Intelligent adjustment and control method for RV air conditioning integrating trip planning

Abstract

This invention relates to the field of intelligent pre-adjustment of RV air conditioning systems. The solution collects structured path navigation data containing terrain and facility attributes, generates a 3D path topology map using a GIS engine, models road nodes, edges, and surfaces in a semantically hierarchical manner, and constructs a hierarchical rule knowledge base based on HVAC engineering principles and multi-source measured data. The system uses a rule graph engine to precisely match path topology features with air conditioning execution parameters, achieving rule-based triggering, temporal conflict resolution, and decision priority ranking, generating a traceable deterministic control command sequence, which is executed and fed back in real time by the microcontroller unit. This method improves the RV air conditioning's adaptive adjustment capability to dynamic terrain changes, achieving pre-adjustment that balances comfort and energy consumption. It also features rule verification and self-iterative optimization mechanisms, promoting system reliability and intelligence.

Description

Technical Field

[0001] This invention relates to the field of intelligent control and route planning technology for RV air conditioning, and in particular to an intelligent adjustment and control method for RV air conditioning that integrates route planning. Background Technology

[0002] Currently, mainstream technologies in the field of intelligent control and route planning for RV air conditioning mainly include predictive response adjustment based on end-to-end deep learning models, dynamic energy consumption optimization using reinforcement learning systems, and scene perception algorithms based on multimodal data fusion. These technological trends aim to improve the adaptive adjustment capabilities of RV air conditioning in complex travel scenarios, achieving a balance between energy efficiency and passenger comfort. Existing industry practices commonly employ black-box neural network structures, training models using multi-source data such as historical route data, climate parameters, and high-frequency sensor inputs. Deep learning or neural networks are then used to fit user behavior and environmental characteristics, enabling automatic adjustment within the vehicle's air conditioning control system. For example, some publicly available intelligent vehicle air conditioning control solutions utilize end-to-end predictive models, combining route elevation changes, real-time weather forecasts, and user preference data to generate adjustment commands to drive actuators, thereby achieving travel linkage and environmental adaptation.

[0003] Current mainstream technologies are applied in intelligent air conditioning control scenarios for RVs, new energy vehicles, and some high-end passenger vehicles. Their typical characteristic is the use of neural network models to perceive and respond to the future path environment, thereby improving user experience and reducing energy consumption. However, due to the black-box nature of deep learning models, users cannot clearly understand the role of different travel characteristics (such as altitude changes, tunnel ratios, slope, and curvature) in the specific air conditioning strategy generation process, resulting in poor model interpretability and affecting subsequent strategy intervention, system maintenance, and user trust. While some existing technologies possess certain environmental parameter mapping capabilities, they lack transparent disclosure of the rule triggering principles and strategy generation chain.

[0004] The industry generally suffers from the following technical shortcomings: First, while neural network-based intelligent control systems have highly automated strategy generation processes, their internal decision-making logic remains invisible, making it difficult for users and maintenance personnel to trace the basis for strategy judgments and lacking reliable auditing and error correction mechanisms. Second, the linkage rules between path characteristics and air conditioning actions are highly complex, and existing black-box models cannot effectively support rule-based consensus and refined manual intervention, especially under multi-scenario concurrent conditions, which can easily lead to strategy response failures or decreased comfort. Third, in actual RV scenarios, user needs are highly diverse, requiring high levels of strategy controllability and process transparency, while existing systems struggle to provide targeted explanations and feedback, affecting system promotion and application satisfaction. Fourth, during model iteration and maintenance, due to the lack of rule traceability, it is difficult to promptly correct abnormal responses to personalized needs or special climatic and geographical environments, resulting in limited system adaptability and security. Summary of the Invention

[0005] This application provides a method for intelligent adjustment and control of RV air conditioning that integrates travel planning, aiming to solve one of the problems or issues of the prior art mentioned in the background section.

[0006] The intelligent adjustment and control method for RV air conditioning that integrates trip planning provided in this application specifically includes: S1: Obtain structured navigation data of the RV's travel route, and perform topology enhancement processing on the structured navigation data to generate a three-dimensional path topology map containing nodes, edges, and faces; S2: Based on the principles of HVAC engineering and measured thermal response data, a hierarchical rule knowledge base is constructed; S3: Input the three-dimensional path topology graph into the rule graph engine, traverse the hierarchical rule knowledge base for each node in the three-dimensional path topology graph, match rule entries that meet the triggering conditions, and generate an initial rule activation set with execution priority markers; S4: Perform temporal conflict resolution and logical consistency verification on the initial rule activation set, use a directed acyclic graph to model the dependency relationship between each rule node, and output a structured strategy logic flow graph; S5: Compile the strategy logic flow graph into a deterministic control instruction sequence. The deterministic control instruction sequence includes the target operating parameters of each actuator of the air conditioning system and their effective timestamps. The parameters include compressor frequency, damper opening, fan speed and operating mode. The deterministic control instruction sequence is used by the vehicle microcontroller unit to control the execution adjustment of the air conditioning actuator.

[0007] The intelligent adjustment and control method for RV air conditioning that integrates trip planning provided in this application has the following beneficial effects: (1) By constructing a collaborative reasoning mechanism based on a 3D path topology map enhanced by geographic information system and a hierarchical rule knowledge base, the forward-looking control capability and operating condition adaptability of the air conditioning system in complex dynamic driving environments are significantly improved. Traditional vehicle temperature control strategies mostly rely on real-time feedback from sensors for passive response, which is difficult to cope with the thermal environment lag effect caused by long-term, large-scale terrain changes. Especially in typical scenarios such as high-altitude climbing, continuous tunnel groups, or long downhill slopes, problems such as condensation and freezing, insufficient fresh air, or excessive energy consumption are likely to occur. This solution pre-analyzes key terrain feature nodes in the travel path and combines them with physical constraint rules to intervene in core parameters such as compressor preheating, damper opening and closing, and cooling power distribution in advance. This upgrades the control timing from a "perception-response" mode to a "prediction-decision-execution" closed loop, enabling key equipment state switching to be completed hundreds of milliseconds to minutes in advance, significantly reducing the risk of thermal shock and effectively ensuring driving comfort and system reliability. At the same time, relying on a graph-based rule engine to traverse and match the path topology node by node, it ensures that each control command has clear engineering basis and spatiotemporal traceability, fundamentally overcoming the technical defects of black box models that are unexplainable, difficult to debug, and difficult to verify.

[0008] (2) By adopting a hierarchical rule modeling and a strategy generation architecture driven by a directed acyclic graph, explicit decoupling and logical consensus among multi-objective control requirements are achieved, significantly enhancing the system's configurability, controllability, and user trust. Addressing the common conflict between comfort, safety, and energy efficiency in RV operating environments, this solution decomposes HVAC engineering experience into three rule levels: "physical constraints," "comfort compensation," and "energy conservation," corresponding to different priority objectives of safety baseline, experience optimization, and resource balance, respectively. Nonlinear coordination between rules is achieved through confidence weights and execution sequence markers. For example, on continuous uphill sections at high altitudes, the system can automatically suppress aggressive precooling to conserve power, while dynamically adjusting the set temperature to compensate for perceived temperature differences, avoiding overall performance degradation due to single-objective optimization. More importantly, all rule activation chains are output in DAG form, allowing users to view the strategy source node by node in the interactive interface, and enabling the disabling, delaying, or weighting of specific rules, forming a transparent decision-making process of human-machine co-governance. This design not only meets the needs of professional users for fine-grained intervention in policies, but also provides complete log support for subsequent auditing, fault attribution and knowledge iteration, effectively solving the problems of "policy black box" and "user silence" in traditional intelligent control systems.

[0009] (3) By explicitly encoding engineering knowledge into an executable rule graph and combining it with a local trusted execution environment to achieve encrypted evidence storage of the entire policy process, a highly reliable, verifiable, and lightweight autonomous inference system for vehicle air conditioning was constructed, possessing excellent deployment efficiency and system robustness. Compared to deep learning or reinforcement learning methods that rely on large-scale training data and cloud computing power, this solution does not require online learning or continuous parameter tuning, avoiding the risks of model drift, overfitting, and privacy leakage, and is particularly suitable for mobile travel scenarios with unstable network coverage and limited computing resources. The rule graph engine adopts deterministic matching logic, which can complete the policy derivation of the entire path in milliseconds, and the control instruction sequence is completely parsed and executed directly by the MCU, without the need to deploy an additional neural network inference framework, significantly reducing the complexity of hardware and software coupling. At the same time, the key derivation steps, rule matching paths, and traces of manual intervention in each policy generation process are encrypted and stored in the TEE module, which not only ensures the integrity and anti-tampering capability of the control logic, but also provides a secure and reliable data foundation for future optimization of the rule base based on real running data. The overall technical approach has achieved a paradigm shift from "data-driven" to "knowledge-driven", ensuring high-performance control while fully considering the system's interpretability, security, and sustainable evolution capabilities.

[0010] In summary, this solution integrates path topology perception, engineering rule modeling, and graph-based strategy reasoning to construct an intelligent vehicle air conditioning control system that is forward-looking, transparent, and adaptive. It not only effectively improves the accuracy and stability of temperature control response under complex road conditions, but also innovatively establishes a trust consensus mechanism between users and the system, providing a replicable, auditable, and scalable technical model for intelligent environmental management of high-end RVs, special vehicles, and long-distance transportation equipment. Attached Figure Description

[0011] Figure 1 The main flowchart of the RV air conditioning intelligent adjustment and control method integrating trip planning is a preferred embodiment of this application.

[0012] Figure 2 This is a sub-flowchart of a preferred embodiment of the intelligent adjustment and control method for RV air conditioning that integrates trip planning.

[0013] Figure 3 This is another sub-flowchart of the RV air conditioning intelligent adjustment and control method integrating trip planning, which is a preferred embodiment of this application. Detailed Implementation

[0014] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0015] The following disclosure provides many different embodiments or examples for implementing different structures of the invention. To simplify the disclosure, specific examples of components and arrangements are described below. Of course, these are merely examples and are not intended to limit the invention. Furthermore, reference numerals and / or letters may be repeated in different examples; such repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed.

[0016] like Figure 1 As shown, this application provides a method for intelligent adjustment and control of RV air conditioning that integrates trip planning, specifically including: S1: Obtain structured navigation data of the RV's travel route, and perform topology enhancement processing on the structured navigation data to generate a three-dimensional path topology map containing nodes, edges, and faces; S2: Based on the principles of HVAC engineering and measured thermal response data, a hierarchical rule knowledge base is constructed; S3: Input the three-dimensional path topology graph into the rule graph engine, traverse the hierarchical rule knowledge base for each node in the three-dimensional path topology graph, match rule entries that meet the triggering conditions, and generate an initial rule activation set with execution priority markers; S4: Perform temporal conflict resolution and logical consistency verification on the initial rule activation set, use a directed acyclic graph to model the dependency relationship between each rule node, and output a structured strategy logic flow graph; S5: Compile the strategy logic flow graph into a deterministic control instruction sequence. The deterministic control instruction sequence includes the target operating parameters of each actuator of the air conditioning system and their effective timestamps. The parameters include compressor frequency, damper opening, fan speed and operating mode. The deterministic control instruction sequence is used by the vehicle microcontroller unit to control the execution adjustment of the air conditioning actuator.

[0017] Step S1: Obtain structured navigation data for the RV's travel route. This data includes the start and end geographic coordinates, length, elevation changes, radius of curvature, road surface type, and ancillary facility labels for each road segment. A Geographic Information System (GIS) engine is used to perform topology enhancement processing on this data, generating a 3D path topology map containing nodes, edges, and faces. Nodes represent key terrain turning points, edges represent road segments with attribute vectors, and faces represent regional environmental feature areas. Specifically, this includes: S1.1: Obtain structured navigation data of the RV's travel route. The data is provided by the in-vehicle navigation system or cloud-based route planning service, including the starting and ending geographic coordinates, length, elevation change, radius of curvature, road surface type, and ancillary facility labels of each road segment. Perform a unified projection transformation on the coordinate data based on the Global Positioning System (GPS) standard coordinate system to eliminate coordinate offsets between different map sources and obtain a standardized spatial trajectory sequence as the basic input for subsequent processing.

