Vehicle energy-saving navigation method and system based on historical behavior learning
By using historical behavior learning methods to predict intersection waiting probabilities using historical vehicle driving data, personalized driving guidance is provided, solving the problem that existing navigation systems cannot accurately predict traffic lights, and achieving energy saving and improved comfort.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-23
- Publication Date
- 2026-04-10
AI Technical Summary
Existing in-vehicle navigation systems cannot predict traffic lights differently based on individual drivers' habits and frequently used routes, making it impossible for drivers to accurately anticipate whether they will need to wait at the next intersection and failing to alleviate range anxiety.
By acquiring historical vehicle driving data, a personalized intersection passage memory database is established. Historical behavior learning methods are used to predict the probability of waiting at the next intersection and provide predictive driving guidance, including generating tiered driving guidance instructions.
It achieves significant energy savings, alleviates range anxiety, improves driving comfort and safety, provides highly customizable driving strategies, and reduces decision-making confusion caused by uncertainty.
Smart Images
Figure CN121829587A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of vehicle control technology, specifically a vehicle energy-saving navigation method and system based on historical behavior learning. Background Technology
[0002] The statements in this section merely refer to the background technology related to this invention and do not necessarily constitute prior art.
[0003] Current in-vehicle navigation systems have traffic signal prediction capabilities. They can predict traffic light changes at the next intersection based on the vehicle's current location by calculating the speed changes of a large number of vehicles passing through the intersection; or, with authorization, they can directly access officially released traffic light change information.
[0004] Navigation systems with traffic light prediction functions receive the same information from all vehicles, making it impossible to differentiate predictions based on individual driving habits and frequently used routes. This prevents electric vehicle drivers from developing a stable expectation of whether they will need to wait at the next intersection, and fails to truly alleviate range anxiety caused by uncertainty. Summary of the Invention
[0005] This invention provides a vehicle energy-saving navigation method and system based on historical behavior learning. By utilizing historical vehicle driving data, it establishes a personalized "intersection passage memory bank" for car owners. By accurately predicting the probability of waiting at the next intersection and providing predictive driving guidance, it achieves significant energy savings, effectively alleviates range anxiety, and improves driving comfort.
[0006] The first aspect of this invention discloses a vehicle energy-saving navigation method based on historical behavior learning, comprising the following steps: Obtain the travel data of the target vehicle within a historical time period. The travel data includes geographical location sequence, speed sequence, and corresponding timestamps. Map matching is performed on the trip data to generate a historical route sequence composed of standard road segment identifiers; Based on the frequency of occurrence of historical path sequences, one or more high-frequency travel paths of the target vehicle are identified. On high-frequency traffic paths, trajectory points that meet preset low-speed conditions are identified based on speed sequences, and these trajectory points are clustered and verified as intersection waiting points. For each intersection waiting point, based on its historical passage records at different time segments, a time-dependent waiting probability model is statistically analyzed and established. When the target vehicle is traveling in real time and is matched with a high-frequency traffic route, the waiting probability model of the waiting point at the next intersection is queried based on the current location, current time and vehicle speed. Based on the waiting probability value obtained from the query, the graded driving guidance instructions are generated and output to the vehicle's human-machine interface.
[0007] Furthermore, one or more high-frequency travel routes of the target vehicle are identified. Specifically, the historical routes of the vehicle are periodically read from the historical route sequence database; aggregating and analyzing routes with the same starting and ending points and repeated occurrences within a preset long-term window to obtain a list of commonly used routes; and determining high-frequency travel routes by calculating the frequency of occurrence of routes or key road segments within the long-term window.
[0008] Furthermore, the trajectory points are clustered and verified as intersection waiting points, specifically as follows: On a high-frequency path, find trajectory points where the vehicle speed is below a first speed threshold; then, spatially cluster the obtained low-speed trajectory points according to their geographical location and path order to form candidate intersection points. Extract multiple trip data of the target vehicle on different dates to verify whether the low-speed waiting conditions are met at the candidate intersection points; mark the candidate intersection points that pass the multi-trip verification as intersection waiting points.
