A method for passenger flow decision based on dual-antenna RFID and RSSI priority

Abstract

The application discloses a passenger flow method based on double-antenna RFID and RSSI priority decision, and belongs to the technical field of intelligent transportation and Internet of Things. The method comprises the following steps: deploying an outer antenna and an inner antenna at a vehicle door; collecting, by a data acquisition module, antenna port, RSSI signal strength and timestamp data of an RFID tag carried by a passenger; triggering detection and snapshot creation; and determining the getting-on and getting-off directions of the passenger by an algorithm processing module through a double-antenna sequence direction analysis method. The application can reasonably utilize the spatial layout of the FRID multi-antenna, accurately identify the getting-on and getting-off directions of the bus passenger, avoid repeated counting, and realize high-precision bus passenger flow statistics.

Description

Technical Field

[0001] This invention belongs to the field of intelligent transportation and Internet of Things technology, specifically referring to a passenger flow decision-making method based on dual-antenna RFID and RSSI priority. Background Technology

[0002] In the field of public transportation passenger flow statistics, accurately obtaining the number of passengers getting on and off the bus is crucial for operation scheduling, capacity planning, and benefit analysis.

[0003] In recent years, Radio Frequency Identification (RFID) technology has been explored for passenger flow statistics due to its advantages such as contactless operation, fast identification speed, and low tag cost. A typical approach involves installing a single RFID reader antenna outside the vehicle door to read passive or active electronic tags with unique identification codes carried by passengers. However, this solution has revealed the following serious shortcomings in engineering practice, falling far short of practical application requirements:

[0004] (1) Lack of direction determination: A single antenna can only sense the appearance of the tag signal and cannot distinguish whether the passenger enters the vehicle from outside or exits the vehicle from inside, resulting in the statistical "passage times" not being mapped to the net flow of passengers getting on and off the vehicle.

[0005] (2) Serious duplicate counting: When passengers hesitate, avoid or pass slowly at the door, the tag is read repeatedly by the reader. If there is no deduplication mechanism, the same passenger may be recorded multiple times within a few seconds, resulting in an inflated count.

[0006] (3) Processing timeliness in multi-person scenarios: When multiple passengers get on and off the vehicle at the same time, the signals of multiple tags are intertwined. A single antenna cannot match the reading events of each tag with the specific passenger's passage behavior, which is very easy to cause mismatch or omission.

[0007] (4) Lack of utilization of signal strength information: The Received Signal Strength Indication (RSSI) available to the reader was not used for analysis, while RSSI is closely related to the distance and orientation of the tag from the antenna. The lack of RSSI analysis means that it is impossible to determine whether the tag has actually crossed the physical plane of the door, nor can it distinguish the spatial position inside and outside.

[0008] (5) Lack of time series analysis: Existing simple counting only focuses on whether the tag is read, while ignoring the time series characteristics such as the time interval and sequence between each reading time. Therefore, it is impossible to construct an effective "through event" judgment logic and it is difficult to eliminate invalid wandering and random interference.

[0009] Therefore, there is an urgent need for an innovative counting method that can rationally utilize the spatial layout, signal strength, and time series characteristics of FRID multi-antenna systems to overcome the shortcomings of existing technologies. Summary of the Invention

[0010] This invention provides a passenger flow decision-making method based on dual-antenna RFID and RSSI priority. It can make reasonable use of the spatial layout of RFID multi-antennas, accurately identify the boarding and alighting directions of bus passengers, and avoid double counting, thereby achieving high-precision bus passenger flow statistics.

[0011] To achieve the above objectives, this invention provides a passenger flow priority decision-making method based on dual-antenna RFID and RSSI, comprising the following steps: deploying an outer antenna and an inner antenna at the vehicle door, with the outer antenna installed on the steps or door frame outside the door and the inner antenna installed inside the vehicle compartment; a reader device collecting antenna port, RSSI signal strength, and timestamp data of RFID tags carried by passengers; triggering detection and snapshot creation; and an algorithm program module determining the boarding and alighting directions using a dual-antenna sequence direction analysis method.

[0012] As a further aspect of the present invention: the dual-antenna sequence direction analysis method is the main algorithm, which includes the following steps:

[0013] Boarding direction determination: The algorithm module extracts the earliest continuous outer antenna sequence and the latest continuous inner antenna sequence. If the timestamp of the latter is greater than that of the former and the time difference is ≥1000ms, then it is determined to be boarding.