[0018] Input data is provided by in-vehicle navigation systems or cloud-based route planning services, including the starting and ending geographic coordinates, length, elevation changes, radius of curvature, road surface type, and ancillary facility labels for each road segment. The data source may come from different map service providers and the coordinate systems may differ.

[0019] A coordinate system standardization method (parameter: GPS Global Positioning System Standard Coordinate System WGS-84) is adopted to achieve unified processing of multi-source geographic coordinate data.

[0020] Furthermore, by using a coordinate projection transformation algorithm (parameters: ellipsoidal model WGS-84, geodetic coordinates → projected coordinates), the mapping of different map service source coordinate systems to a unified projected coordinate system is realized, and the spatial position vector after eliminating coordinate offset is obtained.

[0021] Furthermore, a spatial trajectory resampling algorithm (parameter: sampling interval fixed at 10 meters) is adopted to reconstruct the path data at equal intervals and generate a high-precision trajectory point sequence as a spatial reference for subsequent topology enhancement.

[0022] Furthermore, a coordinate error filtering algorithm (parameters: Kalman filter model, observation noise covariance matrix R and process noise covariance matrix Q are set according to the accuracy of the navigation equipment) is adopted to suppress coordinate sequence noise and generate a smoothed trajectory matrix, so as to reduce the misjudgment of node offset caused by GPS drift.

[0023] By using a unified projection transformation and error filtering method, the raw navigation data from the previous step is transformed into a standardized and smoothed spatial trajectory sequence, providing a basic input with positional accuracy that meets engineering requirements for the construction of a 3D path topology map.

[0024] For example, navigation data for a certain RV trip comes from the GCJ-02 coordinate system provided by Map A and the WGS-84 coordinate system provided by Map B. These coordinates include start and end coordinates (in degrees), road segment length (in meters), elevation change (in meters), radius of curvature (in meters), road surface type (e.g., asphalt, gravel), and labels for ancillary facilities (e.g., tunnel entrance, bridge start). Using coordinate system standardization, the GCJ-02 coordinate system data from Map A is first transformed to the geodetic coordinate system under the WGS-84 ellipsoidal model. Then, ellipsoidal projection is applied to transform it to a Cartesian coordinate system. A spatial trajectory resampling algorithm is used to resample the path nodes to equidistant 10-meter intervals, forming approximately 3500 trajectory points. A Kalman filter algorithm is then applied to this trajectory sequence, with the noise covariance matrix R set to... The plane noise level is given by the process noise covariance matrix Q. The dynamic error level is adjusted to achieve drift suppression, resulting in a smoothed trajectory matrix. In validation, the positional error of trajectory nodes after this processing is significantly reduced in comparisons with adjacent map sources, and the standard deviation of trajectory offset is reduced from that of the original data. Rice reduced to The data provided in meters provided a high-precision, unified coordinate system for the subsequent identification of key terrain turning points in S1.2.

[0025] S1.2: Perform segmented discretization processing on the standardized spatial trajectory sequence obtained in S1.1, and identify key terrain turning points based on road attribute abrupt change points (such as slope jumps, significant curvature changes, and road surface type switching) and the boundaries of ancillary facilities (such as tunnel entrances / exits, bridge start / end points). Define each key terrain turning point as a node in the three-dimensional path topology map to form path event anchor points with clear physical meaning, which are used to support the accurate matching of rule conditions and the timing annotation of triggering time.

[0026] For the standardized spatial trajectory sequence output in S1.1, a road attribute mutation detection algorithm (parameters: slope change threshold Δslope = 6%, curvature mutation coefficient Δcurvature = 0.05 / m, road surface type coding difference value ≥ 1) is used to achieve discretization and segmentation of the spatial trajectory.

[0027] Furthermore, by using an ancillary facility boundary identification method (parameters: tunnel entrance / exit GeoTag, bridge start / end point GeoTag, facility perception radius R=30m), facility boundary marking of trajectory points is achieved, and an index dataset containing facility type and boundary location is obtained.

[0028] Furthermore, by using a node feature clustering algorithm (parameters: Euclidean distance threshold d=50m, feature vector dimension=slope, curvature, road surface type code), the fusion and matching of road attribute change points and facility boundary points are achieved, and a set of candidate key terrain turning points is generated.

[0029] Furthermore, by using a spatial projection verification method (parameters: GPS standard coordinate system, EPSG:4326 projection criterion), the coordinate accuracy of candidate key terrain turning points is corrected, and a node coordinate table that meets the requirements of physical location uniqueness and data accuracy is generated.

[0030] By using a node event semantic encoding method (parameters: event type enumeration = abrupt slope change, significant curvature change, road surface switching, facility entry and exit), the node coordinate table is transformed into path event anchor points with clear physical meaning, realizing structured three-dimensional path topology node output, and providing basic data for accurate matching of rule conditions and trigger timing annotation.

[0031] For example, in a journey traversing a mixed plateau and mountainous route, the input standardized spatial trajectory sequence contains 2000 trajectory points, each recording altitude, slope, radius of curvature, road surface type code, and facility label. Using a road attribute abrupt change detection algorithm, with a slope change threshold Δslope set to 6% and a curvature abrupt change coefficient Δcurvature set to 0.05 / m, the trajectory is segmented, detecting 45 slope jump points and 38 significant curvature change points. Through ancillary facility boundary identification, tunnel entrances / exits and bridge start / end points are marked, identifying 12 tunnel boundary points and 7 bridge constraint points. Using a node feature clustering algorithm, slope abrupt changes and facility boundary points are merged within a 50m Euclidean distance range, generating 58 candidate key terrain inflection points. Spatial projection verification was employed, transforming the coordinates of all candidate points to the EPSG:4326 coordinate system, correcting deviations to less than 1m, and eliminating nodes with coordinate repetition greater than 90%, ultimately yielding 51 key terrain inflection points. Node event semantic encoding was used, marking 22 of these as "slope abrupt change," 15 as "significant curvature change," 9 as "road surface switching," and 5 as "facility entry / exit." The output node event anchor points were used in subsequent air conditioning pre-conditioning rule matching to trigger actions such as compressor preheating and damper control. Verification results showed that this node identification process significantly improved the accuracy and temporal rationality of rule triggering.

[0032] S1.3: Based on the road segments between adjacent nodes identified in S1.2, construct edges with attribute vectors; the attribute vectors include the average slope of the road segment, cumulative elevation change, average radius of curvature, driving length, estimated travel time, and estimated road surface friction coefficient; use the GIS engine to interpolate and smooth the original navigation data, calculate the differential geometric feature parameters of each road segment, and generate a structured edge attribute matrix, which serves as the functional carrier for connecting nodes in the path topology.

[0033] Based on the set of key terrain turning point nodes output by S1.2, a node connection mapping algorithm (parameters: node sequence, spatial coordinates, road attribute labels) is adopted to realize the unique identification and index construction of road segments between adjacent nodes, and to instantiate each road segment as an edge as a connection element of the path topology relationship.

[0034] Furthermore, the average slope of the road segment is calculated using a road segment attribute calculation algorithm (parameters: node elevation value, node plane coordinates, road segment length). The calculation, in which The elevation difference between adjacent nodes. The horizontal projection length of the road segment is used as the longitudinal characteristic index of the road segment, and the cumulative elevation change value is obtained.

[0035] Furthermore, a curvature analysis algorithm (parameters: node curve fitting coefficients, chord length, arc length) is employed to realize the average curvature radius of the road segment. The calculation, in which The curvature estimate is output as a geometric feature parameter that measures the degree of road curvature.

[0036] Furthermore, by combining road speed limit information with road segment length, a traffic condition prediction algorithm (parameters: speed limit, gradient correction coefficient, road condition weighting factor) is used to estimate travel time. The calculation, in which The corrected average vehicle speed yields the time-series characteristic index.

[0037] Furthermore, by using a friction coefficient estimation model (parameters: road surface type label, climate condition coefficient, historical accident rate compensation factor), the estimated value of the road segment friction coefficient is obtained. The determination of this value is used for subsequent energy consumption safety rule triggering judgment.

[0038] By using the interpolation completion method of the GIS engine (parameters: node spatial coordinates, missing sample point identifiers), the elevation and curvature sampling points of the road segment are reconstructed by interpolation to generate a continuous geometric feature sequence. Furthermore, a smoothing filtering algorithm (parameters: window length, filter weights) is used to suppress noise fluctuations in the original navigation data and improve the stability of parameter calculation.

[0039] Through data structuring, the above six types of road segment attribute parameters are integrated into the edge attribute matrix in a unified format, thereby realizing the construction of the functional carrier of edge elements in the three-dimensional path topology map.

[0040] For example, in a certain RV travel route, a segment consisting of node A (elevation 650m, coordinates: 120.123E, 30.123N) and node B (elevation 740m, coordinates: 120.456E, 30.456N) has a horizontal projected length L of 3.2km, an elevation difference Δh of 90m, and an average slope calculated as follows: =0.0281, with a cumulative elevation change of 90m. The radius of curvature was calculated by fitting the chord length and arc length of the curve for this road segment. =222.22m. The speed limit for this section is 80km / h, the gradient correction factor is 0.95, and the estimated travel time is calculated as follows: =0.0421h, or approximately 151.6 seconds. The road surface type is asphalt, and considering the dry climate conditions, the estimated friction coefficient μ is 0.85. Through GIS interpolation, the elevation sampling points were increased from every 200m to every 50m, and weighted smoothing filtering was used to eliminate elevation noise fluctuations within ±5m, generating a stable geometric feature sequence. Finally, the average slope, cumulative elevation change, average radius of curvature, travel length, estimated travel time, and friction coefficient estimate for this road segment were uniformly included in the edge attribute matrix to support subsequent identification of regional environmental feature surfaces and precise triggering of rule conditions.

[0041] S1.4: Based on the edge attribute matrix generated in S1.3, further identify regional environmental feature areas composed of multiple consecutive edges, including plateau elevation zones (elevation ≥ 2500m and continuous length > 10km), frequent curve zones (more than five consecutive curves with curvature > 0.08 / m), long uphill and downhill slope cluster zones (cumulative absolute slope > 6% and total length > 5km), and dense closed passage zones (the number of tunnels per unit distance exceeds the threshold). The above regions are abstracted into surface units in a three-dimensional path topology map, and each surface is labeled with its environmental category and the time window range it covers, forming a macro-environmental perception layer.

[0042] S1.5: Integrate the nodes generated in S1.2, the edge attribute vectors constructed in S1.3, and the face units divided in S1.4, and use the GIS engine to perform topology enhancement modeling to establish hierarchical relationships between nodes, edges, and faces; generate a 3D path topology map containing spatial location, geometric features, semantic labels, and time series indexes. The topology map provides a traversable graph structure input for the subsequent rule graph engine, ensuring that path features can be systematically parsed and mapped to the trigger logic chain of air conditioning adjustment actions.

[0043] The input conditions include the key terrain inflection point node data generated in S1.2, the edge attribute matrix constructed in S1.3, and the set of surface units divided in S1.4, all of which are structured spatial data projected and attribute-completed in a unified coordinate system.

[0044] A hierarchical topological association algorithm (parameters: node ID index table, edge attribute matrix, and face cell category label) is adopted to realize the bidirectional association relationship between nodes and adjacent edges, and between edges and their respective face cells. The spatial location coordinates and time index of the association are explicitly recorded in the model, so that the hierarchical connection has traversability and temporal traceability.

[0045] Furthermore, by using GIS 3D topology modeling methods (parameters: spatial location coordinates, edge geometric vectors, and polygon boundaries), the geometric nesting and interactive constraint calculation of nodes, edges, and faces in 3D space are realized, and a geometric relationship library containing distance and angle matrices is generated, providing underlying support for rule-triggered spatial determination.

[0046] Furthermore, a semantic label mapping algorithm (parameters: node event category label, edge traffic attribute set, and face environment category label) is used to achieve a bidirectional mapping between spatial topological elements and environmental semantics, generating a semantically enhanced topological dataset so that the rule graph engine can directly identify environmental features and match corresponding conditions during graph traversal.

[0047] Furthermore, by using the time series index binding method (parameters: node timestamp, edge travel time estimate, and surface coverage time window), the cumulative travel time sequence from a node to its neighboring nodes is calculated, and this sequence is bound to the corresponding topological element, so that the three-dimensional path topology map has both spatial and temporal information, enabling the timing labeling of subsequent air conditioning adjustment actions.