[0009] Furthermore, a time-dependent waiting probability model is established, specifically as follows: For each intersection waiting point, traffic records at different time segments are extracted from historical data; the proportion of times vehicles wait at the intersection waiting point in each time segment is calculated to generate the waiting probability corresponding to the time segment; the waiting behavior is determined by the vehicle speed being lower than the second speed threshold and the position not changing significantly in multiple consecutive data frames.
[0010] Furthermore, generate and output graded driving guidance instructions, specifically by comparing the obtained waiting probability value with at least two preset probability thresholds, and triggering different levels of guidance instructions based on the comparison results; The boot instructions include at least the following: When the probability of waiting is higher than the first probability threshold, the "slow down" command is triggered. When the waiting probability is between the first probability threshold and the second probability threshold, a "rational planning" instruction containing dynamic speed suggestions is triggered.
[0011] Furthermore, the dynamic speed recommendations in the "rational planning" instruction are generated in the following way: Based on the historical passage records of the intersection waiting point, the minimum historical vehicle speed that can successfully pass through in the corresponding time segment is analyzed and used as the dynamic speed threshold. The real-time vehicle speed is compared with the dynamic speed threshold. If the real-time vehicle speed is greater than or equal to the dynamic speed threshold, the guidance of "maintain current speed" or "suggest acceleration" is generated; otherwise, the guidance of "suggest deceleration" is generated.
[0012] Furthermore, it also includes a continuous update step for the model and thresholds, specifically: The vehicle will transmit traffic data, including real-time vehicle speed, passage results, and environmental scene information, back to the cloud. The cloud grouped the transmitted data by intersection, time period and scenario, and used the new data to incrementally update the waiting probability model and dynamic speed threshold; the updated model and threshold were sent to the vehicle for subsequent real-time guidance.
[0013] A second aspect of the present invention discloses a vehicle energy-saving navigation system based on historical behavior learning, comprising: The data acquisition module is configured to acquire the travel data of the target vehicle within a historical time period. The travel data includes a geographical location sequence, a speed sequence, and the corresponding timestamps. The high-frequency path mining module is configured to: perform map matching on the travel data and generate a historical path sequence composed of standard road segment identifiers; The high-frequency path mining module is also configured to: mine one or more high-frequency travel paths of the target vehicle based on the frequency of occurrence of historical path sequences; The intersection waiting module is configured to: identify trajectory points that meet preset low-speed conditions based on speed sequences on high-frequency traffic paths, cluster the trajectory points and verify them as intersection waiting points; The intersection waiting module is also configured to: for each intersection waiting point, based on its historical passage records in different time segments, statistically analyze and establish a time-dependent waiting probability model; The intersection waiting module is also configured to: when the target vehicle is traveling in real time and is matched with a high-frequency traffic path, query the waiting probability model of the next intersection waiting point based on the current location, current time and vehicle speed; The navigation instruction output module is configured to generate and output graded driving guidance instructions to the vehicle's human-machine interface based on the waiting probability value obtained from the query.
[0014] A third aspect of the present invention discloses a computer program product including computer-readable instructions, which, when executed on an electronic device, enable the electronic device to implement the above-described vehicle energy-saving navigation method based on historical behavior learning.
[0015] A fourth aspect of the present invention discloses an electronic device, including at least one processor and a memory connected to the processor, the memory being used to store a computer program; the processor being used to execute the computer program, enabling the electronic device to implement the above-described vehicle energy-saving navigation method based on historical behavior learning.
[0016] Compared with existing technologies, one or more of the above technical solutions have the following beneficial effects: 1. By deeply mining and analyzing the historical driving data of individual vehicles, a personalized intersection traffic knowledge base and predictive model are constructed, upgrading navigation from "general information broadcasting" to "personalized driving strategy guidance." Compared to current navigation systems that are only based on real-time traffic conditions or fixed signal cycles, this system can provide drivers with highly customized and predictive traffic suggestions, thereby producing significant beneficial effects in multiple dimensions such as energy saving, alleviating range anxiety, and improving driving comfort and safety.