[0014] Disembarkation determination: Extract the earliest continuous inner antenna sequence and the latest continuous outer antenna sequence. If the timestamp of the latter is greater than that of the former and the time difference is ≥1000ms, then it is determined as disembarkation.

[0015] As a further aspect of the present invention, it also includes an RSSI signal strength analysis step:

[0016] When there is only one candidate label, calculate its RSSI difference range;

[0017] If the RSSI difference is less than the preset threshold of 3.0dB, it is determined as "no movement at the door" and the calculation is skipped.

[0018] As a further aspect of the present invention: when the main algorithm determines failure and triggers a degradation algorithm, the degradation algorithm includes detecting whether there are consecutive target antenna scan records of ≥2 times; if there are multiple sequences that meet the conditions, the sequence with the largest RSSI difference is selected.

[0019] As a further aspect of the present invention, the specific boundary conditions for triggering the degradation algorithm include the following four categories:

[0020] Insufficient record count boundary: The total number of valid scan records in the tag sliding queue is less than the minimum threshold, and the amount of data is insufficient to support dual-antenna sequence reconstruction;

[0021] Single-sided data missing boundary: When a passenger passes through, one side of the antenna does not read the signal at all due to human body absorption or metal shielding, resulting in a break in the timing chain of the dual antennas;

[0022] Temporal disorder / abnormal boundary: The earliest extracted sequence timestamp is greater than the latest sequence timestamp, and the temporal logic has serious overlap or inversion;

[0023] Loitering deadlock boundary: Although the temporal sequence is satisfied, the time difference calculated from both sides of the sequence is less than the critical anti-loitering threshold. The main algorithm judges it as "loitering at the door" for the purpose of high-precision anti-interference.

[0024] As a further aspect of the present invention, it also includes a four-level prioritization method for decision trees, comprising the following steps:

[0025] First priority level: Algorithm determines the success level decision;

[0026] Decision logic: The system first evaluates the algorithm output results of each candidate snapshot and prioritizes the selection of candidate snapshots that have been successfully determined through dual-antenna sequence direction analysis;

[0027] Flow control: If there are candidates that the main algorithm determines are successful, they are directly sent to the second-level evaluation as the core candidate set. At this time, the candidate snapshots that the downgrade algorithm determines are suspended. Only when the main algorithm of all candidate snapshots fails and returns UNKNOWN will the second-level evaluation open the snapshots that the downgrade algorithm determines are successful.

[0028] Second-level priority: RSSI energy fluctuation explicit decision;

[0029] Decision logic: If multiple candidate snapshots still exist in parallel under the same algorithm success level, the system extracts the range of signal strength difference between each snapshot;

[0030] Flow control: The decision tree prioritizes the candidate snapshot with the largest RSSI difference; the larger the RSSI difference, the more significant the movement trajectory and activity of the label when crossing the physical door beam coverage area, thus accurately locating the primary target with the most obvious current movement characteristics.

[0031] Third-level priority: Time window timeliness convergence constraint;

[0032] Decision logic: If an extreme parallel situation occurs after the second-level energy assessment, the system invokes the timeliness constraint to achieve convergence;

[0033] Flow control: The system compares the last update timestamp of each candidate snapshot and prioritizes the snapshot whose last update time is closest to the current door sensor trigger cutoff time, thereby eliminating outdated dynamic interference left over from past sliding windows;

[0034] Fourth priority level: Single-object dynamic verification and static removal protection;

[0035] Decision logic: When the decision tree filters layer by layer and finally locks in or only has 1 valid candidate snapshot left, the system performs mandatory position state machine transition compliance checks and static object discrimination.

[0036] Static filtering at the entrance: If the RSSI difference of the only candidate snapshot is less than the preset fluctuation threshold, the decision tree directly starts the interruption protection, determines it as "a person standing still at the door when the door opens", completely removes it from the pass queue, skips this calculation, and ensures the consistency of the atomic counter data.

[0037] A system for prioritizing passenger flow using dual-antenna RFID and RSSI based on the above-described method includes: a reader device, which comprises: an outer antenna, an inner antenna, an RFID reader / writer, a camera, and an algorithm program module; the algorithm program module comprises: a data processing module and a counting and statistics module.

