to further analyze the work area status to obtain, including but not limited to: geographic information, such as information about this party and other work units in a key area; range information, such as the detection range and range of action of this party's work unit; distance information, such as the distance between two specific work units, etc. The workflow of this step is as follows: Figure 6 shown.

[0046] Step S33, “Automatic Reasoning”, determines whether a rule is triggered based on the information extracted by the workspace status perception module, and determines the command issued and the work unit that executes the command. Figure 7 As shown, the specific implementation process includes:

[0047] Step S331, "precondition determination", checks the workspace status based on the precondition. If the current workspace status meets every sub-condition in the precondition of a rule, it is considered that the current workspace meets the precondition and the rule is triggered.

[0048] Step S332, “Rule selection”, when the prerequisites of multiple rules are met, three classical social selection theory methods, namely Pareto optimality, worst optimality, and best optimality, are used to screen out the most prioritized rule and only execute this rule.

[0049] Step S333, “preference screening”, performs a depth-first search on the graph data structure representing the preferences generated in step S14, filters out work units that do not meet the conditions, and obtains a priority ranking of the remaining work units.

[0050] Step S334, "obtain post-commands", determines the set of work units that execute the task and the specific commands issued to these units according to the current rules in the current workspace state.

[0051] Step S34, "issuing instructions", converts the command obtained in step S334 into a Lua script, interacts with the deduction platform, and completes the instruction issuance.

[0052] Appendix A. Components of Preference Information

[0053] A rule's preference consists of four parts: "attribute preference," "attribute value preference," "attribute preference with constraints," and "attribute value preference with constraints." This appendix explains the meaning and form of these four preferences.

[0054] 1. Attribute Preference

[0055] Attribute preferences specify and rank certain attributes, specifying the priority of consideration. For example, if attribute Att1 is prioritized over attribute Att2, then when sorting multiple work units, the values ​​of these work units for attribute Att1 will be considered first, followed by the values ​​of attribute Att2. In this invention, an attribute preference is represented by a sequence of attributes. For example, the attribute preference in this example can be represented by (Att1, Att2).

[0056] 2. Attribute value preference

[0057] Attribute value preference specifies certain values ​​of an attribute and sorts them, specifying the priority of consideration. For example, for attribute Att1, its value Val is specified. 11 Takes precedence over Val 12 , when sorting multiple work units, when considering the attribute Att1, the attribute Att1 value is Val 11 takes precedence over the attribute Att1 value of Val 12 In the present invention, an attribute value preference is represented by an attribute and a sequence consisting of the attribute's value. For example, the attribute value preference in this example can be represented by (Att1, (Val 11 , Val 12 ))express.

[0058] 3. Attribute Preference with Constraints

[0059] Attribute preference with constraints specifies an attribute value preference or an attribute value preference with constraints P V A value of the specified attribute, and specify some attributes and sort them, specifying the priority to be considered when obtaining the value. For example, the value of the attribute Att1 is specified as Val. 11 , and then specify that the attribute Att2 takes precedence over the attribute Att3, then when sorting multiple work units, if there are multiple work units with the attribute Att1 value of Val 11 , then the values ​​of these units on the attribute Att2 are considered first, and then the values ​​on the attribute Att3 are considered. In the present invention, a constrained attribute preference is expressed through its parent attribute value preference P V、P V It is represented by an attribute A, a value V that A can take, and a sequence of attributes. For example, the attribute preference with constraints in this example can be represented by (P V , Att1, Val 11 , (Att2, Att3)).

[0060] 4. Attribute value preferences with constraints

[0061] Constrained attribute value preference specifies certain values ​​of an attribute in a constrained attribute preference and sorts them, specifying the priority of consideration. For example, if a constrained attribute preference P is specified A For the attribute Att1 considered, its value Val is specified 11 Takes precedence over Val 12 , then when using P A When sorting multiple work units, when considering the attribute Att1, the value of the attribute Att1 is Val 11 takes precedence over the attribute Att1 value of Val 12 In the present invention, a constrained attribute value preference is expressed through a constrained attribute preference P A 、P A It is represented by a sequence of an attribute A and the value of attribute A. For example, the attribute value preference in this example can be expressed as (P A ,Att1,(Val 11 , Val 12 )) to express it.