[0017] 2. By mining drivers' unique "high-frequency travel routes," the analysis focuses on their real, repetitive travel scenarios, avoiding the neglect of individual differences by general models. Furthermore, rigorous validation of "intersection waiting points" based on individual historical data accurately identifies the driver's actual, fixed signal control points (such as traffic lights), rather than temporary stopping points, ensuring the reliability of the predictions. This makes drivers' expectations of road conditions ahead highly stable and reliable, resolving the decision-making dilemma of "should I accelerate or coast" caused by inaccurate navigation predictions.
[0018] 3. The established "time-dependent waiting probability model" can inform drivers of the likelihood of waiting at a specific intersection at a specific time, providing richer information dimensions. It transforms the abstract information of "possibly waiting at a red light" into specific, actionable "speed-action" guidance instructions, promoting energy-efficient driving behavior. By classifying probabilities and providing "dynamic speed suggestions" in low-to-medium probability scenarios, it avoids unnecessary energy consumption. For example, when a "high probability of waiting" is predicted, it advises drivers to smoothly decelerate and coast, avoiding sudden braking and idling, thus reducing the waste of kinetic energy converted into heat.
[0019] 4. By providing clear action guidelines (such as "Please maintain a speed of xx km / h or higher to pass"), drivers shift from passively accepting "There are still XX kilometers of range left" to actively implementing "Driving according to this strategy can save energy." This partially transforms the uncertainty of "remaining range" into the certainty of "crossing the intersection ahead," alleviating range anxiety on both psychological and practical levels. Attached Figure Description
[0020] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.
[0021] Figure 1 This is a schematic diagram of a vehicle energy-saving navigation process based on historical behavior learning, provided for one or more embodiments of the present invention. Detailed Implementation
[0022] The present invention will be further described below with reference to the accompanying drawings and embodiments.
[0023] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.
[0024] Therefore, this solution provides a vehicle energy-saving navigation method and system based on historical behavior learning. Without relying on external hardware, the system analyzes the historical GPS and speed data of a single vehicle to automatically learn its high-frequency paths and intersection waiting patterns, establishes a time-sensitive personalized prediction model, and matches it in real time during driving, giving operation suggestions such as "slow down and coast" or "maintain speed" through voice or light, forming a data-driven energy-saving driving closed loop.
[0025] like Figure 1 As shown, the vehicle energy-saving navigation method based on historical behavior learning includes the following steps: Obtain the travel data of the target vehicle within a historical time period. The travel data includes geographical location sequence, speed sequence, and corresponding timestamps. Map matching is performed on the trip data to generate a historical route sequence composed of standard road segment identifiers; Based on the frequency of occurrence of historical path sequences, one or more high-frequency travel paths of the target vehicle are identified. On high-frequency traffic paths, trajectory points that meet preset low-speed conditions are identified based on speed sequences, and these trajectory points are clustered and verified as intersection waiting points. For each intersection waiting point, based on its historical passage records at different time segments, a time-dependent waiting probability model is statistically analyzed and established. When the target vehicle is traveling in real time and is matched with a high-frequency traffic route, the waiting probability model of the waiting point at the next intersection is queried based on the current location, current time and vehicle speed. Based on the waiting probability value obtained from the query, the graded driving guidance instructions are generated and output to the vehicle's human-machine interface.
[0026] The specific process of this plan is described in detail below.
[0027] Step S1: Obtain the travel data of the target vehicle within the historical time period.
[0028] The data in this solution comes from user authorization and is periodically reported by the vehicle.
[0029] The data includes "speed, latitude and longitude, and time" (i.e., "GPS Data"), and is organized into trip data in units of trip ID, with different vehicles distinguished by the ID number.
[0030] The collected data is reported to the cloud platform and written to a message queue for further processing.
[0031] Step S2: Perform map matching on the itinerary data.
[0032] This step matches the raw, potentially drifting GPS latitude and longitude sequences to the actual road network. Data is consumed from Kafka by a Spark or Flink job, which then calls a map matching service (such as Valhalla).
[0033] The map matching service returns a sequence of matched road segments. For example, the original GPS points are converted into a list of standard segment_ids (road segment IDs) such as ["seg_101", "seg_102", "seg_105"]. The returned road segment sequence is written to the ClickHouse database for storage and subsequent analysis.
[0034] Step S3: Discover high-frequency traffic paths.