[0038] Compared with existing technologies, the beneficial effects of this invention are as follows: Accurate direction recognition: Through dual-antenna sequence analysis, the direction of boarding and alighting is accurately distinguished, with an accuracy rate of over 95%; Anti-wandering interference: Time difference threshold detection (1000ms) effectively filters out misjudgments caused by wandering at the door; RSSI signal utilization: The RSSI difference is used to determine whether a passenger is moving; a single candidate time difference <3dB is considered stationary; Multi-candidate processing: When multiple people board and alight simultaneously, the candidate with the largest RSSI difference is selected, prioritizing the identification of the most obvious movement; Degradation mechanism: When the main algorithm fails, a degradation algorithm is automatically activated to improve system robustness; High-precision bus passenger flow statistics are achieved. Attached Figure Description

[0039] Figure 1 This is an architecture diagram of a vehicle door RFID personnel entry and exit counting system according to an embodiment of the present invention.

[0040] Figure 2 This is a timing diagram of the dual-antenna sequence analysis algorithm for the vehicle loading scenario in an embodiment of the present invention.

[0041] Figure 3 This is a schematic diagram illustrating the RSSI difference judgment principle of an embodiment of the present invention.

[0042] Figure 4 This is a state transition diagram according to an embodiment of the present invention.

[0043] Figure 5 This is a flowchart of the parallel extraction of temporal features in an embodiment of the present invention. Detailed Implementation

[0044] The invention will now be further described with reference to the accompanying drawings.

[0045] like Figures 1 to 5 As shown, taking a vehicle door RFID personnel entry and exit counting system as an example, the reader device includes: an outer antenna (ANT4), an inner antenna (ANT1), an RFID reader / writer, a camera, and an algorithm program module. The algorithm program module includes: a data processing module, a counting and statistics module, and a triggering device, which includes a vehicle door sensor or a camera, used to trigger calculations. A passenger flow priority decision method based on dual-antenna RFID and RSSI includes the following steps:

[0046] Step 1: Dual-antenna layout configuration:

[0047] Two RFID antennas are deployed at the bus doors:

[0048] External antenna (ANT4): Installed on the step or door frame outside the vehicle door to detect the area outside the vehicle.

[0049] Inner antenna (ANT1): Installed inside the car door, used to detect the area inside the vehicle.

[0050] Step 2: Real-time RFID data collection:

[0051] When passengers carrying RFID tags pass through the door area, the system continuously collects RFID scanning data. Each record includes: Tid, EPC, RSSI, antenna number, and timestamp.

[0052] Step 3: Sliding window data caching:

[0053] Maintain a sliding time window queue for each label:

[0054] Queue length limit: 70 records;

[0055] Update the maximum and minimum RSSI values ​​of the tags in real time;

[0056] Calculate the RSSI difference range: Range = max(RSSI) - min(RSSI);

[0057] Step 4: Parallel extraction of temporal features (ECS / LCS):

[0058] This step targets the specific tag record queue cached in step three. Before judgment, data isolation is performed according to the target antenna port (inner antenna 1 or outer antenna 4). Records with an adjacent timestamp interval Δt ≤ 1000ms in the same antenna are aggregated into independent continuous subsequences to construct a valid sequence set (C_valid). Parallel dual-segment extraction is then performed.

[0059] Branch A: Earliest Continuous Sequence (ECS) Extraction

[0060] Objective: To determine the timing characteristics of the initial physical critical point of the target antenna's radio frequency field when the passenger tag first enters the target antenna.

[0061] Extraction mechanism: Traverse the valid sequence set (C_valid) of the target antenna, extract the record with the smallest timestamp within each subsequence, and use it as the starting timestamp (t_start) of that subsequence. Sort all subsequences in ascending order on the time axis according to t_start, accurately isolate and extract the first subsequence block with position index [0], and define it as the earliest continuous sequence (ECS).

[0062] Branch B: Latest Continuous Sequence (LCS) Extraction

[0063] Objective: To identify the departure wake characteristics of a passenger tag that has completely exited the target antenna's radio frequency field termination state.

[0064] Extraction mechanism: Traverse the valid sequence set (C_valid) of the target antenna, extract the record with the largest timestamp within each subsequence, and use it as the end timestamp (t_end) of that subsequence. Sort all subsequences in descending order on the time axis according to t_end, accurately isolate and extract the first subsequence block with position index [0], and define it as the latest continuous sequence (LCS).