[0062] Appendix B. Description of rule patterns and their representation specifications

[0063] Task-level rules are composed of three types of entities: precondition entities, preference entities, and conclusion entities. This appendix first explains these three types of entities.

[0064] 1. Precondition Entity

[0065] 1.1. Work Area: string type, multiple sub-areas divided by a large work area;

[0066] 1.2. Worker number constraint: This constraint consists of two integer values ​​and specifies the range of work units that can be mobilized within the sub-region.

[0067] 1.3. Target type: string type, the type of the task target;

[0068] 1.4. Target party: string type, party of the task target;

[0069] 1.5. Target constraint: This consists of two floating-point values ​​that define the ratio of the number of task targets to the number of work units that can be mobilized within the sub-region.

[0070] 1.6. Current status of target: the current status of the task target;

[0071] 1.7. Rule validity period: This period consists of a start time and an end time. The rule only applies to this period.

[0072] 2. Conclusion Entity

[0073] 2.1. Action: string type, the action performed by the work unit;

[0074] 2.2. Quantity: integer type, the number of work units required to execute the task;

[0075] 3. Preference Entity

[0076] 3.1.Unit type: string type, the selected work unit type;

[0077] 3.2.Unit function: string type, selected work unit function;

[0078] 3.3.Unit model: string type, selected work unit model;

[0079] 3.4. Unit Task Status: The current task status of the selected work unit, that is, whether the task is currently being executed and the name of the task;

[0080] 3.5.Unit-to-target spacing constraint: string type, preference for the spacing between the selected work unit and the task target;

[0081] 3.6. Unit mount: string type, the content currently mounted by the selected work unit.

[0082] The following table gives the representation specifications of task-level rules:

[0083] project meaning area work area num_constraint Constraints on the number of work units target_type Target Type target_group Target party target_constraint Target Constraint target_state Current status of target start_time Start time end_time End Time action_type action unit_type Unit Type unit_function Unit Function unit_model Unit Model unit_mission_state Unit mission status unit_dist_constraint Unit-to-target distance constraint unit_mount Unit mounts

[0084] Example: Now given a specific task-level rule and its corresponding JSON representation:

[0085] Rules: When there is a damaged bridge in area A within the work zone, and there are fewer than 5 work units in area A, a specific work unit must be dispatched to area A between t1 and t2 to perform bridge repair tasks until there are at least 5 work units in the area. The work unit preferences are described as follows:

[0086] 1. Select the unit type according to the following priority: ① Van; ② Truck

[0087] 2. Select unit mounts according to the following priority: ① Mounted with construction materials; ② No construction materials mounted;

[0088] 3. Select the unit task status according to the following priority: ① Currently not executing tasks; ② Currently executing tasks with lower priority;

[0089] 4. If you choose Truck, select the unit function according to the following priority: ① Heavy Truck; ② Medium Truck;

[0090] 5. If you select Truck, select the unit function according to the following priority: ① Heavy Truck; ② Light Truck

[0091] 6. If you choose a heavy truck, select the unit model according to the following priority: ① Ural 6320; ② FAW Jiefang J7;

[0092] 7. If you choose a medium truck, select the unit model according to the following priority: ① FAW Jiefang J6L; ② Dongfeng Kaiput K6L

[0093] 8. Distance constraint between units and targets: proximity principle.

[0094] A set of attribute preferences with a clear order and no value constraints and their value preferences is considered a preference set. The corresponding expression of the preference description in this example is as follows:

[0095] Preference Set 0:

[0096] Unit Type Preference:

[0097] Value 1: Truck → Preference Set 1

[0098] Value 2: Truck → Preference Set 2

[0099] Unit loadout preferences:

[0100] Value 1: Construction materials are mounted

[0101] Value 2: No construction materials are mounted

[0102] Unit mission status preferences:

[0103] Value 1: The task is not executed.