[0048] Through the above-mentioned topology enhancement modeling process, the processing results of the previous step are transformed into a complete three-dimensional path topology map data structure, which includes spatial location, geometric features, semantic labels and time series indexes, realizing the systematic analysis of path features in the subsequent rule matching process and the mapping of the air conditioning adjustment action trigger logic chain.

[0049] For example, in a motorhome trip traversing mountains and plateaus, the node set contains 112 key terrain turning points (such as slope abrupt changes and tunnel entrances), with each node associated with its 3D coordinates and event labels. The edge attribute matrix records parameters such as the average slope of adjacent nodes (5.4%), cumulative elevation change (320m), average radius of curvature (460m), and travel time (4 minutes). The polygon set labels the plateau elevation zone (average elevation 2800m, continuous length 12.5km) and the frequent curve area (7 consecutive curvatures > 0.1 / m, covering length 3.2km). During the topology association construction process, node ID 45 is bound to edge ID 87, and edge ID 87 is associated with polygon ID 5 (plateau elevation zone). The spatial overlap rate between this edge and polygon is calculated to be 0.95 through GIS 3D modeling, and a bidirectional mapping is established between the event label "long uphill start point" of node ID 45 and the label "plateau elevation zone" of polygon ID 5 during the semantic label mapping process. In the time index binding stage, the timestamp of node ID 45 is the 146th minute of the journey, the estimated travel time of edge ID 87 is 540 seconds, and the time window covered by face ID 5 is from the 146th minute to the 164th minute. The final generated 3D path topology map has spatial and temporal dual-domain information and environmental semantic parsing capabilities. When the rule graph engine traverses to node ID 45, it triggers a "high-altitude low-temperature preheating" rule and implements a pre-adjustment action to adjust the compressor frequency 180 seconds in advance.

[0050] Step S2: Based on HVAC engineering principles and measured thermal response data, a hierarchical rule knowledge base is constructed. This knowledge base includes physical constraint rules, comfort compensation rules, and energy conservation rules. The physical constraint rules set compressor preheating conditions based on slope and duration; the comfort compensation rules set temperature and wind speed adjustment parameters based on altitude rise and continuous curve characteristics; and the energy conservation rules set the upper limit of cooling power and the timing of cold energy recovery based on subsequent terrain energy distribution. All rules are expressed in parsable condition-action statement form. Specifically, this includes: S2.1: Based on the ASHRAE HVAC design standard and the measured thermodynamic response dataset of RVs, key variable pairs such as slope-compressor start-stop sequence, altitude-perceived temperature offset, and curvature-airflow disturbance intensity are extracted. Correlation analysis and threshold segmentation algorithms are used to determine the critical action interval between each variable, and a preliminary set of physical constraint rules with clear trigger boundaries is generated.

[0051] The input object is a dataset of measured thermodynamic responses of motorhomes under different path scenarios, which combines the ASHRAE HVAC design standard clauses. The dataset includes slope information, altitude change records, road curvature parameters, and corresponding air conditioning system compressor start and stop times, in-vehicle perceived temperature offset, and airflow disturbance intensity values ​​obtained by vehicle sensors.

[0052] A correlation analysis method (parameter: Pearson correlation coefficient calculation window length set to 10 minutes) was used to quantitatively assess the correlation between the slope change rate and the compressor start-stop sequence.

[0053] Furthermore, by using correlation analysis (parameters: altitude gradient and perceived temperature offset within the same dataset), the linear dependence coefficient between altitude change and comfort loss is extracted, and the direction and intensity of the effect are obtained.

[0054] Furthermore, a correlation analysis method (parameter: the inverse of the radius of curvature represents the curvature compactness and the intensity of airflow disturbance) is used to realize the mapping relationship between curvature change characteristics and the decrease in airflow stability, and to generate a curvature sensitivity index.

[0055] Furthermore, by using a threshold segmentation algorithm (parameter: threshold setting is based on the statistical range of the 95% confidence interval), the critical action interval of the above-mentioned relevant feature variables is defined, and the upper and lower limit values ​​of each variable triggering the physical constraint rules are obtained.

[0056] The adjustment value of altitude change to target temperature is calculated using the following formula:

[0057] in, This is the temperature adjustment value (in degrees Celsius). The elevation gain is in meters, with a denominator of 500 meters.

[0058] Furthermore, the compressor preheating time corresponding to the critical slope value is calculated using the formula:

[0059] in, Preheating time (seconds) is the time required for the product to warm up. This represents the slope percentage value, with the denominator 6 corresponding to the slope trigger threshold.

[0060] Through the above algorithm and formula processing, the correlation analysis results and threshold interval division results of the previous step are transformed into a set of initial draft physical constraint rules with clear trigger boundaries, realizing the quantitative definition and interpretable expression of rule triggering conditions.

[0061] For example, in a motorhome trip including long uphill and high-altitude sections, input data came from onboard IMU sensors and ambient temperature and humidity sensors. The average slope change rate was 7%, the altitude increased by 1500 meters from the start of the trip to the target section, and the average radius of curvature corresponded to a curve compactness of 0.09 / m. The compressor start-stop log showed that preheating started 160 seconds in advance when the slope exceeded 6%. In the correlation analysis, the slope change rate and start-stop timing Pearson coefficient reached 0.82, the altitude change and perceived temperature shift linearity coefficient was 0.002℃ / m, and the curvature compactness and airflow disturbance intensity coefficient was 0.75. Based on threshold segmentation calculation, the slope trigger threshold was 6%, and the preheating time calculation formula result was... =210 seconds; the formula for calculating the temperature adjustment value based on altitude change is: =0.9℃. The initial draft of the rule included the following constraints: when the slope is ≥6% and the continuous length is ≥2km, the compressor needs to preheat at low frequency for 210 seconds in advance; when the altitude increase is ≥500m, the target set temperature is increased by 0.3℃, and when the altitude increase is 1500m, it is increased by 0.9℃. The actual implementation effect showed that the perceived temperature inside the vehicle was significantly increased when reaching the plateau section, and no airflow fluctuation was observed, verifying the accuracy of the rule's triggering and the effect of improving comfort.

[0062] S2.2: Perform engineering feasibility verification on the initial set of physical constraint rules generated in S2.1. Based on the technical specifications of the vehicle air conditioning system, limit the action boundaries of the actuators—perform parameter mapping processing on the minimum operating frequency of the compressor, the mechanical travel limit of the damper, and the speed regulation resolution of the fan to ensure that the action commands output by the rules are within the controllable range of the equipment, thus forming a physical constraint rule sub-library with physical implementability.

[0063] Based on the initial set of physical constraint rules obtained from S2.1, the input data includes the trigger condition parameters (slope critical value, duration, tunnel advance, etc.) and action setting parameters (compressor start-stop frequency, damper opening, fan speed, etc.) defined in the rules, as well as the physical limit configuration data of the actuator in the technical specifications of the vehicle air conditioning system.

[0064] A parameter mapping processing method is adopted (the mapping table is based on the technical specifications provided by the equipment manufacturer) to realize the correspondence between the virtual action parameters in the initial draft of the rules and the actual executable range of the equipment. For example, the compressor preheating frequency in the rules is mapped to the lowest sustainable operating frequency level in the technical specifications, and parameter truncation or normalization is performed when the range exceeds the specification.

[0065] Furthermore, by using a mechanical stroke limit detection algorithm (parameter: maximum reading θ_max of the damper stepper motor angle sensor), the boundary comparison between the target value of the damper opening in the rules and the maximum stroke of the mechanism is realized. When the target opening exceeds θ_max × safety factor, it is adjusted to the corresponding gear value within the safety range, and the adjustment factor is recorded.

[0066] Furthermore, by using a digital speed regulation resolution verification method (parameters: fan speed regulation step Δv, minimum resolution σ_v), the alignment of the target fan speed value in the rules with the speed gears that the controller can distinguish is achieved. When the target value does not meet the gear dispersion, it is rounded to the nearest gear to ensure that the MCU control signal accurately matches the hardware capability.

[0067] Furthermore, the boundary value of the compressor's preheating power requirement is calculated using a formula under the conditions of slope and duration:

[0068] in, The calories that need to be replenished, To preheat the target temperature difference, For the thermal efficiency of the air conditioning system, The allowable preheating time is determined by the P value output by the formula. The compressor operating frequency in the rules is then reversed to ensure that the energy consumption is within the equipment's power supply capacity.

[0069] Through the above algorithm or processing method, the initial draft of the rules in the previous step is transformed into physical constraint rule entries that have been verified by the device boundary constraints and whose parameters can be directly executed, thereby realizing the executability and safety of the rule sub-library in the actual RV air conditioning system.

[0070] For example, under a long uphill condition on a plateau, the slope detection module outputs a slope value of 8.5% and a continuous length of 3.2km. The initial draft of the physical constraint rules sets the compressor preheating frequency to 30Hz. The vehicle air conditioning technical specifications stipulate a minimum sustainable operating frequency of 28Hz. Since the target value is higher than the minimum setting but within the rated range, the parameter mapping maintains the original value of 30Hz. The target damper opening is set to 95%, and the stepper motor angle sensor detects an opening of 92% corresponding to the maximum stroke, with a safety factor of 0.98. After calculation by the limit adjustment algorithm, the target value is corrected to 90%. The target fan speed is set to 1470rpm, and the controller can recognize a step of 50rpm, which is rounded to the nearest setting and corrected to 1450rpm. In the compressor preheating power boundary calculation, the heat demand Q = 5.8kW·min, the target temperature difference ΔT = 4℃, the thermal efficiency η = 0.85, and the allowable preheating time t = 180s. Substituting these values ​​into the formula:

[0071] The calculated value of P is approximately 0.152kW, which is less than the rated power supply capacity of the air conditioning system, therefore the operating frequency does not need to be reduced according to the rules. After the above steps, the generated physical constraint rule entries are feasible within the hardware capabilities and safety boundaries, and can maintain the expected thermal response effect during MCU execution.

[0072] S2.3: Based on user comfort survey data and physiological perception experiment reports from three typical regions—plateau, mountain, and plain—a linear regression modeling method was used to calculate the equivalent temperature compensation value corresponding to every 500m increase in altitude. Combining the fitting relationship between the average curvature of continuous curves and the airflow discomfort scores reported by users, a comfort compensation rule sub-library was constructed with 'altitude increase' and 'number of continuous curves' as inputs. All rules are expressed in the form of 'when...then...' as the target temperature rise and the fan speed reduction level.

[0073] Based on user comfort survey data and physiological perception experiment reports processed by regional classification, the input data includes altitude gradient sequences of three typical regions: plateau, mountain, and plain; historical curves of ambient temperature and humidity; user body surface temperature measurements; and user airflow discomfort scores synchronized with the average curvature of continuous curves and vehicle travel time.

[0074] A linear regression modeling method (parameters: independent variable is altitude gradient ΔH, dependent variable is equivalent temperature compensation value ΔT) is used to fit the quantitative relationship between altitude change and user perceived temperature, and the temperature compensation slope coefficient is obtained.

[0075] Furthermore, the equivalent temperature compensation value corresponding to each 500m increase in altitude was calculated by fitting an equation, using the formula:

[0076] in, The regression slope coefficient, To measure the elevation gain, a set of temperature compensation coefficients characteristic of the region is generated.

[0077] Furthermore, the Kendall correlation coefficient calculation method (parameters: x is the average curvature of continuous curves, y is the airflow discomfort score) is used to quantify the rank correlation between curvature and discomfort, and obtain the significance level p-value.

[0078] Furthermore, a quadratic polynomial fitting method is adopted (parameters: the independent variable is the average curvature of the continuous curves, and the dependent variable is the fan speed adjustment level) to establish the wind speed compensation function caused by airflow disturbance and generate a complete curvature-wind speed adjustment mapping table.

[0079] By defining rules using templates, the aforementioned regionalized temperature compensation coefficients and curvature-wind speed adjustment mapping table are embedded into conditional-action statements, expressed as IF<altitude rise>AND<number of consecutive curves>THEN<target temperature rise>AND<fan speed reduction level>WITH<confidence weight>, thus achieving logical encoding of the comfort compensation rule sub-library.