[0065] Step 5: Trigger detection and snapshot creation:

[0066] When the car door opens (triggered by a camera or door sensor), the system executes:

[0067] Delayed waiting: Wait 2 seconds to ensure passengers have completed the passage;

[0068] Snapshot Filtering: Create candidate snapshots with filter criteria:

[0069] Status matching: When boarding, the passenger must be in an "outside the vehicle" state; when alighting, the passenger must be in an "inside the vehicle" state.

[0070] Time window: Updated within the last 3 seconds;

[0071] Single-object RSSI check: If there is only one candidate and the RSSI difference is <3.0dB, it is determined as "no movement at the door" and the calculation is skipped.

[0072] Step Six: Dual-Antenna Sequence Direction Analysis (Main Algorithm)

[0073] Perform a dual-antenna sequence analysis algorithm on each candidate snapshot:

[0074] Vehicle entry assessment (outer side → inner side: 4 → 1):

[0075] 1. Traverse the record queue and extract the earliest consecutive sequence of outer antenna 4;

[0076] 2. Traverse the record queue and extract the latest consecutive inner antenna 1 sequence;

[0077] 3. Calculate the time difference: Δt = latest inner time - earliest outer time;

[0078] 4. Judgment Logic:

[0079] If Δt ≥ 1000ms: it is determined to be boarding;

[0080] If Δt < 1000ms: it is determined to be loitering at the door (direction unknown).

[0081] Disembarkation assessment (inner side → outer side: 1 → 4):

[0082] 1. Traverse the record queue and extract the earliest consecutive inner antenna 1 sequence;

[0083] 2. Traverse the record queue and extract the latest consecutive sequence of outer antenna 4;

[0084] 3. Calculate the time difference: Δt = latest outermost time - earliest innermost time;

[0085] 4. Judgment Logic:

[0086] If Δt ≥ 1000ms: determine that you are getting off the bus;

[0087] If Δt < 1000ms: it is determined to be loitering at the door (direction unknown).

[0088] Step 7: One-sided sequence detection (downgrade algorithm):

[0089] When the main algorithm cannot determine the direction, a fallback algorithm is activated.

[0090] The collaborative relationship and switching logic between the main algorithm and the degradation algorithm:

[0091] 1) Overview of Collaborative Relationships:

[0092] This invention constructs a dual guarantee mechanism of "high-precision topology of the main algorithm + strong robustness and fault tolerance of the degradation algorithm".

[0093] The main algorithm (dual-antenna sequence analysis) is the preferred high-precision judgment method. It relies heavily on the spatial topology and time difference characteristics of the dual antennas (outer ANT4 and inner ANT1) to accurately identify standard single-person passage and crowded passage scenarios, and to perform high-precision filtering of loitering behavior at the door.

[0094] Degradation Algorithm (Single-sided Sequence Detection): Serves as a fallback mechanism for system fault tolerance. In complex real-world public transportation scenarios (such as severe data loss due to human obstruction causing significant missed readings on one side of the antenna, timing breaks, or signal anomalies), the main algorithm may fall into a deadlock state of "unable to determine" due to incomplete data. In this case, the degradation algorithm, through in-depth analysis of continuous high-frequency scanning and RSSI difference energy fluctuations of the single-sided antenna, enforces convergence on the passage behavior, ensuring that the system does not miss any data under extreme interference.

[0095] 2) Algorithm switching trigger conditions

[0096] The system makes decisions on a snapshot-by-snapshot basis. The system immediately activates a state switching instruction and calls the degradation algorithm if and only if the main algorithm returns UNKNOWN (unknown direction) as the result of its decision on a candidate snapshot.

[0097] The specific boundary conditions under which the main algorithm fails (outputs UNKNOWN) and triggers the degradation algorithm include the following four categories:

[0098] Record count insufficiency boundary: The total number of valid scan records in the tag sliding queue is less than the minimum threshold (records.size() < 4), and the amount of data is insufficient to support dual-antenna sequence reconstruction.

[0099] Single-sided missing data boundary: When a passenger passes through, one side of the antenna may not read any signal at all due to human body absorption or metal shielding (firstSequence.isEmpty() or lastSequence.isEmpty()), resulting in a break in the timing chain of the two antennas.

[0100] Temporal disorder / abnormal boundary: The earliest extracted sequence timestamp is greater than the latest sequence timestamp (earliestTime > latestTime), indicating severe overlap or inversion of temporal logic.