[0035] Periodically (e.g., daily), all historical route sequences (segments) of bicycles are read from ClickHouse. By aggregating and analyzing trips with "same start and end points" and "recurring routes", a list of "frequently used routes" is obtained.
[0036] In this embodiment, the starting point of the path is defined as follows: For new energy vehicles, according to relevant specifications, the reported data includes login and logout messages, which serve as the basis for determining the starting and ending points; for non-new energy vehicles, if no data is reported for about five minutes, the last frame of data is the ending point, and the first start is the first starting point of the day.
[0037] In this embodiment, "high frequency" is defined as follows: First, one month's worth of data is extracted from the global historical data, divided into multiple trip data sets by day, and trip matching data is constructed. For example, coarse-grained analysis determines the start and end points, while fine-grained analysis calculates distances by inputting the latitude and longitude points before and after the two points. This refines the coarse-grained trajectory, resulting in identical and partially deviated road segments. For instance, from the company to home, taking main roads and other branch roads, the monthly frequency of each road segment is calculated, as shown in the following formula: ; ; ; in, Latitude Longitude It is the Earth's radius (approximately 6371 km). To determine the latitude of two points ( 1, 2) Longitude difference (Δ) λ The relevant parameters for converting ) into a spherical triangle; Central angle (in radians) is the angle between two points and the center of the sphere, and it is the key bridge between "angle" and "distance". The final actual distance is obtained by multiplying the central angle of the sphere by the Earth's radius, which gives the shortest distance on the sphere between the two points.
[0038] In this embodiment, the path is based on long-term data, such as data from one month. The main road is extracted. For example, if you take this road for 10 days in a month, and then take a fork in the road at a certain intersection for three to five days, it is considered a secondary path of the main road.
[0039] In this embodiment, an algorithm that extracts subsequences and counts their frequencies using a sliding window ("common sub-path mining") is used to find the most frequently traveled route of the vehicle, such as the "ABC route" for commuting to and from get off work.
[0040] Step S4: Identify and verify waiting points at intersections.
[0041] On a high-frequency path, identify trajectory points where vehicle speeds are less than 5 km / h (or "and below") and where "multiple frames of reported data are pending". Cluster these low-speed points based on geographical location and path order, and initially label them as candidate intersection points.
[0042] To confirm that the candidate intersection points are traffic lights rather than temporary stops, multiple trips are extracted, and it is checked whether the vehicles meet the low-speed waiting conditions at these points in trips on different dates. Intersection points that pass the multi-trip verification are finally marked as "intersection trajectory points" (i.e., intersection waiting points).
[0043] In this embodiment, the multi-trip data is generally one month's worth of data, and is classified into two types: long routes and short routes. Long routes are generally like the route from home to the company, while short routes are generally those where some vehicles also pass through the same point, and common waiting situations are extracted.
[0044] Based on the second-level and millisecond-level data reported by the vehicles, different data frames are extracted. First, frames that meet the speed requirements are extracted, and then it is concluded whether these points are intersections. Generally, the second-level data frames are about 10 consecutive frames.
[0045] In this embodiment, "waiting" is mainly determined based on the vehicle's forward and backward speed per second and the engine or motor status, obtained from the data reported by the vehicle.
[0046] The "waiting to be excluded" situation is generally around 30%. Generally, excluding temporary roadside parking, if 10 data points are collected normally per month, and three or more data points are collected, it can be identified as an "intersection".
[0047] Step S5: Establish a time-dependent waiting probability model.
[0048] For each confirmed "intersection waiting point," extract all travel records from historical travel data for the route "from the previous intersection trajectory point to the next intersection trajectory point." Analyze each record to determine if there is a wait at the target intersection point and the duration of the wait.
[0049] Based on historical data, the percentage of waiting times at this intersection during different time periods (e.g., 8:00-9:00 AM) is calculated relative to the total number of passages. Probability labels such as "100% waiting," "90% waiting," or "40%" are output and linked to specific time periods to form a probability model for the intersection.
[0050] In this scheme, "waiting" means stopping at an intersection, with a speed of 0 or less than 5 km / h, and the latitude and longitude generally remain unchanged for the time being. For example, there is no change for at least three consecutive frames, the vehicle speed is less than 5 km / h, or the vehicle is in a parked waiting state (speed is 0).