[0080] For example, during a trip in a high-altitude region, the input parameters include an altitude gradient ΔH = 1000m, an average curvature of continuous curves c = 0.09 / m, and a user airflow discomfort score of 7. The linear regression model has a regression slope coefficient k = 0.0006, which is then substituted into the formula. An equivalent temperature compensation value of ΔT = 0.6℃ was obtained. Kendall correlation analysis yielded a rank correlation coefficient τ = 0.68, with a significance level of p < 0.05, confirming a high correlation between curvature and discomfort scores. The quadratic polynomial fitted curve at c = 0.09 / m corresponds to a fan speed reduction of level 1. The rule is expressed as: IF Altitude rise ≥ 1000m AND Number of consecutive curves ≥ 5 THEN Target temperature rise = 0.6℃ AND Fan speed reduction level = 1 WITH Confidence weight = 0.85. This rule significantly improved the user's subjective comfort score during the vehicle journey in the test scenario, and reduced wind speed disturbance, achieving simultaneous assurance of improved comfort and rule interpretability.

[0081] S2.4: Based on the vehicle dynamics model and braking energy recovery characteristic curve, analyze the energy distribution pattern of long uphill and downhill sections in the subsequent journey, and use the power balance equation to calculate the net energy consumption budget under different terrain combinations; based on the budget value, execute the dynamic programming algorithm, set the upper limit ratio of cooling power and the timing of cold energy regeneration during the coasting stage, and generate a sub-library of energy conservation rules, in which each rule outputs two control parameters: the proportion of cooling power and the cold energy storage time window.

[0082] S2.5: The physical constraint rule sub-libraries, comfort compensation rule sub-libraries, and energy conservation rule sub-libraries constructed in S2.2, S2.3, and S2.4 respectively are standardized in format. Standardized condition-action statement templates (IF<condition>THEN<parameter adjustment>WITH<confidence weight>) are used for encoding conversion, and reference source tags (such as standard number, experimental batch number) are added. Finally, they are integrated into a hierarchical rule knowledge base with consistent structure, clear semantics, and machine parsing, which serves as the input for the decision-making basis of the rule graph engine.

[0083] Using the physical constraint rule sub-libraries, comfort compensation rule sub-libraries, and energy conservation rule sub-libraries formed in S2.2, S2.3, and S2.4 respectively as input objects, a rule syntax parser (parameters: support condition-action structured abstraction, template matching threshold = 0.95) is used to extract rule semantic elements and standardize the mapping of condition logic.

[0084] Furthermore, by using a condition-action template matching algorithm (parameters: IF-THEN structural constraint, action parameter type inference mode = strong type), the structure of rules from different sources is unified, and a unique condition expression tree is generated for each rule to eliminate the differences in condition field naming and operator expression across sub-libraries, thus obtaining unified condition set data.

[0085] Furthermore, a confidence weight normalization processing method (parameters: weight range [0,1], normalization strategy = min-max linear scaling) is adopted to achieve cross-domain comparability of rule weight values ​​of each sub-library, and the normalized weights are embedded in the WITH field of the standardized statement template to maintain the accuracy of subsequent rule graph inference.

[0086] Furthermore, by using a source label appending algorithm (parameters: source type = [standard number, experimental batch number], and label sequence ID automatically generated rule), a clear reference basis, such as the ASHRAE standard number or thermal response experiment ID, is appended to the metadata area of ​​each rule, forming a set of rule metadata with traceability capabilities.

[0087] The standardized condition set, action parameter set, confidence weight, and source label are encapsulated into a unified format condition-action statement using a template encoding converter, for example: <conditional expression> <Parameter Adjustment> The confidence weights are ultimately integrated into a hierarchical rule knowledge base that is structurally consistent, semantically clear, and machine-parsable, serving as the input for the rule graph engine's decision-making process.

[0088] For example, in the construction of a strategy for a motorhome trip to high altitudes, the physical constraint rule sub-library generated by S2.2 includes rule R_01, with the condition that the slope is >6% and the duration is >2km, and the action is compressor low-frequency preheating for 3 minutes with a confidence weight of 0.92; the comfort compensation rule sub-library generated by S2.3 includes rule R_15, with the condition that the altitude increase is ≥1500m and the number of consecutive curves is ≥8, and the action is temperature increase of 0.8℃ and fan speed reduction by 2 levels with a confidence weight of 0.88; the energy conservation rule sub-library generated by S2.4 includes rule R_25, with the condition that there are ≥2 long uphill sections within the next 10km, and the action is cooling power limit of 65% and cold energy recovery time window of 30s with a confidence weight of 0.90. A rule syntax parser is used to extract the condition sets and action sets of the three rules, and template matching is used to unify the condition fields, mapping the slope threshold of R_01 to... The length threshold is mapped to The confidence weights are normalized using a min-max method while maintaining their original proportions. Source tags are added, such as R_01 referencing ASHRAE-55:2017, R_15 referencing the user comfort experiment batch UEX_2023_07, and R_25 referencing the RV energy consumption curve experiment EC_2023_11. Finally, the three rules are encoded into a standard template: IF<condition set>THEN<action parameter set>WITH<confidence weight>[reference source], resulting in hierarchical knowledge base entries that can be directly parsed by the rule graph engine. These entries are successfully matched and activated during subsequent rule node traversal in S3, significantly improving the accuracy of condition parsing and the transparency of rule triggering.

[0089] like Figure 2 As shown, step S3 involves inputting the 3D path topology map generated in S1 into the rule graph engine. For each node in the map, the hierarchical rule knowledge base defined in S2 is traversed to match rule entries that meet the triggering conditions. The rule number, action parameters, and timing requirements are recorded to generate an initial rule activation set with execution priority markers. Specifically, this includes: S3.1: Based on the set of nodes in the 3D path topology map output by S1, extract the topological attribute vector associated with each node. The attribute vector includes the slope change rate, elevation difference, curvature change intensity, road surface type transition identifier, and ancillary facility label status of the road segments before and after the node. Use GIS spatial interpolation algorithm to reconstruct the continuous terrain features between discrete nodes and generate a locally terrain feature descriptor with physical meaning, which serves as the basis for the input conditions of rule matching.

[0090] Based on the node set in the 3D path topology map output by step S1, an attribute vector extraction algorithm (parameters: node ID, spatial coordinates, connection edge attribute matrix, and surface cell label) is used to retrieve and aggregate the preceding and following road segment data associated with each node. Furthermore, a slope change rate calculation method (parameters: elevation data of adjacent nodes, road segment length) is used to extract the slope change rate of the road segments before and after a node, obtaining numerical results reflecting the terrain undulations.

[0091] Furthermore, by using an elevation difference calculation method (parameters: elevation data before and after the node), the cumulative elevation difference between adjacent road segments at the node is obtained, and an elevation index matrix describing the vertical height change is generated. Furthermore, by using a curvature change intensity analysis method (parameters: road segment curvature radius sequence, curvature change threshold), the geometric turning characteristics at the node are quantitatively determined, and a curvature change intensity vector reflecting the severity of the route's bending is generated.

[0092] Furthermore, a pavement type transition detection algorithm (parameters: pavement type encoding sequence, transition boundary recognition rules) is used to determine the pavement type switching between adjacent road segments of a node and output a pavement type transition identifier. Further, an ancillary facility label parsing method (parameters: node ancillary facility metadata, facility category code table) is used to determine the facility status at the node and generate a facility status vector containing semantic labels such as tunnel entrance / exit, bridge start / end point, and plateau road section start point.

[0093] Furthermore, a GIS spatial interpolation algorithm (parameters: node spatial coordinates, road segment differential geometric features, interpolation kernel function type) is used to reconstruct continuous terrain features between discrete nodes, resulting in a set of physically meaningful local terrain feature descriptors. This processing method transforms the node topological attributes from the previous step into rule-matching input conditions, achieving a standardized expression of node environmental features.

[0094] For example, in a certain RV travel route, the elevations before and after node N12 are 1450m and 1505m respectively, the road length is 3km, the radii of curvature are 450m and 200m respectively, the road surface types are asphalt and gravel respectively, and the facility label at the node is a tunnel exit. The slope change rate calculation formula is used:

[0095] in, The elevation of the subsequent road section is 1505m. The elevation of the preceding road section is 1450m. For a road segment of 3km in length, the calculated slope change rate is 1.83%. The formula for calculating the curvature abrupt change intensity is used: in, It is 450m. The curvature abrupt change intensity was calculated to be 0.556, given a depth of 200m. Using a GIS spatial interpolation algorithm, the terrain was reconstructed based on the coordinate data of nodes N11 to N13, resulting in a local terrain feature descriptor for node N12: [slope change rate 1.83%, elevation difference 55m, curvature abrupt change intensity 0.556, road surface type transition: asphalt → gravel, facility label: tunnel exit]. This descriptor, after formatting, serves as input for rule matching, enabling the rule graph engine to accurately assess the node's trigger adaptability in air conditioning pre-conditioning strategies, achieving strategy generation that balances comfort and energy efficiency.

[0096] S3.2: Perform pattern matching between the local terrain feature descriptors generated in S3.1 and the condition-action statements in the hierarchical rule knowledge base constructed in S2. Use a rule parsing engine based on predicate logic to determine the truth value of the preconditions of each rule. When all the input conditions of a rule are satisfied by the feature descriptor of the current node, the rule is determined to be triggered. Output the rule number corresponding to the rule and the set of its bound action parameters as the preliminary matching result.

[0097] S3.3: Mark the timing attributes of all triggered rule entries in S3.2, calculate the effective time window of each rule's corresponding action based on the direction of travel and the position of the node sequence, and combine the advance parameters defined by the rule itself (such as '15s before entering the cave' or '3 minutes before going uphill') to calculate the target execution timestamp of each rule action. Then, attach the timestamp as a timing requirement to the rule record to form a rule activation candidate set with timing constraints.

[0098] For the rule entries triggered in S3.2, the spatial positioning of the rule's action timing is achieved using the path node sequence analysis method (parameters: node index, travel direction vector). Furthermore, the estimated travel time between nodes is calculated using a vehicle speed prediction model (parameters: road type, slope, elevation difference, radius of curvature), and the time distance data from each rule trigger point to the actual execution point is obtained.

[0099] Furthermore, an advance correction algorithm (parameters: the advance value and unit form defined in the rule) is used to calculate the offset of the target time of the rule action and generate candidate values ​​for the offset target time. Further, a formula is derived based on the time coordinate:

[0100] in The execution timestamp of the rule target. For the arrival time of the trigger node, This is the time offset calculated based on the rule lead time. The precise action timestamp data is obtained through this formula.

[0101] Furthermore, a time-series constraint attachment method (parameters: rule number, action parameter set, execution timestamp) is used to bind the calculated target timestamp to the rule entry metadata structure and update the "time-series requirement" field in the rule record to form a rule activation candidate set containing time constraints.

[0102] By using the combined algorithm based on spatial positioning, velocity prediction, lead time correction and timestamp binding, the initial set of triggering rules from the previous step is transformed into candidate rule data with a clear execution time, thereby achieving the determinism and executability of the rule set in the time dimension.

[0103] For example, in a RV trip route topology map, node N5 corresponds to a road gradient of 6% and a length of 2km. The estimated travel time calculated by the vehicle speed prediction model is 180s. The lead time defined by the physical constraint rule matched to this node is 3 minutes, i.e. =180s. Node arrival time. =12:00:00, then according to the formula The target execution timestamp was calculated to be 11:57:00. This timestamp was then appended to the rule metadata, forming a rule activation candidate set containing rule number R002, action parameter "compressor frequency low-frequency mode", target execution timestamp 11:57:00, and type label "physical constraint". In simulation verification, this rule action was triggered at the predetermined time, and the compressor operating state was adjusted in advance, significantly improving the stability of air conditioning thermal management during long uphill runs, without significantly increasing energy consumption.

[0104] S3.4: Based on the priority levels (physical constraint level, comfort compensation level, and energy conservation level) of each rule in the knowledge base as preset in S2 and the timestamp order marked in S3.3, the rule activation candidate set is sorted by priority. A multidimensional sorting algorithm is used to sort the rules in ascending order with priority as the primary key and timestamp as the secondary key to ensure that physical constraint rules with high security level are retained and passed on priority over other types of rules, and an initial rule activation set with execution priority mark is generated.

[0105] S3.5: The initial rule activation set output by S3.4 is formatted and encapsulated. Each record contains the rule number, original triggering condition, actual matching value, action parameters, target execution timestamp and priority level. It is transmitted to the conflict resolution module of the next stage through the vehicle communication bus protocol. At the same time, a copy is cached locally for auditing and tracing to ensure that the rule triggering process is verifiable and traceable.