[0101] Loitering deadlock boundary: Although the temporal sequence is satisfied, the time difference calculated from both sides is less than the critical anti-loitering threshold (timeDiff < 1000ms). For the purpose of high-precision anti-interference, the main algorithm judges it as "loitering at the door (direction unknown)".

[0102] 3) Algorithm switching logic and workflow

[0103] The system's execution state machine strictly follows the following control logic for seamless algorithm switching and data closure:

[0104] Step 1: Triggering and Snapshot Loading

[0105] Once the door opens (triggered by the door sensor or camera) and after a delay of TRIGGER_DELAY_MS (2000ms), the system completes the data accumulation for the passenger's passage. The system creates candidate snapshots, filters, and loads valid tag data that meets the time window and state requirements.

[0106] Step 2: Prioritize calling the main algorithm

[0107] The system first sends the snapshot dataset into the main algorithm engine analyzeDirection to perform dual-antenna timing and time difference verification.

[0108] Successful branch of the main algorithm: If the main algorithm successfully solves a clear direction (returns BOARDING or EXITING), the system directly skips the downgrade algorithm and proceeds to Step 4 for multi-candidate arbitration.

[0109] Step 3: State Switching and Degradation Algorithm Activation

[0110] If the main algorithm triggers any of the above boundary conditions and returns UNKNOWN, the system state machine immediately issues a fallback control instruction to activate the fallbackDirectionAnalysis algorithm:

[0111] Target-side positioning: Based on the currently triggered service type (onboard determination or offboard determination), dynamically lock onto a single-sided target antenna (lock onto the outer ANT4 when onboard, lock onto the inner ANT1 when offboard).

[0112] Single-sided continuous search: Search the history of this tag for the existence of a valid target antenna sequence with ≥ 2 consecutive scans.

[0113] Energy Feature Selection: If multiple subsequences that meet the continuity condition are retrieved on one side, the degradation algorithm no longer compares the temporal order, but calculates the RSSI difference (max(RSSI) - min(RSSI)) of each subsequence, selects the sequence with the largest RSSI difference (i.e. the most obvious movement feature) as the degradation judgment basis, and forces the output of the passage direction.

[0114] Complete failure handling: If the downgrade algorithm still does not find ≥ 2 consecutive scans after retrieval, the system determines that the tag is non-traffic interference (such as pedestrians passing by outside the vehicle), completely abandons this calculation, and keeps the counter unchanged.

[0115] Step 4: Multi-candidate arbitration and counting loop closure

[0116] During the final data update phase, if both a successful snapshot of the main algorithm and a successful snapshot of the degradation algorithm exist simultaneously, the system executes the combinatorial optimization selection logic:

[0117] 1. Prioritize candidate snapshots that were successful with the main algorithm.

[0118] 2. Among candidates with the same success level, prioritize the snapshot with the largest RSSI difference (most obvious movement) to update the position state, and drive the closed-loop update of the atomic counter (rfidIn++ or rfidOut++) and the real-time number of people in the vehicle (rfidCurrent).

[0119] Vehicle downgrade: Check for the presence of ≥2 consecutive external antenna 4 scan records;

[0120] Degradation upon exiting the vehicle: Detect whether there are ≥2 consecutive scan records of the inner antenna 1;

[0121] Selection rule: If multiple sequences meet the criteria, select the sequence with the largest RSSI difference.

[0122] Step 8: Multiple Candidate Selections and State Updates

[0123] If multiple candidate snapshots all meet the conditions:

[0124] 1. Prioritize candidates whose main algorithm has succeeded;

[0125] 2. Select the candidate with the largest RSSI difference (most obvious shift).

[0126] 3. Update tag status:

[0127] Successfully boarded: OUTSIDE → INSIDE

[0128] Successfully disembarked: INSIDE → OUTSIDE

[0129] 4. Update the counter:

[0130] Get on board: rfidIn++, rfidCurrent = rfidIn - rfidOut

[0131] Get off the bus: rfidOut++, rfidCurrent = rfidIn – rfidOut.

[0132] To achieve high-precision passenger flow calculation in concurrency scenarios such as crowded conditions or multiple people passing through the doors simultaneously, when a detection is triggered, if multiple candidate snapshots that meet the time window and state requirements are selected in the sliding window cache, the system will strictly follow the following four-level priority decision tree to perform layer-by-layer filtering and target selection from top to bottom:

[0133] First priority level: Algorithm determines success level (Main algorithm > Degradation algorithm)

[0134] Decision logic: The system first evaluates the algorithm output results of each candidate snapshot. It prioritizes selecting the set of candidate snapshots that have been successfully determined (returning BOARDING or EXITING) through dual-antenna sequence direction analysis (main algorithm).