[0051] The probability model constructs a sequence based on different average speeds to determine the waiting situation at intersections at different speeds. For example, at the intersection from one to two, the probability of passing is 30% at an average speed of 40-45 km / h, and 100% at 60 km / h. If there is a situation where one is always waiting at a certain intersection, it is considered that there is no possibility of passing through the intersection directly.
[0052] Step S6: Real-time driving matching and probability query.
[0053] While the vehicle is in motion, it reports its location (latitude and longitude) and speed to the cloud at set intervals. The cloud then matches the location information with the vehicle's stored "high-frequency travel routes" to determine which frequently traveled route the vehicle is currently on.
[0054] Based on the matched path and real-time location, determine which intersection the vehicle is waiting at next. Combining the current time (used for matching time periods) and current speed, query the corresponding waiting probability model for that intersection to obtain the predicted waiting probability value.
[0055] Step S7: Generate and output graded driving guidance instructions Based on the queried probability value, different instructions are generated according to a preset threshold.
[0056] For example: High probability of waiting (e.g., "usually will wait"): Triggers the "slow down" command; Low to medium probability waiting (e.g., "40%)": "Collect speed information. If the average vehicle speed is greater than a certain threshold, the vehicle will pass through the intersection and trigger a prompt such as "Please plan your route accordingly". The instructions are sent to the vehicle and guided through methods such as "voice module broadcast" or "light reminder".
[0057] The "threshold" in "low-to-medium probability waiting" is a dynamic threshold. It constructs the waiting situation at each intersection based on historical data and analyzes the waiting situation and waiting time intervals based on data reported by vehicles. It analyzes how many times the intersection was crossed during the waiting period and constructs speed ranges. For example, if the average speed is below 50km / h and the vehicle has not crossed at the intersection, it is considered that the vehicle cannot cross at that speed.
[0058] Based on historical data, intermediate data is calculated in real time to construct the distance from the current point to the intersection. Combined with the current speed and historical passing speed ranges, the required speed is calculated. Specifically: Group by "intersection ID + time period + signal cycle" and filter out samples with label 1 (pass); The minimum passing speed for each sample group (i.e. the lowest speed that can be passed in history) is calculated, which is the dynamic threshold for that scenario. When the real-time vehicle speed exceeds the "minimum passing speed", it is likely to pass; otherwise, it "wait".
[0059] To avoid interference from extreme values, statistical corrections were made, and the minimum passing speed of the 90th percentile was used (excluding 10% of outlier samples).
[0060] Expressed using the following formula: Threshold for historical excavation v three With the critical speed range [ v min ,v max Combine them, and take the intersection as the final decision threshold. v final-three : v final-three =max( v three-hist ,v min ); when v real ≥ v final-three and v real ≤ v max It is possible; when v real < v final-three or v real >v max Wait or slow down; in, v final-three The final decision threshold, v three-hist The threshold is obtained through historical data mining (e.g., the reasonable lower limit of vehicle speed calculated from historical data). v real The actual detected current vehicle speed. v min The minimum permissible speed, v max The maximum permissible speed.
[0061] Step S8, continuously updated.
[0062] The updated parameters mainly include the "dynamic speed threshold" for waiting / passing at intersections, i.e., "at what speed is a vehicle required to pass through an intersection on a green light," and "correction coefficients" for various influencing factors (such as weather and congestion weights), used to fine-tune the base thresholds to better adapt them to current road conditions.
[0063] During vehicle operation, new "vehicle speed-passability" data is uploaded to the cloud. The cloud periodically updates the threshold mapping table incrementally and then distributes it to the vehicle.
[0064] When the number of samples passing through a certain intersection increases by a set number, such as more than 100, an "update" is triggered, and the threshold is recalculated.
[0065] The vehicle-side supports differential updates of the threshold table (downloading only the changed intersection data), reducing traffic consumption.
[0066] When there is real-time congestion (vehicle speed in front <20km / h), the threshold will be forcibly lowered by 50%, or a message will be displayed saying "Currently congested, waiting for download and update".
[0067] In extreme weather (rain or snow), the threshold is increased by 20% based on road sensor data (either from sensors installed on the vehicle or by accessing meteorological data) (to reduce the speed of passage and improve safety).