[0106] like Figure 3As shown, step S4 involves resolving temporal conflicts and verifying logical consistency of the initial rule activation set generated in S3. A directed acyclic graph (DAG) is used to model the dependencies between rule nodes, eliminating action contradictions caused by concurrent multiple rules. A structured strategy logic flow graph is output, where each node includes a rule number, triggering condition, target parameter, expected effect, and confidence weight. Specifically, this includes: S4.1: Based on the initial rule activation set output in S3, obtain each activated rule entry and its associated attributes, including rule number, action parameters (such as compressor frequency setting value, damper opening level), triggering conditions (such as slope > 6% and continuous length > 2km), execution timing requirements (such as starting 3 minutes in advance), and confidence weight labels from the hierarchical rule knowledge base. These are used as input objects for building the temporal dependency model to support subsequent detection of time overlap and action conflict identification among multiple rules.

[0107] S4.2: Perform timeline mapping processing on the execution timing requirements of the rule entries obtained in S4.1. Utilize the geographic coordinates of nodes in the travel path topology map and the vehicle's expected driving speed model to calculate the absolute effective timestamp interval of each rule action and generate a rule execution time period sequence with time window markers. Transform the abstract 'X seconds in advance' into a specific time interval data structure to provide a comparable time benchmark for judging concurrent conflicts.

[0108] S4.3: Based on the rule execution time sequence generated in S4.2, execute the multi-rule concurrent conflict detection algorithm. For rule pairs that act on the same air conditioning actuator (such as fan speed, evaporator damper) within the same or overlapping time window but have inconsistent set parameters, identify them as potential action contradictions. For example, one rule requires 'damper closed' while another simultaneously requires 'damper fully open', and output a list of conflicting rule pairs and their related control variables and time overlap indicators.

[0109] Based on the rule execution time period sequence generated in S4.2, a time window overlap detection method (parameters: execution time period interval list, time overlap threshold) is used to realize the interrelationship analysis of all rule entries under the same time base.

[0110] Furthermore, by using a control variable matching algorithm (parameters: execution component identifier, action parameter type), rule entries acting on the same air conditioning execution component are categorized, and key execution components such as fan speed and evaporator damper are taken as objects of interest for conflict detection, resulting in a rule set divided by component dimension.

[0111] Furthermore, a parameter consistency comparison method (parameters: target operating parameters in the classification rule set) is adopted to compare the target parameter values ​​of the same execution component in the same or overlapping time windows one by one. When a value inconsistency or mode conflict is detected, a potential action contradiction marker is generated.

[0112] Furthermore, the potential action conflicts that have been marked are quantitatively assessed using the conflict intensity calculation formula, as follows:

[0113] in, and These are the target parameter values ​​for the conflict rules on the same component. The normalized difference output by this formula serves as an index of conflict intensity, representing the upper limit of the rated parameters of the actuator.

[0114] Furthermore, the time overlap ratio of conflicting rule pairs is evaluated using the time overlap calculation formula:

[0115] in, The length of the intersection of the conflict rules and their effective periods. The ratio is the length of the union of the effective periods of the two events, and it serves as an indicator of the degree of concurrency as a time overlap index.

[0116] The conflict rule output generator records the detected potential action conflict rule pairs in a structured manner according to the execution component, conflict intensity and time overlap, forming a list of conflict rule pairs, thereby realizing the explicit identification and quantitative description of concurrent conflicts.

[0117] For example, in a motorhome travel route, two rules are triggered before and after a certain plateau section: Rule R12 requires the evaporator damper to close at node N5, with a target opening parameter. The effective time interval is from t=300s to t=360s; Rule R17 requires that the dampers be fully open at the same node, with the target opening parameter... The effective time interval is from t=320s to t=380s. Using a control variable matching algorithm, two rules are categorized and applied to the evaporator damper, resulting in the following parameter pairs: , Substituting into the conflict intensity calculation formula, we get: = This indicates that the intensity has reached the rated upper limit. The time overlap calculation formula is output. = This indicates that the two rules overlap by 50% during their effective period. The final conflict rule, with component ID EVF001, conflict intensity 1.0, and time overlap 0.5 recorded in the list, is used for subsequent priority arbitration and parameter reconciliation, achieving accurate quantification and traceability of conflict detection.

[0118] S4.4: The conflict rules identified in S4.3 are processed by the priority arbitration mechanism. Based on the priority order of rule types preset in the hierarchical rule knowledge base (physical constraint rules > comfort compensation rules > energy conservation rules), high-priority rule actions are retained, and low-priority rules are subject to delayed execution, partial disabling, or parameter reconciliation operations. The arbitration decision basis is recorded, and a modified conflict-free rule activation sequence is generated to ensure that all actions to be executed are compatible in the spatiotemporal dimension.

[0119] Based on the list of conflicting rule pairs identified by S4.3, a priority arbitration algorithm (parameters: rule type priority table, conflicting rule pair list, time overlap index) is used to retain the actions of high-priority rules. Furthermore, a delayed execution adjustment method (parameters: low-priority rule effective timestamp, delay calculation formula) is used to achieve staggered execution of low-priority rules in the time dimension, resulting in an updated timestamp sequence. Further, a partial disabling strategy (parameters: low-priority rule trigger condition matching degree, conflict variable importance coefficient) is used to temporarily block some conflicting rules and generate a list of blocked markers. Further, a parameter reconciliation algorithm (parameters: conflict variable value range, acceptable deviation range of the executing agency) is used to take the weighted average of the output parameters of high and low priority rules when a conflict occurs, obtaining the reconciled target operating parameter value. Through the arbitration decision record module (parameters: arbitration type, involved rule number, calculation basis), the decision log generated by the above arbitration process is bound to the corresponding rule record, achieving traceability and verifiability of rule conflict handling.

[0120] For example, during a long RV trip, the conflict rule list includes a physical constraint rule (rule number R101, trigger condition: gradient > 6% and duration > 2km, target parameter: compressor frequency 40Hz) and a comfort compensation rule (rule number R205, trigger condition: number of consecutive curves > 5 and average curvature > 0.08 / m, target parameter: fan speed at medium speed). The priority arbitration algorithm determines that R101 has a higher priority than R205 based on the rule type table, and retains the action parameters of R101 unchanged. The delayed execution adjustment method postpones the target execution timestamp of R205. Seconds are taken to ensure no overlap with R101 compressor preheating. The partial disable strategy detects that the R205 trigger condition matching degree is below the set threshold. A blocking flag is generated to prevent the rule from being executed on that road segment. The parameter harmonic algorithm detects a conflicting variable at another location: damper opening. R102 requires the damper to be closed, while R208 requires it to be fully open. The system calculates the weighted average: Obtain the harmonic value %, and allocated to the control instruction sequence. The above arbitration decision records include arbitration type "parametric reconciliation", involve rule numbers R102 and R208, and are calculated based on the acceptable opening deviation of the enforcement agency ± In the final output of the conflict-free rule activation sequence, all rule actions are compatible in terms of time, space, and logical conditions, improving the stability and comfort of the control strategy execution.

[0121] S4.5: Input the conflict-free rule activation sequence generated in S4.4 into the Directed Acyclic Graph (DAG) construction module. With each rule as a graph node and the time sequence and logical dependencies as directed edges, construct a structured policy logic flow graph. Each node encapsulates the rule number, triggering condition, target parameter, expected effect and confidence weight. Edges represent execution order constraints. Finally, output a DAG structure that satisfies the topological sorting condition, which serves as a verifiable and traceable formal intermediate product for the next step of compiling the deterministic control instruction sequence.

[0122] Based on the conflict-free rule activation sequence generated in S4.4, a directed acyclic graph construction algorithm (parameters: node set, timestamp sequence, logical dependency matrix) is used to initialize the graph structure of the rule nodes and prepare the topological sorting input.

[0123] Furthermore, by encapsulating the node processing module (parameters: rule number, trigger condition, target parameter, expected effect, confidence weight), the attribute binding of each rule node is realized, transforming the unstructured control logic into a traversable graph node object, and obtaining a complete node attribute table.

[0124] Furthermore, by using a dependency edge generation method (parameters: timestamp order, rule dependency matrix, conflict resolution result), directed edges in the graph are constructed, and edge set data containing execution order constraints is generated to ensure that the execution path satisfies temporal causality and logical dependency constraints.

[0125] Furthermore, a DAG verification algorithm based on topological sorting (parameters: node set, edge set, cycle detection threshold) is adopted to realize the directed acyclicity detection of the constructed graph, generate a sorted node visit sequence, and calculate the level depth index of the node in the graph.

[0126] The DAG serialization module (processing method: merging and encapsulating node attribute tables and ordered edge sets) transforms the results of the previous step into structured policy logic flow graph data, realizing a verifiable and traceable formal intermediate product, and providing consistent input for the compilation of deterministic control instructions.

[0127] For example, in a typical high-altitude travel route, the input conflict-free rule activation sequence contains 15 rule nodes. Rule number R001 corresponds to the trigger condition "altitude ≥ 2500m and duration > 10km", with a target parameter of a 0.5℃ temperature increase and an expected effect of improved air conditioning comfort; its confidence weight is 0.92. Rule number R007 corresponds to the trigger condition "slope > 6% and duration > 2km", with a target parameter of a compressor frequency reduction to 35Hz and an expected effect of preventing condensation and freezing; its confidence weight is 0.95. The DAG construction algorithm first creates each rule as a node object and establishes a parallel hierarchy for R001 and R007 due to their independent timestamps. Then, it identifies in the logical dependency matrix that R010 needs to be executed after R007 is completed, thus generating a directed edge.<R007 → R010> This forms a partially ordered graph structure. When performing directed acyclic detection on the graph, no loops were found. The node access sequence obtained from topological sorting is {R001, R007, R010, R003,...}, where R001 and R007 are at level 0, and R010 is at level 1. The final serialized policy logic flow graph contains all 15 nodes. Each node includes a rule number, trigger condition, target parameter, expected effect, and confidence weight, along with an execution level index, which is used in the subsequent control instruction compilation stage. In this scenario, the system's policy transparency is significantly improved, and the policy execution path remains stable.

[0128] Step S5: Compile the strategy logic flow graph output in S4 into a deterministic control instruction sequence. This instruction sequence includes the target operating parameters of each executing component of the air conditioning system and their effective timestamps. The parameters include compressor frequency, damper opening, fan speed, and operating mode. The instruction sequence is arranged along a timeline and has traceable rule source annotations. Specifically, it includes: S5.1: Based on the policy logic flow graph output in S4, obtain the rule number, trigger condition, target parameter, expected effect and confidence weight contained in each node, and use the graph traversal algorithm to extract all rule nodes that need to be activated in time priority order, and generate a rule execution queue with time sequence mark to build the input basis for subsequent instruction compilation.

[0129] S5.2: According to the HVAC control system execution interface specification, define the standard control variable set for each execution component of the air conditioning system, including compressor frequency, damper opening, fan speed and operating mode, and establish a mapping relationship table between target parameters and actual control quantities to achieve standardized conversion of logical parameters into physically executable signals.

[0130] Based on the HVAC control system execution interface specification, the categories, adjustable parameter ranges, and control signal types of each actuator in the air conditioning system are obtained to form an equipment capability description file, which serves as the data source for defining control variables.

[0131] The interface protocol parsing method (parameters: CAN bus frame structure, Modbus register address mapping) is adopted to parse the signal formats and encoding methods that are acceptable at the physical layer for each execution component, establish control signal templates for compressor, damper, fan and mode switching, and realize the establishment of equipment control path.

[0132] Furthermore, through a standardized parameter abstraction method (parameters: rated frequency range, speed level, opening percentage), the target parameter set in the strategy logic flow diagram is mapped to normalized control variables, including... Indicates compressor frequency, Indicates damper opening degree, Indicates fan speed, This indicates the operating mode to ensure consistency of parameter definitions across different devices.

[0133] Furthermore, an algorithm for constructing a mapping table (parameters: logical parameter - physical signal correspondence) is used to derive the encoded value of the target parameter in the actual control signal based on the engineering value range of the target parameter, such as the compressor frequency. Mapped to PWM duty cycle damper opening Mapped to stepper motor pulse count Fan speed Mapped to voltage level index values.