[0135] Flow control: If there are candidates that the main algorithm determines are successful, they are directly sent to the second-level evaluation as core candidates. At this time, the candidate snapshots that are successful by one-sided sequence detection (downgrade algorithm) will be suspended. Only when the main algorithm of all candidate snapshots fails and returns UNKNOWN will the second-level evaluation open the snapshots that are successful by the downgrade algorithm.

[0136] Second-level priority: RSSI energy fluctuation dominant decision (larger difference > smaller difference)

[0137] Decision logic: If there are still multiple candidate snapshots with the same success level, the system extracts the range of signal strength difference between each snapshot (Range = max(RSSI) - min(RSSI)).

[0138] Flow control: The decision tree prioritizes candidate snapshots with the largest RSSI difference. The larger the RSSI difference, the more significant the movement trajectory and activity of the tag when crossing the physical door beam coverage area (e.g., passenger C is prioritized over passenger D in Example 3), thereby accurately locating the primary target with the most obvious current movement characteristics.

[0139] Third-level priority: Time window timeliness convergence constraint (most recent update > earliest update)

[0140] Decision logic: If an extreme parallel situation occurs after the second energy assessment (such as multiple people passing through at the same speed with the same type of label, resulting in the algorithm level and RSSI difference range being completely consistent), the system calls the timeliness constraint to converge.

[0141] Flow control: The system compares the last update timestamp (lastUpdateTime) of each candidate snapshot and prioritizes the snapshot whose last update time is closest to the current door sensor trigger cutoff time (i.e., the most up-to-date snapshot), thereby eliminating outdated dynamic interference left over from past sliding windows.

[0142] Fourth priority level: Single object dynamic verification and static culling protection (dynamic crossing > doorway immobile culling)

[0143] Decision logic: When the decision tree filters layer by layer and finally locks in or only has 1 valid candidate snapshot, the system performs mandatory location state machine transition compliance checks and static object identification (corresponding to step five and the security threshold).

[0144] Static filtering at the entrance: If the RSSI difference of the only candidate snapshot is less than the preset fluctuation threshold (rssiRange < 3.0dB), the decision tree directly activates interrupt protection, determines it as "a person standing still at the door when the door opens", completely removes it from the pass queue, skips this calculation, and ensures the consistency of atomic counter data.

[0145] In the control algorithm proposed in this invention, the time threshold (default 1000ms) and energy threshold (default 3.0dB) are set based on critical benchmarks determined strictly according to human dynamics, the physical topology of bus doors, and the radio frequency field propagation model. To ensure the system's high adaptability to different vehicle models and different metal shielding environments, the above thresholds are all default initial parameters. In actual engineering deployments, they all support flexible secondary adjustments and calibrations based on the actual site conditions.

[0146] Direction determination time threshold (default value 1000ms):

[0147] Theoretical Basis: The physical width of a standard bus door is typically between 0.8 and 1.2 meters. According to traffic engineering statistics, the average speed of a passenger stepping through the door passage is approximately 0.8 m / s to 1.5 m / s. Therefore, the physical passage time for a passenger holding a card to completely cross the polarization coverage areas of the outer antenna (ANT4) and the inner antenna (ANT1) is, under normal circumstances, a parabolic range of 0.8 to 1.5 seconds. Setting 1000 ms as the lower limit of the baseline is, on the one hand, to match the minimum physical passage time for a human body, tolerating the sudden radio frequency blind zone caused by human bodies during crowding (typically within 300 to 600 ms); on the other hand, if the Δt detection value is less than 1000 ms, it can be physically determined that this is not an effective passage displacement across the door, but rather a lingering pseudo-timing caused by the passenger leaning against the door or in the overlapping area of ​​the side lobes of the two door antennas, and the system must implement deadlock interception.

[0148] On-site dynamic adjustment mechanism: This 1000ms is the system default baseline value. In actual engineering implementation, technicians can adjust this time threshold proportionally (e.g., adjust it to 800ms or 1200ms) based on the actual width of the bus door passage, the slope of the steps (e.g., double or triple steps will cause longer travel time), and the specific congestion characteristics of the site, in order to achieve the best anti-wandering filtering effect.