[0068] The speed threshold correction is divided into two dimensions, and the final output is the optimal threshold fused from multiple factors: Historical dimension: Accumulated data on "vehicle speed-passability-scenario" ensures that the thresholds align with long-term traffic patterns; Real-time dimensions: weather, congestion, and signal phase change data, enabling thresholds to be adapted to the current intersection status.
[0069] v final =v hist × w i ×k real ; in, v hist This serves as the basic threshold for historical data mining. w i Adjust weights for various scenarios (such as weather weights and congestion weights). k real This is the calibration coefficient.
[0070] Based on historical extreme weather data, a weather type-threshold correction coefficient mapping table is constructed and distributed to the vehicle in advance.
[0071] The correction factor is calculated by statistically analyzing the difference in vehicle speed between extreme weather and clear weather to determine the correction factor. k w = Average speed of vehicles passing through under extreme weather conditions / Average speed of vehicles passing through under clear weather conditions; The revised rule is as follows: In rainy / snowy weather, the road surface friction coefficient decreases by approximately 30% to 50%, leading to an increase in braking distance and a higher set threshold, i.e., a correction coefficient. k w =1.2~1.5; In foggy weather, visibility is low, vehicle speed decreases, and the set threshold is lowered, i.e., the correction coefficient. k w =0.7~0.9.
[0072] The entire threshold correction and extreme weather protection process is achieved through a closed loop of cloud-based distribution, vehicle-side computation, and data feedback. On the cloud side, historical data is mined offline to generate a basic threshold table and a weather correction coefficient table; real-time traffic data uploaded by vehicles is received and incremental corrections are performed; and full / differential threshold tables and weather correction coefficients are sent to vehicles.
[0073] The vehicle-side system preloads threshold tables and correction coefficients from the cloud; it collects vehicle status (vehicle speed) and roadside data (weather, congestion) in real time; it integrates historical thresholds, weather coefficients, and real-time road conditions to calculate the final threshold; it outputs guidance prompts to the driver and sends the passage results back to the cloud.
[0074] Closed-loop optimization: The vehicle sends back "actual speed - pass result" data to continuously enrich the cloud training set; the cloud regularly evaluates the accuracy of the model and thresholds, and iteratively optimizes and corrects the strategy.
[0075] The data is updated using a grouping and accumulation method, grouped in three dimensions: "Intersection ID + Time Period + Weather Type," and continuously accumulated through sample data. Grouping example: crossroad_1001_morning_rain (intersection 1001 + morning rush hour + rainy day); During sample screening, only valid samples that pass with a "green light" are retained, and the basic threshold is set at the 10th percentile of the passing samples to avoid interference from extreme values.
[0076] The incremental update formula is: Q new = a × Q old +(1- a )× Q batch ; Q old Historical quantile threshold, Q batch The quantiles of the newly added batch of samples, a The smoothing coefficient (ranging from 0.7 to 0.9) is used to balance the influence of historical patterns and new data; the trigger condition is: when the number of new samples in a group is greater than 50, incremental calculation is performed.
[0077] This solution leverages in-depth mining and analysis of individual vehicle historical driving data to construct a personalized intersection traffic knowledge base and predictive model, upgrading navigation from "general information broadcasting" to "personalized driving strategy guidance." Compared to current navigation systems that rely solely on real-time traffic conditions or fixed signal cycles, it provides drivers with highly customized and predictive traffic suggestions, resulting in significant benefits in multiple dimensions, including energy conservation, alleviating range anxiety, and improving driving comfort and safety.
[0078] This solution focuses its analysis on drivers' unique "high-frequency travel routes," concentrating on their real and repetitive travel scenarios, thus avoiding the neglect of individual differences by general models. Furthermore, rigorous validation of "intersection waiting points" based on individual historical data accurately identifies the driver's actual, fixed signal control points (such as traffic lights), rather than temporary stopping points, ensuring the reliability of the predictions. This makes the driver's expectations of the road ahead highly stable and reliable, resolving the decision-making dilemma of "should I accelerate or coast" caused by inaccurate navigation predictions.