[0134] A calibration compensation method (parameters: equipment response curve, ambient temperature and humidity coefficient) is used to linearly or nonlinearly correct the physical signals in the mapping table, ensuring that the error between the control signal and the target operating parameters is minimized under specific environmental conditions. This algorithm or processing method transforms the logical parameters from the previous step into standardized physical control quantities that the equipment can directly execute, achieving seamless integration from strategy logic to hardware execution.

[0135] For example, in a vehicle with a rated compressor frequency of Hz, maximum damper opening is In % of RV air conditioning systems, the target compressor frequency is Hz, damper opening is %, fan speed is rpm, operating mode is refrigeration. Compressor frequency. Hz mapping formula Obtain the PWM duty cycle The opening degree of the damper is through × Calculate the number of stepper motor pulses The fan speed, obtained from a table lookup, corresponds to voltage level index 5. After calibration and compensation, the PWM duty cycle was fine-tuned to... To match the compressor efficiency curve under low-temperature conditions, the damper pulse count is maintained. The fan voltage index is fine-tuned to 4 to reduce airflow disturbance in high humidity environments. The final output physical control signal is analyzed by the MCU, which can significantly improve the accuracy and response speed of air conditioning control in actual operation, meeting the dual requirements of comfort and energy saving of the pre-adjustment strategy.

[0136] S5.3: For each rule node in the rule execution queue generated in S5.1, calculate its corresponding target running parameter value by combining it with the mapping relationship table defined in S5.2, and assign the corresponding effective timestamp according to the time prediction model of the travel path to generate the original control instruction entry with accurate timing information.

[0137] S5.4: Sort and merge the original control instruction entries generated in S5.3 according to the time axis, use the time window sliding algorithm to eliminate duplicate or conflicting instruction items, optimize instruction density, and output a structured time-seriesd control instruction package to ensure the stability and real-time performance of the MCU parsing process.

[0138] Based on the original control instruction entries generated in S5.3 and their corresponding target operating parameters and effective timestamp data, a time-axis sorting algorithm (parameters: ascending order, primary key is timestamp, secondary key is execution component ID) is used to achieve the initial time-series arrangement of the control instruction entries.

[0139] Furthermore, a time window sliding algorithm is used (parameters: window width Δt = 2s, step size Δt). step =1s) to realize the temporal proximity detection of adjacent instructions, and perform similar clustering of the instruction set in the window based on the execution unit identifier to obtain the candidate set of duplicate instructions.

[0140] Furthermore, a duplicate instruction resolution strategy (parameter: target operating parameter tolerance threshold ε = 0.5 units) is adopted to compare the parameter differences of instructions within the cluster. When the difference is lower than the threshold, they are merged into a single instruction and the timestamp is updated to a weighted average. When the difference exceeds the threshold, the instruction with the earliest timestamp is retained, and the remaining instructions are marked as conflicting items to be resolved.

[0141] Furthermore, a priority arbitration algorithm (parameters: rule type priority table, trigger condition urgency index) is used for conflicting items to be resolved. While retaining high-priority instructions, the final execution parameter value is calculated through a parameter reconciliation formula:

[0142] in, These are high-priority parameter values. These are low-priority parameter values. The weighting coefficients are calculated based on the rule-based confidence weights.

[0143] Furthermore, the instruction set after conflict resolution is sorted again along the time axis, and an instruction density optimization algorithm (parameter: maximum instruction interval threshold T_max=10s) is used to merge overly dense instruction sequences into time segment instruction blocks, thereby reducing the MCU's parsing burden and improving real-time response stability.

[0144] Through the above-mentioned time window sliding detection, repetition resolution, conflict arbitration and density optimization processing, the original control instruction entries in the previous step are transformed into structured time-sequential control instruction packages, thereby achieving stability and real-time performance of control instructions during MCU parsing.

[0145] For example, in a RV trip that included high-altitude sections and long tunnels, a total of 120 original control command entries were generated, including 45 compressor frequency commands, 35 damper opening commands, and 40 fan speed commands. A sliding detection was performed on the compressor frequency commands with a time window width Δt=2s. Twelve sets of duplicate commands were detected, with an average parameter difference of 0.3Hz per set, all below the tolerance threshold ε=0.5Hz. Therefore, they were merged into a single command, and the timestamp was updated to the average time within the window. Regarding conflicting damper opening commands, a set containing both "fully open" and "closed" commands was detected. Based on the priority table, the "closed" command was determined to be of higher priority (physical constraint rules), and the reconciliation formula was invoked. The final execution parameter value is 20% opening. After density optimization, the compressor frequency command is reduced to 38, the damper opening command to 30, and the fan speed command to 34. The final output time-sequential control command package shows an instruction parsing delay of less than 50ms and a significantly reduced response fluctuation in actual MCU operation tests, meeting the requirements for real-time performance and stability.

[0146] S5.5: Attach traceable rule source annotation information to each control command. The annotation information includes the original rule number, trigger condition snapshot and associated path topology node ID, and encapsulate the complete command sequence into a data frame conforming to the vehicle communication protocol format to support policy auditing and deviation attribution analysis in the subsequent execution feedback process.

[0147] In the S5, the onboard microcontroller controls the air conditioning actuator to perform adjustment actions, including initiating low-frequency preheating before long uphill climbs, increasing the set temperature on high-altitude roads, and closing the evaporator damper before entering long tunnels. Specifically, this includes: S6.1: Obtain the deterministic control command sequence output in S5. The command sequence is a structured control data packet arranged on a time axis. Each data packet contains target operating parameters and their effective timestamps. The target operating parameters include compressor frequency, damper opening, fan speed and operating mode. The effective timestamps are used to mark the absolute trigger time of each control action. The command sequence is encapsulated in a frame format based on the vehicle communication protocol to conform to the CAN bus transmission standard, generating a binary control message stream that can be recognized by the vehicle microcontroller unit (MCU).

[0148] The deterministic control command sequence output from S5 is obtained as the initial input object for chained processing. The command sequence is a structured control data package arranged in chronological order, containing a target operating parameter set and a precise effective timestamp. The target operating parameter set consists of four types of standard control variables: compressor frequency, damper opening, fan speed, and operating mode. It is generated through mapping in the previous steps and ensured to have consistency and traceability through conflict resolution and rule labeling.

[0149] A protocol parsing and field verification method (parameters: vehicle air conditioning control variable specification table, field type definition) is employed to perform item-by-item syntax verification and data field integrity checks on structured control data packets, eliminating null values, illegal codes, or instruction fields with inconsistent units, and outputting the verified control data sequence. Furthermore, a device driver compatibility mapping algorithm (parameters: MCU control register mapping table, actuator driver protocol identifier) ​​is used to achieve bidirectional mapping between logical parameters and physical signal definitions of control variables, obtaining a list of intermediate control signals conforming to the vehicle MCU driver interface.

[0150] Furthermore, a frame format encapsulation algorithm (parameters: CAN 2.0B standard field definition, priority flag bit, data length code) is employed to encapsulate the intermediate control signal list into a complete communication data frame, and an arbitration ID is generated in the frame header area to identify the functional domain, ensuring the accuracy of subsequent command routing on the bus. Furthermore, a cyclic redundancy check (CRC-15) algorithm is used to calculate the integrity check code for each frame's control data payload, and the check code field is appended to the end of the data frame to obtain a complete set of communication frames with error detection capabilities.

[0151] By performing binary serialization, the aforementioned set of communication frames is converted into a continuous binary control message stream, maintaining the instruction timing order and field alignment structure, thus achieving a message format that can be directly parsed by the MCU side.

[0152] For example, in an air conditioning pre-adjustment task on a long uphill section, the deterministic control command sequence output by S5 contains three data packets, the first of which has a target operating parameter of compressor frequency. Hz, damper opening %, fan speed RPM, working mode "warm-up", effective timestamp is Seconds; the second is the compressor frequency. Hz, damper opening %, fan speed rpm, operating mode "energy saving", effective timestamp is Seconds; the third is the compressor frequency. Hz, damper opening %, fan speed rpm, operating mode "constant temperature", effective timestamp is Seconds. During protocol parsing and field validation, a field type definition library is used to perform type matching checks on each control data field, and a device driver compatibility mapping algorithm is used to map frequency parameters to MCU register numbers. The damper opening parameter is mapped to the register number. Fan speed is mapped to register number Operating modes are mapped to register numbers. During the frame format encapsulation stage, the data length is set to 8 bytes according to the CAN 2.0B standard, and the arbitration ID is set to... Then, the CRC-15 checksum algorithm is called to calculate the checksum using the following formula:

[0153] in For the control data stream to be encapsulated, For standard polynomial masks The formula calculation yields a 15-bit checksum. The data packets are appended to the end of each frame to form a complete message. Binary serialization encapsulates the three data packets to generate a 144-bit control message stream, maintaining the original timestamp order. This allows for direct parsing and execution on the MCU side, ensuring that the air conditioning preheating action before a long uphill climb is triggered precisely according to preset conditions.

[0154] S6.2: The binary control message stream generated in S6.1 is transmitted to the vehicle microcontroller unit (MCU) of the air conditioning system through the vehicle communication interface. The MCU performs verification and frame de-framing processing on the received messages, extracts the target operating parameters and effective timestamps from each control data packet, and stores them in the real-time scheduling queue. Based on the system timestamp provided by the real-time clock module, the control tasks in the scheduling queue are time-matched to determine whether the triggering conditions for any control action have been met.

[0155] S6.3: When the MCU detects that the current system time matches the effective timestamp in a certain control data packet, it triggers the corresponding control action execution process; according to the target operating parameters contained in the data packet, it generates the corresponding PWM modulation signal or digital switch signal, and outputs them to the compressor frequency conversion drive module, the damper stepper motor controller and the fan speed control circuit respectively, so as to drive the air conditioning actuator to adjust according to the preset parameters, such as adjusting the compressor frequency to a low frequency operating state 3 minutes before a long uphill climb to achieve preheating.

[0156] The set of control tasks is based on the real-time scheduling queue in S6.2, and the input conditions are the verified target running parameters and their absolute effective timestamp.

[0157] The system time matching algorithm (parameters: real-time clock timestamp, control task timestamp) is used to realize the control task triggering and judgment function. When the comparison result meets the precise matching condition, the execution process is initiated.

[0158] Furthermore, by using parameter parsing methods (parameters: compressor frequency setpoint, damper opening level, fan speed target value, and operating mode identifier), the logical operating parameters are mapped to a set of low-level drive signal control variables, and three types of core execution data are obtained: compressor control quantity, damper position command quantity, and fan PWM duty cycle.

[0159] Furthermore, a PWM modulation signal generation algorithm (parameters: target frequency, target duty cycle, carrier period) is adopted to achieve accurate synthesis of the input signal of the compressor frequency converter drive module, and a digital switch signal generation method (parameters: position encoding, number of step control pulses) is used to generate the closing and opening adjustment commands of the damper stepper motor controller.

[0160] Furthermore, based on the control logic of the fan speed regulation circuit, the output generation and processing of PWM signal or voltage modulation (parameters: target speed, speed regulation step size) are executed to achieve precise setting of fan speed and generate the expected value of fan speed feedback as a reference for subsequent closed-loop correction.

[0161] The above-mentioned multi-channel signal output algorithm transforms the parameter parsing results of the previous step into a set of physical layer drive signals, thereby achieving the predetermined technical effect that each actuator of the air conditioning system performs actions according to preset operating parameters at the target time.

[0162] For example, in a scenario where the altitude rises to 2800m and the path prediction model determines that a long uphill climb will occur in 3 minutes, the target compressor in the control data packet operates at a low frequency of 35Hz, with the damper opening at 0° (fully closed), the fan speed at 800rpm, and the operating mode in preheating mode. The MCU matches the control task with an effective timestamp of 10:32:15 when the system time is read from the real-time clock module. Parameter parsing is then performed to obtain the compressor control input of 35Hz, the damper position code of 0°, and the fan PWM duty cycle of 28%. For the low-frequency operating parameters of the compressor, the PWM modulation formula is used:

[0163] in For output frequency, The carrier frequency is 500Hz. With a frequency division factor of 14, the calculated output frequency is approximately 35Hz, and the modulation signal output is completed through the frequency converter drive module. The damper stepper motor control command uses digital pulse generation, generating a 0-step pulse output at a step angle of 1.8° / step to maintain a 0° closed state. The fan speed control uses a linear mapping relationship formula between the target speed and the PWM duty cycle:

[0164] in For PWM duty cycle, The target speed is 800 rpm. With a maximum speed of 2800 rpm, the calculated duty cycle is approximately 0.285, which is highly consistent with the analytical result of 28%. The performance results show that the compressor begins low-frequency preheating operation 3 minutes before entering a long uphill section, effectively suppressing the risk of condensate freezing; the damper is closed to prevent the entry of high-humidity air; the fan maintains a low to medium speed to reduce airflow disturbance, and the temperature and humidity inside the vehicle remain stable before entering the tunnel and during the uphill climb, achieving a dual improvement in comfort and energy consumption control.