[0149] Minimum RSSI difference threshold for a single object (default = 3.0dB):

[0150] Theoretical Basis: According to the Friis Transmission Equation for electromagnetic wave propagation, the Received Signal Strength Indication (RSSI) is logarithmically inversely proportional to the absolute spatial distance between the tag and the antenna. When a passenger undergoes a real displacement across the door (moving closer to the antenna and then further away), a classic parabolic dynamic energy trajectory is generated, and the corresponding dynamic RSSI difference range is usually extremely significant, stabilizing between 6.0 dB and 15.0 dB. Conversely, when the door is open and there is only one isolated static candidate object inside the sliding window (such as a passenger leaning against the door frame), although the RSSI will fluctuate randomly due to slight movements of human breathing or multipath fading from surrounding metal, statistically, the maximum range of instantaneous energy level fluctuations caused by this static stray noise is strictly limited to a closed-loop range of 3.0 dB. Therefore, 3.0 dB is set as a safety protection boundary; below this value, the tag can be considered spatially stationary.

[0151] On-site dynamic adjustment mechanism: This 3.0dB is the system's default energy threshold. In actual engineering implementation, the reflective metallic materials inside the vehicle body, the distribution of glass windows, and varying intensities of external electromagnetic environments all affect the fluctuation range of multipath noise. On-site technicians can flexibly adjust this energy difference threshold (e.g., to 2.5dB or 4.0dB) based on the real-time noise floor of the onboard RFID reader and the RF fading model specific to the vehicle structure, thus perfectly adapting to the on-site electromagnetic environment of various bus doors.

[0152] Example 1: Single passenger boarding scenario

[0153] Passenger A enters the vehicle from outside, carrying an RFID tag:

[0154] The outer antenna (4) scans the tag 5 times continuously, time: t1~t5, RSSI range: -45~-60dBm;

[0155] The inner antenna (1) scanned the tag 4 times in a row, with a time interval of t6 to t9 and an RSSI range of -50 to -65 dBm.

[0156] The car door closes, triggering the boarding calculation.

[0157] The system extracts the earliest sequence number 4 [t1~t5] and the latest sequence number 1 [t6~t9].

[0158] Time difference = t9 - t1 = 1500ms > 1000ms;

[0159] Once boarding is confirmed, the tag status is updated from OUTSIDE to INSIDE.

[0160] rfidIn++, the number of people in the car increases by 1.

[0161] Example 2: Scene of loitering at the doorway

[0162] Passenger B lingered at the door without entering.

[0163] The outer antenna 4 scanned the tag multiple times, with an RSSI range of -55 to -57 dBm (difference of 2 dB < 3 dB).

[0164] The car door closes, triggering the boarding calculation;

[0165] Only one candidate, and the RSSI difference is < 3dB;

[0166] The system determines that the position is "not moving at the door" and skips this calculation.

[0167] The count remains unchanged.

[0168] Example 3: Multiple people boarding the vehicle simultaneously

[0169] Passengers C and D board the bus at the same time:

[0170] Passenger C: 5 times in sequence 4 (RSSI difference 12dB), 4 times in sequence 1 (RSSI difference 10dB);

[0171] Passenger D: Sequence 4, 3 times (RSSI difference 5dB); Sequence 1, 2 times (RSSI difference 3dB);

[0172] Both individuals were determined to have boarded the bus by the main algorithm;

[0173] The system selects passenger C with the larger RSSI difference (12dB > 5dB);

[0174] This trigger only records passenger C boarding; passenger D will wait for the next trigger.

Claims

1. A passenger flow priority decision-making method based on dual-antenna RFID and RSSI, characterized in that, Includes the following steps: An outer antenna and an inner antenna are deployed at the vehicle door. The outer antenna is installed on the step or door frame outside the door, and the inner antenna is installed inside the vehicle compartment. The reader device collects the antenna port, RSSI signal strength, and timestamp data of the RFID tags carried by passengers; triggers detection and snapshot creation. The algorithm program module uses a dual-antenna sequence direction analysis method to determine the direction of getting on and off the vehicle.