[0079] The "time-dependent waiting probability model" established in this solution can inform drivers of the probability of waiting at a specific intersection at a specific time. This provides richer information and transforms the abstract information of "possibly waiting at a red light" into specific, actionable "speed-action" guidance instructions, promoting energy-efficient driving behavior. By classifying probabilities and providing "dynamic speed suggestions" in low-to-medium probability scenarios, ineffective energy consumption can be avoided. For example, when a "high probability of waiting" is predicted, the driver is advised in advance to smoothly decelerate and coast, avoiding sudden braking and idling, thus reducing the waste of kinetic energy converted into heat.
[0080] This solution uses clear action guidelines (such as "Please maintain a speed of xx km / h or higher to pass") to shift drivers from passively accepting "There are still XX kilometers of range left" to actively implementing "Driving according to this strategy can save energy." It partially transforms the uncertainty of "remaining range" into the certainty of "crossing the intersection ahead," alleviating range anxiety on both psychological and practical levels.
[0081] Correspondingly, vehicle energy-saving navigation systems based on historical behavior learning include; The data acquisition module is configured to acquire the travel data of the target vehicle within a historical time period. The travel data includes a geographical location sequence, a speed sequence, and the corresponding timestamps. The high-frequency path mining module is configured to: perform map matching on the travel data and generate a historical path sequence composed of standard road segment identifiers; The high-frequency path mining module is also configured to: mine one or more high-frequency travel paths of the target vehicle based on the frequency of occurrence of historical path sequences; The intersection waiting module is configured to: identify trajectory points that meet preset low-speed conditions based on speed sequences on high-frequency traffic paths, cluster the trajectory points and verify them as intersection waiting points; The intersection waiting module is also configured to: for each intersection waiting point, based on its historical passage records in different time segments, statistically analyze and establish a time-dependent waiting probability model; The intersection waiting module is also configured to: when the target vehicle is traveling in real time and is matched with a high-frequency traffic path, query the waiting probability model of the next intersection waiting point based on the current location, current time and vehicle speed; The navigation instruction output module is configured to generate and output graded driving guidance instructions to the vehicle's human-machine interface based on the waiting probability value obtained from the query.
[0082] Correspondingly, a computer program product includes computer-readable instructions that, when executed on an electronic device, cause the electronic device to implement the aforementioned vehicle energy-saving navigation method based on historical behavior learning.
[0083] Accordingly, an electronic device includes at least one processor and a memory connected to the processor, the memory being used to store computer programs; the processor is used to execute the computer programs, enabling the electronic device to implement the aforementioned vehicle energy-saving navigation method based on historical behavior learning.
[0084] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A vehicle energy-saving navigation method based on historical behavior learning, characterized in that, Includes the following steps: Obtain the travel data of the target vehicle within a historical time period. The travel data includes geographical location sequence, speed sequence, and corresponding timestamps. Map matching is performed on the travel data to generate a historical route sequence composed of standard road segment identifiers; Based on the frequency of occurrence of historical path sequences, one or more high-frequency travel paths of the target vehicle are identified. On high-frequency traffic paths, trajectory points that meet preset low-speed conditions are identified based on speed sequences, and these trajectory points are clustered and verified as intersection waiting points. For each intersection waiting point, based on its historical passage records at different time segments, a time-dependent waiting probability model is statistically analyzed and established. When the target vehicle is traveling in real time and is matched with a high-frequency traffic route, the waiting probability model of the waiting point at the next intersection is queried based on the current location, current time and vehicle speed. Based on the waiting probability value obtained from the query, the graded driving guidance instructions are generated and output to the vehicle's human-machine interface.
2. The vehicle energy-saving navigation method based on historical behavior learning as described in claim 1, characterized in that, The process involves identifying one or more high-frequency travel routes for the target vehicle. Specifically, this includes periodically reading the vehicle's historical routes from a historical route sequence database; performing aggregation analysis on routes with the same start and end points that repeat within a preset long-term window to obtain a list of commonly used routes; and determining high-frequency travel routes by calculating the frequency of occurrence of routes or key road segments within the long-term window.