[0165] S6.4: During the execution of control actions, the MCU continuously monitors the actual response status of each execution component, collects physical quantities such as the actual operating frequency of the compressor, the actual opening angle of the damper, and the actual speed of the fan through feedback sensors, and compares the measured values ​​with the target operating parameters to form a closed-loop error signal; the error signal is dynamically compensated based on the PID adjustment algorithm to generate a corrected drive signal to eliminate the execution deviation caused by mechanical lag or power fluctuations, and ensure that the actual operating state of the air conditioning system is close to the target parameters set by the instruction sequence.

[0166] The input conditions are that after the vehicle-mounted microcontroller unit (MCU) triggers the control action execution process, it has output the corresponding PWM modulation signal or digital switch signal to the compressor frequency conversion drive module, the damper stepper motor controller and the fan speed control circuit, and the system's built-in feedback sensor has been continuously transmitting physical quantities such as the actual operating frequency of the compressor, the actual opening angle of the damper and the actual speed of the fan in the bus network.

[0167] A multi-channel signal acquisition method (parameters: sampling frequency 1kHz, resolution 12bit) is used to realize the compressor operating frequency. <f>damper opening <θ> and fan speed <v>The synchronous acquisition function generates a set of feedback vectors with a unified timestamp.

[0168] Furthermore, through a real-time comparison algorithm (parameters: target parameters are from the instruction sequence in S5, matching window ±50ms), the feedback vector set and the target operating parameter set are compared point by point, and an instantaneous error vector is obtained. The error components include frequency deviation <Δf>, opening deviation <Δθ>, and speed deviation <Δv>.

[0169] Furthermore, a proportional-integral-derivative (PID) control algorithm (parameters: Kp, Ki, Kd obtained by fitting the HVAC system response curve) is used to calculate the closed-loop compensation E(t), the formula of which is:

[0170] in, For the instantaneous error of the corresponding control variable, It is the integral variable.

[0171] Furthermore, based on the above E(t) results, a drive signal correction method is implemented (parameters: compressor drive module PWM duty cycle adjustment resolution 0.1%, damper stepper motor micro-step adjustment angle 0.05°, fan speed control circuit voltage adjustment accuracy 0.01V) to generate a new drive output quantity, which is then applied to the actuator in real time through the MCU's high-speed output channel to achieve dynamic compensation for deviations caused by mechanical lag and power fluctuations.

[0172] By using a closed-loop PID control algorithm and drive signal correction processing method, the feedback error signal from the previous step is transformed into a precise compensation drive command, thereby achieving long-term stable operation of the air conditioning system and closely matching the target parameters set by the command sequence.

[0173] For example, in a scenario where a single travel path includes long uphill sections and continuous tunnel segments, the MCU collects the compressor's actual operating frequency as 48Hz, while the target frequency is 50Hz, resulting in a frequency deviation. Hz; The actual opening of the damper is 32°, the target opening is 30°, and the opening deviation is... °; The actual fan speed is 1350 rpm, the target speed is 1400 rpm, and the speed deviation is... The error value was input into the PID algorithm, with Kp=0.8, Ki=0.15, and Kd=0.05. The compensation amounts for each actuator were calculated, including a 1.6% increase in the compressor PWM duty cycle, a 0.25° increase in the damper stepper motor closure, and a 0.5V increase in the fan voltage. After correction, the compressor operating frequency stabilized at 50Hz, the damper opening was accurate to 30°, and the fan speed remained at 1400rpm. The system maintained a stable and comfortable environment throughout the long uphill preheating and tunnel humidity isolation phases, verifying the effectiveness and accuracy improvement of dynamic compensation.

[0174] S6.5: Record the start and end times, target parameters, measured response curves, and PID compensation values ​​for each control action in S6.4, and generate an execution process log with timestamps. Write this log, along with the original instruction source (i.e., the rule number and strategy logic flow node ID in S5), into the execution feedback database in the local trusted execution environment (TEE) as the data basis for subsequent strategy effect evaluation and rule knowledge base iterative optimization, thereby supporting the traceability and adaptability of the entire pre-adjustment system.

[0175] The method further includes the following step after S5: Step S7: During the air conditioning operation, collect in-vehicle temperature and humidity, user manual adjustment behavior, and actual energy consumption data. Compare the data with the expected effect of the instruction sequence in S5 to generate a strategy execution deviation log. Encrypt the log along with the original rule matching path and manual intervention traces and store them in a local trusted execution environment (TEE). Specifically, this includes: S7.1: Real-time monitoring data during the air conditioning process is acquired based on the vehicle-mounted multi-source sensor network. The data includes in-vehicle environmental parameters output by temperature and humidity sensors, user manual adjustment behavior logs recorded by the human-machine interface, and actual power consumption sequences collected by the power metering module. The heterogeneous data streams are timestamped and normalized to generate a structured real-time operating status snapshot sequence.

[0176] S7.2: Based on the deterministic control command sequence generated in S5, extract the target operating parameters and their effective timestamps for each time period, construct the expected effect model, map the real-time operating status snapshot generated in S7.1 to the corresponding command interval according to the time axis, calculate the integral value of the deviation between the actual temperature and humidity response curve and the theoretical set trajectory, and identify the time point and adjustment range of user manual intervention to quantify the degree of strategy deviation.

[0177] Based on the deterministic control command sequence generated in S5, the target operating parameters and their effective timestamps for each time period are extracted. A time axis mapping method (parameters: command sequence timestamp, control period Δt) is used to establish a one-to-one correspondence between theoretical control actions and time intervals. Furthermore, through the interval indexing method (parameters: start and end timestamps Tk_start, Tk_end), the real-time operating status snapshots generated in S7.1 are divided into corresponding command effective intervals according to the time sequence, and the temperature and humidity response datasets for each interval are obtained.

[0178] A spline interpolation algorithm (parameters: cubic spline, node interval Δt) is used to smoothly reconstruct the temperature and humidity response dataset for each interval, obtaining continuous actual temperature and humidity response curves. Furthermore, the mean square error calculation method (parameter: number of sampling points n within the time window) is used to calculate the integral value of the deviation between the actual curve and the theoretically set trajectory.

[0179] in, This represents the actual temperature value at the i-th sampling point within the time window. Set the theoretical temperature value.

[0180] An event detection algorithm (parameters: snapshot sequence, adjustment parameter change threshold δ) is used to identify the time points of user manual intervention and obtain a list of timestamps for each intervention. Furthermore, an amplitude quantization method (parameter: difference in measurement values ​​before and after adjustment ΔP) is used to calculate the amplitude of each manual intervention and generate an intervention intensity index matrix.

[0181] By using an index fusion processing method, the deviation integral value and intervention intensity matrix are transformed into strategy deviation data, thereby quantifying the difference between the target parameters and the actual response during the air conditioning execution process.

[0182] For example, during a motorhome trip to a high-altitude area, the target temperature for a certain high-altitude section in the deterministic control command sequence is set to... ℃, effective time window is from 09:20:00 to 09:50:00, sampling period is The snapshot sequence output by the temperature and humidity sensor, after being timestamped, corresponds to the actual temperature curve for the effective period, which gradually decreases from 09:20:00 to [the point where the temperature curve gradually decreases]. ℃ rises again to ℃. The mean square error calculated after smoothing and reconstructing using cubic spline interpolation is: The user manually increased the temperature at 09:33:10. ℃, this entry in the intervention intensity matrix is ​​denoted as The strategy deviation is calculated by integrating the integral value of the deviation over this interval with the intervention intensity. The output is then fed into the subsequent deviation type label generation stage to quantitatively characterize the difference between the strategy and the actual execution effect in this scenario.

[0183] S7.3: Based on the policy execution deviation integral value and user intervention intensity index calculated in S7.2, and combined with the rule number, triggering condition and confidence weight in the original rule activation chain, generate policy execution deviation log entries containing deviation type labels (such as 'temperature lag', 'unsuitable wind speed', 'excessive energy consumption'). Each log entry is associated with the corresponding rule action fragment and environmental context feature vector.

[0184] S7.4: The policy execution deviation log entries generated in S7.3 are associated and encapsulated with the rule matching paths in the policy logic flow diagram output in S4 and the human intervention traces recorded in S6 across stages to form a closed-loop audit data package with causal traceability. The data package is organized in JSON-LD format to ensure semantic parsing and field extensibility.

[0185] S7.5: Input the closed-loop audit data packet constructed in S7.4 into the Trusted Execution Environment (TEE) in the vehicle security chip, encrypt and store it using the AES-256-GCM algorithm, and establish a log integrity verification mechanism based on hash chain to prevent data tampering, so as to ensure the authenticity and legal compliance of the model correction basis during subsequent rule knowledge base iteration.

[0186] Following S7, step S8 is also included: based on the policy execution deviation log stored in S7, the interpretability verification and parameter correction of the hierarchical rule knowledge base defined in S2 are performed periodically, rule entries with confidence weights below a threshold are identified, and their triggering conditions or effect strength are optimized in combination with historical verification results to achieve closed-loop iterative updates of the rule graph. Specifically, this includes: S8.1: Obtain the policy execution deviation log stored in the local trusted execution environment (TEE). The log contains the actual temperature and humidity data collected during each air conditioning adjustment process, user manual intervention behavior, comparison results between actual energy consumption value and expected effect of instruction sequence, and extract the application instance and deviation measurement index of each rule in the specific travel scenario to construct a rule verification sample set.

[0187] S8.2: Based on the constructed rule verification sample set, perform interpretability verification on each rule in the hierarchical rule knowledge base defined in S2. Use a logical consistency-based evaluation model to calculate the output stability and prediction accuracy of each rule in a multi-journey scenario, and generate a rule confidence weight vector, where the confidence weight is a weighted statistical value of the expected effect achievement rate in the historical execution of the rule.

[0188] The input object is a rule verification sample set constructed by S8.1, which contains rule application instances from multiple historical travel scenarios and their policy deviation metrics under different environmental parameters.

[0189] An evaluation model based on logical consistency (parameters: rule matching path, deviation type label, and environmental context feature vector) is adopted to calculate the stability of each rule output. The stability is obtained by solving the consistency matrix of the rule's output results in a multi-journey scenario.

[0190] Furthermore, a statistical inference-based accuracy calculation method (parameters: expected effect achievement flag and actual achievement flag) is used to quantitatively evaluate the rule prediction accuracy and obtain the accuracy value of each rule on all validation samples; the accuracy is expressed as follows:

[0191] in, The actual effect achieved during the i-th rule execution is the state. This indicates that the expected results of this rule have been achieved. This represents the total number of samples used in the validation.

[0192] Furthermore, a weighted statistical method (parameters: rule stability index, rule accuracy index, sample environment weight coefficient) is used to comprehensively calculate the rule confidence weight. The formula for calculating the j-th component of the weight vector is:

[0193] in, Let j be the output stability index of the j-th rule. Let j be the prediction accuracy metric for the j-th rule. , The weighting coefficient is determined by the ratio of the importance of user comfort to the importance of energy consumption control.

[0194] Through the above weighted calculation, the stability and accuracy results of the previous step are transformed into a rule confidence weight vector, thereby realizing a comprehensive interpretability evaluation of the rule's execution effect in multiple scenarios.

[0195] For example, in a plain-plateau hybrid route verification scenario, the rule verification sample set contains 120 application instances of rules related to physical constraints, comfort compensation, and energy conservation. The stability index of a particular rule in the plateau and mountain sections is also considered. for Prediction accuracy for Weighting coefficient settings , Substituting into the weighted formula:

[0196] Calculated This indicates that the rule has a high overall confidence level in this hybrid route scenario, and its core triggering conditions can be retained in subsequent iterations, with optimization only applied to energy conservation details to maintain a balance between interpretability and performance.