2. The passenger flow decision-making method based on dual-antenna RFID and RSSI priority as described in claim 1, characterized in that, The dual-antenna sequence direction analysis method is the main algorithm, which includes the following steps: Boarding direction determination: The algorithm module extracts the earliest continuous outer antenna sequence and the latest continuous inner antenna sequence. If the timestamp of the latter is greater than that of the former and the time difference is ≥1000ms, then it is determined to be boarding. Disembarkation determination: Extract the earliest continuous inner antenna sequence and the latest continuous outer antenna sequence. If the timestamp of the latter is greater than that of the former and the time difference is ≥1000ms, then it is determined as disembarkation.

3. The passenger flow decision-making method based on dual-antenna RFID and RSSI priority as described in claim 1, characterized in that, It also includes RSSI signal strength analysis steps: When there is only one candidate label, calculate its RSSI difference range; If the RSSI difference is less than the preset threshold of 3.0dB, it is determined as "no movement at the door" and the calculation is skipped.

4. The passenger flow decision-making method based on dual-antenna RFID and RSSI priority as described in claim 1, characterized in that, When the main algorithm fails and triggers the degradation algorithm, the degradation algorithm includes detecting whether there are two consecutive target antenna scan records; if there are multiple sequences that meet the conditions, the sequence with the largest RSSI difference is selected.

5. The passenger flow decision-making method based on dual-antenna RFID and RSSI priority as described in claim 4, characterized in that, The specific boundary conditions for triggering the degradation algorithm include the following four categories: Insufficient record count boundary: The total number of valid scan records in the tag sliding queue is less than the minimum threshold, and the amount of data is insufficient to support dual-antenna sequence reconstruction; Single-sided data missing boundary: When a passenger passes through, one side of the antenna does not read the signal at all due to human body absorption or metal shielding, resulting in a break in the timing chain of the dual antennas; Temporal disorder / abnormal boundary: The earliest extracted sequence timestamp is greater than the latest sequence timestamp, and the temporal logic has serious overlap or inversion; Loitering deadlock boundary: Although the temporal sequence is satisfied, the time difference calculated from both sides of the sequence is less than the critical anti-loitering threshold. The main algorithm judges it as "loitering at the door" for the purpose of high-precision anti-interference.

6. The passenger flow priority decision-making method based on dual-antenna RFID and RSSI according to claim 1, characterized in that, It also includes a four-level prioritization method for decision trees, comprising the following steps: First priority level: Algorithm determines the success level decision; Decision logic: The system first evaluates the algorithm output results of each candidate snapshot and prioritizes the selection of candidate snapshots that have been successfully determined through dual-antenna sequence direction analysis; Flow control: If there are candidates that the main algorithm determines are successful, they are directly sent to the second-level evaluation as the core candidate set. At this time, the candidate snapshots that the downgrade algorithm determines are suspended. Only when the main algorithm of all candidate snapshots fails and returns UNKNOWN will the second-level evaluation open the snapshots that the downgrade algorithm determines are successful. Second-level priority: RSSI energy fluctuation explicit decision; Decision logic: If multiple candidate snapshots still exist in parallel under the same algorithm success level, the system extracts the range of signal strength difference between each snapshot; Flow control: The decision tree prioritizes the candidate snapshot with the largest RSSI difference; the larger the RSSI difference, the more significant the movement trajectory and activity of the label when crossing the physical door beam coverage area, thus accurately locating the primary target with the most obvious current movement characteristics. Third-level priority: Time window timeliness convergence constraint; Decision logic: If an extreme parallel situation occurs after the second-level energy assessment, the system invokes the timeliness constraint to achieve convergence; Flow control: The system compares the last update timestamp of each candidate snapshot and prioritizes the snapshot whose last update time is closest to the current door sensor trigger cutoff time, thereby eliminating outdated dynamic interference left over from past sliding windows; Fourth priority level: Single-object dynamic verification and static removal protection; Decision logic: When the decision tree filters layer by layer and finally locks in or only has 1 valid candidate snapshot left, the system performs mandatory position state machine transition compliance checks and static object discrimination. Static filtering at the entrance: If the RSSI difference of the only candidate snapshot is less than the preset fluctuation threshold, the decision tree directly starts interrupt protection, determines it as "a person standing still at the door when the door opens", completely removes it from the pass queue, skips this calculation, and ensures the consistency of atomic counter data.

7. A system for prioritizing passenger flow using dual-antenna RFID and RSSI based on the method of any one of claims 1 to 6, characterized in that, The system includes: a reader device, which includes: an outer antenna, an inner antenna, an RFID reader / writer, a camera, and an algorithm program module; the algorithm program module includes: a data processing module and a counting and statistics module.

Images