[0104] Value 2: Execute lower priority tasks

[0105] Unit-target distance constraint preference: proximity principle

[0106] Preference Set 1:

[0107] Unit function preference:

[0108] Value 1: Heavy truck

[0109] Value 2: Medium truck

[0110] Preference Set 2:

[0111] Unit function preference:

[0112] Value 1: Heavy Truck → Preference Set 3

[0113] Value 2: Medium Truck → Preference Set 4

[0114] Preference Set 3:

[0115] Unit Model Preference:

[0116] Value 1: Ural 6320

[0117] Value 2: FAW Jiefang J7

[0118] Preference Set 4:

[0119] Unit Model Preference:

[0120] Value 1: FAW Jiefang J6L

[0121] Value 2: Dongfeng K6L

[0122] The corresponding JSON representation of this example is as follows:

[0123]

[0124] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in a suitable manner in any one or more embodiments or examples. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are mutually inconsistent. Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations of the present invention. Those skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A rule representation and reasoning method with preference for task planning, characterized by: The method comprises the following steps: Step S1, task-level rule representation construction, normalizes the task-level rule information input by the user, establishes a unified representation and stores it, as follows: Step S11, "Acquire rule prerequisite information." Preconditions describe a type of workspace condition by constraining the workspace status. The rule is triggered only when the current workspace meets this condition. Rule representation technology categorizes preconditions into seven subconditions: "Workspace," "Number of work units constraint," "Target type," "Target party," "Target constraint," "Current target status," and "Rule effective time period." Acquiring information about these seven subconditions is equivalent to acquiring the rule prerequisite information. Step S12, "Acquire rule preference information." Rule preferences constrain the attributes and characteristics of local work units, describing the characteristics of the specific execution units to be selected when the rule is triggered, as well as the priority relationships between different units. Rule representation technology, through induction, considers rule preference information to consist of four parts: "attribute preference," "attribute value preference," "attribute preference with constraints," and "attribute value preference with constraints." The system guides the user to input and acquire information about these four components, effectively acquiring the rule preference information. Step S13, "Acquire rule post-command information." Post-commands define the execution of a rule. When a rule is triggered and the priority of a candidate unit is determined, the post-commands determine the number of candidate units to be selected and the operations to be performed on the selected units. By induction, rule representation technology divides commands into two sub-conditions: "action type" and "number of units." Acquiring these two sub-conditions is equivalent to acquiring the post-command information for the rule. Step S14, "Generate JSON file storage rules", the rule representation technology includes a JSON-based rule representation specification, which uses key-value pairs to represent the rule's preconditions and postconditions, and uses a graph data structure to represent the rule's preference information, mapping each rule to a JSON-formatted file for storage; Step S2, conflict detection and resolution: When faced with the same workspace state, different rules may make different decisions, thus causing conflicts. This step checks for possible conflicts between different rules and proposes resolution solutions. In step S2, three types of rule mutual exclusion, including preference contradiction, action mutual exclusion, and action contradiction, are constructed from the logical unsatisfiable types and resources. Based on this, a conflict detection algorithm is designed. By checking and comparing the preconditions, rule preferences, and postconditions, the conflict type is inferred and a resolution solution is generated. This allows rule conflict determination during rule editing and entry, and provides conflict explanations through interface prompts. Step S3, multi-rule collaborative deduction, this step issues instructions that meet the user's intentions to the local work unit based on multiple task-level rules input by the user.

2. The task planning-oriented rule representation and reasoning method with preference according to claim 1, characterized in that: In step S1, the rule classification is formed from the three aspects of rule premises, rule preferences and rule post-commands from the aspects of preconditions, conclusions, preferences, logical operators, workspace status, work unit types, association relationships, and reasoning task types, and the ontology of task-level rules including rule patterns, precondition entities, conclusion entities, preference entities, and mutually exclusive tasks is constructed.

3. The task planning-oriented rule representation and reasoning method with preference according to claim 1, characterized in that: In step S2, conflict resolution is divided into two stages: static conflict resolution and runtime conflict resolution. The former is designed as a human-machine combination of machine prompts and manual resolution based on conflict detection during the rule editing and entry process. The latter is achieved during the rule running stage by designing a strategy including conflict rule set generation and conflict rule selection in the inference engine based on the actual rule content and test results.

4. The task planning-oriented rule representation and reasoning method with preference according to claim 1, characterized in that: In step S3, the preference-based equilibrium model constructed by considering the priority preference relationship in the work unit selection also includes: (1) a single-rule preference processing model, which takes the optimal preference processing strategy on the complete order; (2) a multi-rule preference processing model, which adopts the three classical social choice theory methods of Pareto optimality, worst optimality, and best optimality to form the optimal conclusion selection strategy under the preference conclusion.

Images