[0197] S8.3: Identify inefficient rule entries with confidence weights below a preset threshold, mark such rules as objects to be optimized, and associate them with their original construction basis (such as ASHRAE standard clause number, measured thermal response experiment ID, or user preference confirmation record) to form a rule defect traceability list, which serves as the input basis for subsequent parameter correction.

[0198] S8.4: For the marked rule entries to be optimized, combined with the verification results in multiple historical trips and relevant environmental variables (such as altitude change gradient, external temperature and humidity fluctuation range, battery SOC state), the incremental linear regression algorithm is used to dynamically adjust the trigger condition threshold or the intensity of the action parameter, generate candidate correction schemes, and retain the original rule version for A / B test comparison.

[0199] S8.5: Embed the generated candidate correction scheme into the rule graph engine, perform backtracking simulation verification on the simulated travel path, compare the comfort gain and energy consumption cost of the policy logic flow graph before and after correction, if the new version improves the comfort compliance rate by more than a set proportion without being lower than the energy-saving performance of the original version, then update the rule to the hierarchical rule knowledge base and update its confidence weight calculation benchmark simultaneously to complete the closed-loop iterative update of the rule graph.

[0200] For those skilled in the art, various other corresponding changes and modifications can be made based on the technical solutions and concepts described above, and all such changes and modifications should fall within the protection scope of the claims of this invention.

[0201] Unless otherwise defined, the technical or scientific terms used herein shall have the ordinary meaning as understood by one of ordinary skill in the art to which this application pertains. The terms "first," "second," "third," and similar terms used in this patent application specification and claims do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, "an" or "a" and similar terms do not indicate a quantity limitation, but rather indicate the presence of at least one. The terms "comprising" or "including" and similar terms mean that the elements or objects preceding "comprising" or "including" encompass the elements or objects listed following "comprising" or "including" and their equivalents, and do not exclude other elements or objects. The "multiple" involved in the embodiments of this application refers to two or more. A and / or B indicate three possibilities: A; B; and A and B.

[0202] The above description is merely an exemplary embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and such modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.< / v> < / f>

Claims

1. A method for intelligent adjustment and control of RV air conditioning integrating trip planning, specifically including: S1: Obtain structured navigation data of the RV's travel route, and perform topology enhancement processing on the structured navigation data to generate a three-dimensional path topology map containing nodes, edges, and faces; S2: Based on the principles of HVAC engineering and measured thermal response data, a hierarchical rule knowledge base is constructed; S3: Input the three-dimensional path topology graph into the rule graph engine, traverse the hierarchical rule knowledge base for each node in the three-dimensional path topology graph, match rule entries that meet the triggering conditions, and generate an initial rule activation set with execution priority markers; S4: Perform temporal conflict resolution and logical consistency verification on the initial rule activation set, use a directed acyclic graph to model the dependency relationship between each rule node, and output a structured strategy logic flow graph; S5: Compile the strategy logic flow graph into a deterministic control instruction sequence. The deterministic control instruction sequence includes the target operating parameters of each actuator of the air conditioning system and their effective timestamps. The parameters include compressor frequency, damper opening, fan speed and operating mode. The deterministic control instruction sequence is used by the vehicle microcontroller unit to control the execution adjustment of the air conditioning actuator.

2. The RV air conditioning intelligent adjustment and control method integrating trip planning according to claim 1, characterized in that, In the S5, the onboard microcontroller controls the air conditioning actuator to perform adjustment actions, including starting low-frequency preheating before long uphill climbs, increasing the set temperature on plateau roads, and closing the evaporator damper before entering long tunnels.

3. The intelligent adjustment and control method for RV air conditioning based on integrated travel planning according to claim 1, characterized in that, After step S5, the method further includes: S7: During the air conditioning process, collect data on in-vehicle temperature and humidity, user manual adjustment behavior, and actual energy consumption. Compare the data with the expected effect of the deterministic control command sequence to generate a strategy execution deviation log. Encrypt and store the log, along with the original rule matching path and manual intervention traces, in the local trusted execution environment.

4. The intelligent adjustment and control method for RV air conditioning based on integrated travel planning according to claim 1, characterized in that, After step S7, the method further includes: S8: Based on the policy execution deviation log stored in S7, the interpretability verification and parameter correction of the hierarchical rule knowledge base are performed periodically, rule entries with confidence weights below the threshold are identified, and their triggering conditions or effect strength are optimized in combination with historical verification results to achieve closed-loop iterative update of the rule graph.

5. The intelligent adjustment and control method for RV air conditioning based on integrated travel planning according to claim 1, characterized in that, In step S1, the structured navigation data includes the starting and ending geographic coordinates, length, elevation change, radius of curvature, road surface type, and ancillary facility labels for each road segment.

6. The intelligent adjustment and control method for RV air conditioning based on integrated travel planning according to claim 1, characterized in that, Step S1 specifically includes: S1.1: Obtain structured navigation data of the RV's travel route, and perform a unified projection transformation on the coordinate data based on the global positioning system standard coordinate system to obtain a standardized spatial trajectory sequence; S1.2: Perform segmented discretization processing on the standardized spatial trajectory sequence obtained in S1.1, identify key terrain turning points based on road attribute change points and ancillary facility boundaries; define each key terrain turning point as a node in the three-dimensional path topology map, forming path event anchor points with clear physical meaning; S1.3: Based on the road segments between adjacent nodes identified in S1.2, construct edges with attribute vectors. The attribute vectors include the average slope of the road segment, cumulative elevation change, average radius of curvature, driving length, estimated travel time, and estimated road surface friction coefficient. Use a GIS engine to interpolate and smooth the original navigation data, calculate the differential geometric feature parameters of each road segment, and generate a structured edge attribute matrix. S1.4: Based on the edge attribute matrix generated in S1.3, further identify the regional-level environmental feature areas formed by multiple consecutive edges and the dense closed channel areas; abstract the above areas into surface units in the three-dimensional path topology map, and label each surface with its environmental category and the time window range it covers, forming a macro-environmental perception layer. S1.5: Integrate the nodes generated in S1.2, the edge attribute vectors constructed in S1.3, and the face units divided in S1.4, and use the GIS engine to perform topology enhancement modeling to establish hierarchical relationships between nodes, edges, and faces; generate a 3D path topology map containing spatial location, geometric features, semantic labels, and time series indexes.

7. The RV air conditioning intelligent adjustment and control method integrating trip planning according to claim 1, characterized in that, Step S2 specifically includes: S2.1: Based on the ASHRAE HVAC design standard and the measured thermodynamic response dataset of RVs, key variable pairs such as slope-compressor start-stop sequence, altitude-perceived temperature offset, and curvature-airflow disturbance intensity are extracted. Correlation analysis and threshold segmentation algorithms are used to determine the critical action range between each variable, and a preliminary set of physical constraint rules with clear trigger boundaries is generated. S2.2: Perform engineering feasibility verification on the initial set of physical constraint rules generated in S2.1, and form a physical constraint rule sub-library with physical implementability based on the preset action boundary of the actuator; S2.3: Based on user comfort survey data and physiological perception experiment reports from three typical regions—plateau, mountain, and plain—the equivalent temperature compensation value corresponding to each preset value of altitude increase is calculated using the linear regression modeling method. Combined with the fitting relationship between the average curvature of continuous curves and the airflow discomfort scores reported by users, a comfort compensation rule sub-library is constructed with "altitude increase" and "number of continuous curves" as inputs. S2.4: Based on the vehicle dynamics model and braking energy recovery characteristic curve, analyze the energy distribution pattern of long uphill and long downhill sections in the subsequent journey, and use the power balance equation to calculate the net energy consumption budget under different terrain combinations; based on the budget value, execute the dynamic programming algorithm, set the upper limit ratio of cooling power and the timing of cooling energy regeneration during the coasting stage, and generate a sub-library of energy conservation rules. S2.5: The physical constraint rule sub-library, comfort compensation rule sub-library, and energy conservation rule sub-library constructed in S2.2, S2.3, and S2.4 respectively are standardized in format to generate a hierarchical rule knowledge base.

8. The intelligent adjustment and control method for RV air conditioning based on integrated travel planning according to claim 1, characterized in that, Step S3 specifically includes: S3.1: Based on the set of nodes in the 3D path topology map output by S1, extract the topological attribute vector associated with each node, and use the GIS spatial interpolation algorithm to reconstruct the continuous terrain features between discrete nodes to generate a physically meaningful local terrain feature descriptor. S3.2: The local terrain feature descriptors generated in S3.1 are matched one by one with the condition-action statements in the hierarchical rule knowledge base constructed in S2. The predicate logic-based rule parsing engine is used to determine the truth value of the preconditions of each rule. When all the input conditions of the preset rule are satisfied by the feature descriptor of the current node, the rule is determined to be triggered, and the rule number corresponding to the rule and the set of binding action parameters are output as the preliminary matching result. S3.3: Mark the timing attributes of all triggered rule entries in S3.2, calculate the effective time window of each rule's corresponding action based on the travel direction and node sequence position, and calculate the target execution timestamp of each rule action by combining the advance parameter defined by the rule itself. Then, attach the timestamp as a timing requirement to the rule record to form a rule activation candidate set with timing constraints. S3.4: Based on the priority levels of each rule in the knowledge base as preset in S2 and the timestamp order marked in S3.3, the rule activation candidate set is sorted by priority. A multidimensional sorting algorithm is used to sort the rules in ascending order with priority as the primary key and timestamp as the secondary key, generating an initial rule activation set with execution priority markers. S3.5: Format and encapsulate the initial rule activation set output by S3.4, transmit it to the conflict resolution module in the next stage, and cache a copy locally for audit tracing.

9. The intelligent adjustment and control method for RV air conditioning based on integrated travel planning according to claim 1, characterized in that, Step S4 specifically includes: S4.1: Based on the initial rule activation set output in S3, obtain each activated rule entry and its associated attributes, as well as the confidence weight label from the hierarchical rule knowledge base; S4.2: Perform timeline mapping processing on the execution timing requirements of the rule entries obtained in S4.

1. Utilize the geographic coordinates of nodes in the travel path topology map and the vehicle's expected driving speed model to calculate the absolute effective timestamp interval of each rule action and generate a rule execution time period sequence with time window markers. S4.3: Based on the rule execution time sequence generated in S4.2, execute the multi-rule concurrent conflict detection algorithm. For rule pairs that act on the same air conditioning actuator within the same or overlapping time window but have inconsistent set parameters, identify them as potential action contradictions and output a list of conflict rule pairs and their related control variables and time overlap index. S4.4: The conflict rules identified in S4.3 are processed by the priority arbitration mechanism. Based on the priority order of the rule types preset in the hierarchical rule knowledge base, the actions of high-priority rules are retained, and the low-priority rules are subject to delayed execution, partial disabling, or parameter reconciliation operations to generate a modified conflict-free rule activation sequence. S4.5: Input the conflict-free rule activation sequence generated in S4.4 into the directed acyclic graph construction module. With each rule as a graph node and the time sequence and logical dependency relationship as directed edges, construct a structured strategy logic flow graph, and finally output a DAG structure that satisfies the topological sorting condition.

10. The intelligent adjustment and control method for RV air conditioning based on integrated travel planning according to claim 1, characterized in that, Step S5 specifically includes: S5.1: Based on the strategy logic flow graph output in S4, obtain the rule number, trigger condition, target parameter, expected effect and confidence weight contained in each node, and use the graph traversal algorithm to extract all rule nodes to be activated in time priority order to generate a rule execution queue with time sequence mark. S5.2: According to the HVAC control system execution interface specification, define the standard control variable set for each execution component of the air conditioning system, and establish a mapping relationship table between target parameters and actual control quantities; S5.3: For each rule node in the rule execution queue generated in S5.1, calculate its corresponding target running parameter value by combining the mapping relationship table defined in S5.2, and assign the corresponding effective timestamp according to the time prediction model of the travel path to generate the original control instruction entry with accurate timing information. S5.4: Sort and merge the original control instruction entries generated in S5.3 according to the time axis, use the time window sliding algorithm to eliminate duplicate or conflicting instruction items, optimize instruction density, and output a structured time-seriesd control instruction package. S5.5: Attach traceable rule source labeling information to each control command and encapsulate the complete command sequence into a data frame conforming to the vehicle communication protocol format.

Images