3. The vehicle energy-saving navigation method based on historical behavior learning as described in claim 1, characterized in that, Clustering and validating trajectory points as intersection waiting points is specifically as follows: On a high-frequency path, find trajectory points where the vehicle speed is below a first speed threshold; then, spatially cluster the obtained low-speed trajectory points according to their geographical location and path order to form candidate intersection points. Extract multiple trip data of the target vehicle on different dates to verify whether the low-speed waiting conditions are met at the candidate intersection points; mark the candidate intersection points that pass the multi-trip verification as intersection waiting points.
4. The vehicle energy-saving navigation method based on historical behavior learning as described in claim 1, characterized in that, Establish a time-dependent waiting probability model, specifically as follows: For each intersection waiting point, extract its passage records at different time segments from historical data; The probability of waiting at intersections within each time segment is calculated as the proportion of times a vehicle waits at the intersection waiting point relative to the total number of passages. The waiting behavior is determined by the vehicle speed being lower than the second speed threshold and the position not changing significantly within multiple consecutive data frames.
5. The vehicle energy-saving navigation method based on historical behavior learning as described in claim 1, characterized in that, Generate and output graded driving guidance instructions, specifically by comparing the waiting probability value obtained from the query with at least two preset probability thresholds, and triggering different levels of guidance instructions based on the comparison results; The boot instructions include at least the following: When the probability of waiting is higher than the first probability threshold, the "slow down" command is triggered. When the waiting probability is between the first probability threshold and the second probability threshold, a "rational planning" instruction containing dynamic speed suggestions is triggered.
6. The vehicle energy-saving navigation method based on historical behavior learning as described in claim 5, characterized in that, The dynamic speed recommendations in the "Rational Planning" instruction are generated in the following way: Based on the historical passage records of the intersection waiting point, the minimum historical vehicle speed that can successfully pass through in the corresponding time segment is analyzed and used as the dynamic speed threshold. The real-time vehicle speed is compared with the dynamic speed threshold. If the real-time vehicle speed is greater than or equal to the dynamic speed threshold, the guidance of "maintain current speed" or "suggest acceleration" is generated; otherwise, the guidance of "suggest deceleration" is generated.
7. The vehicle energy-saving navigation method based on historical behavior learning as described in claim 1, characterized in that, It also includes a continuous update step for the model and thresholds, specifically: The vehicle will transmit traffic data, including real-time vehicle speed, passage results, and environmental scene information, back to the cloud. The cloud grouped the transmitted data by intersection, time period and scenario, and used the new data to incrementally update the waiting probability model and dynamic speed threshold; the updated model and threshold were sent to the vehicle for subsequent real-time guidance.
8. A vehicle energy-saving navigation system based on historical behavior learning, characterized in that, include: The data acquisition module is configured to acquire the travel data of the target vehicle within a historical time period. The travel data includes a geographical location sequence, a speed sequence, and the corresponding timestamps. The high-frequency path mining module is configured to: perform map matching on the travel data and generate a historical path sequence composed of standard road segment identifiers; The high-frequency path mining module is also configured to: mine one or more high-frequency travel paths of the target vehicle based on the frequency of occurrence of historical path sequences; The intersection waiting module is configured to: identify trajectory points that meet preset low-speed conditions based on speed sequences on high-frequency traffic paths, cluster the trajectory points and verify them as intersection waiting points; The intersection waiting module is also configured to: for each intersection waiting point, based on its historical passage records in different time segments, statistically analyze and establish a time-dependent waiting probability model; The intersection waiting module is also configured to: when the target vehicle is traveling in real time and is matched with a high-frequency traffic path, query the waiting probability model of the next intersection waiting point based on the current location, current time and vehicle speed; The navigation instruction output module is configured to generate and output graded driving guidance instructions to the vehicle's human-machine interface based on the waiting probability value obtained from the query.
9. A computer program product, characterized in that, Includes computer-readable instructions that, when executed on an electronic device, cause the electronic device to perform the steps in the vehicle energy-saving navigation method based on historical behavior learning as described in any one of claims 1-7.
10. An electronic device, characterized in that, It includes at least one processor and a memory connected to the processor, the memory being used to store a computer program; the processor is used to execute the computer program, enabling the electronic device to perform the steps in the vehicle energy-saving navigation method based on historical behavior learning as described in any one of claims 1-7.