A multi-source cooperative early warning control system and method for cycling safety

Abstract

The application discloses a kind of multi-source collaborative early warning control systems and methods for cycling safety.Vehicle-mounted sensing terminal collects at least three kinds of heterogeneous data in physiological, motion state, posture, environment, team and external target and pre-processes;Risk determination module extracts risk characteristics, generates risk score using dynamic weight normalization algorithm, divides risk level and determines event type;Strategy generation unit generates tail light strategy including light frequency, brightness and light language mode accordingly;Intelligent tail light terminal outputs matched warning light language;When normal, report abnormality through cellular module, perform local high-reliability cache when network interruption, after network recovery, preferentially slice upload dangerous level event through cascading supplement transmission mechanism;Remote monitoring platform realizes vehicle fleet cascading early warning and collaborative intervention based on formation topology.The application realizes the integration of local tail light early warning and remote platform monitoring, significantly improves the safety and data integrity of cycling and vehicle fleet formation training.

Description

Technical Field

[0001] This invention belongs to the fields of cycling safety, intelligent traffic assistance systems and Internet of Things sports monitoring technology, specifically relating to a multi-source collaborative early warning control system and method for cycling safety. Background Technology

[0002] With the popularization of cycling and the continuous improvement of training intensity of professional cycling teams (such as road cycling teams and triathlon training teams), road traffic safety and exercise physiological safety during cycling have received widespread attention.

[0003] Existing cycling safety aids can be broadly categorized into two types: The first type consists of local safety warning devices, such as smart taillights equipped with accelerometers. These taillights can sense the vehicle's deceleration and increase their brightness during braking to warn following vehicles or teammates. Alternatively, radar taillights with rearward millimeter-wave radar can detect the approaching distance and relative speed of vehicles behind and send a warning to the front-end cycling computer. The second type comprises motion data monitoring devices, such as smart cycling computers, heart rate monitors, dual-sided power meters, and cadence sensors. These devices primarily collect the cyclist's physiological indicators and exercise performance, uploading the data to the cloud for instructors or cyclists to review after training.

[0004] However, these two existing technologies have the following significant technical drawbacks in practical applications:

[0005] 1. Limited Risk Identification Dimensions and Prone to Misjudgment: Traditional smart taillights rely solely on a single deceleration sensor to trigger the brake light. However, in real-world riding scenarios, non-braking behaviors such as road bumps, driving over speed bumps, or pushing the bike can easily cause strong disturbances in the accelerometer, leading to frequent false triggering of the taillights. Furthermore, this single hardware component cannot detect physiological risks such as extreme fatigue or cardiac overload experienced by the rider, nor can it detect the potential risk of falling due to lateral tilting and instability.

[0006] 2. Disconnection and lack of coordination between local early warning and remote monitoring: In existing technologies, the physiological and power consumption data (such as heart rate and power) collected by bike computers are only used for sports performance analysis and are not transformed into control variables to guide road safety. Local taillights cannot detect when a rider is physically exhausted, and therefore cannot proactively change their light signals to warn following vehicles to give way. In addition, in the context of team training, although the coach can see some riders' data, when a rider experiences a sudden crash, tire blowout, or emergency braking, the risk cannot be cascaded within the team in seconds. Riders following closely behind often experience large-scale chain-reaction crashes due to the extremely close following distance of road bikes (usually less than 50cm).

[0007] 3. Loss of critical safety data in offline environments: In cycling environments with poor cellular network signals or intermittent network outages, such as mountainous or suburban areas, existing network monitoring systems typically lose all telemetry data during the outage period. This means that when a cyclist has a serious accident in an offline area (such as a crash and loss of consciousness), the remote monitoring platform neither receives an alarm nor can it obtain physiological, power, and posture slice data of the cyclist in the seconds before the accident, making it impossible to accurately reconstruct the cause of the accident.

[0008] The information disclosed in this background section is intended only to enhance the understanding of the overall background of the invention and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0009] The purpose of this invention is to provide a multi-source collaborative early warning control system and method that can deeply integrate six categories of heterogeneous data, including physiological, motion, posture, environment, queue, and external target data, and realize local refined light signal early warning, convoy cascaded closed-loop intervention, and highly reliable cascaded supplementary transmission mechanism. This is to solve the problems of existing cycling safety equipment having a single risk dimension, a disconnect between local early warning and remote monitoring, a lack of cascaded early warning in convoy training, and the easy loss of critical emergency data in offline environments.

[0010] To address the aforementioned technical problems, this invention provides a multi-source collaborative early warning and control method for cycling safety, the method comprising the following steps:

[0011] S01. Multi-source data acquisition and preprocessing: By connecting to cycling peripherals and onboard sensors through an onboard sensing terminal installed on the cycling vehicle, at least three types of heterogeneous data are collected and received from the following data: cyclist physiological data, vehicle motion state data, posture data, environmental data, team data, and external target data; and the heterogeneous data are preprocessed to obtain preprocessed cycling data.

[0012] S02, Multi-source risk feature extraction: Based on the preprocessed cycling data, extract risk features related to cycling safety within a preset time window;

[0013] S03, Cycling Risk Fusion and Determination: The risk characteristics are input into the risk determination module for fusion analysis. A cycling risk score is generated based on the dynamic weight normalization algorithm. The cycling risk level is divided according to the cycling risk score. At the same time, the risk event type is determined by combining the dynamic changes of various risk characteristics.

[0014] S04. Taillight warning strategy generation: Based on the riding risk level and the risk event type, dynamically calculate and generate a taillight control strategy that includes light emission frequency, brightness and light signal mode.

[0015] S05, Local Taillight Warning and Abnormal Event Reporting: The taillight control strategy is sent to the intelligent taillight terminal, which outputs a warning light message that matches the riding risk level and risk event type. When the communication network is normal, the risk event containing the risk event type, event timestamp, and event fragment data is reported to the remote monitoring platform through the communication module.

[0016] S06. Remote Collaborative Monitoring and Intervention: The remote monitoring platform receives cycling data and risk events uploaded by one or more vehicle-mounted sensing terminals, generates a monitoring status of individual cycling status or team training status, and issues intervention suggestions to the corresponding vehicle-mounted sensing terminals based on the cycling risk level and risk event type, or issues team cascade warning instructions to the vehicle-mounted sensing terminals of related cycling vehicles in the same formation.

[0017] Preferably, in the technical solution, in step S01: the cyclist's physiological data includes at least one of heart rate, heart rate change rate, blood oxygen saturation, body temperature, or heart rate recovery rate after exercise; the vehicle's motion status data includes at least one of real-time speed, acceleration, deceleration, current cadence, real-time output power, cumulative cycling mileage, or current continuous cycling duration; the posture data includes at least one of three-axis tilt angle, yaw rate, vehicle vibration intensity, or posture disturbance frequency; the environmental data includes at least one of ambient light intensity, ambient temperature, current weather conditions, or current road gradient; the team data includes at least one of the registered order of team members, relative distance between vehicles in front and behind, formation density, or overall speed distribution of the formation; the external target data includes at least one of the relative speed, azimuth angle, relative distance, or collision risk approach trend of the target approaching from behind or to the side of the vehicle.

[0018] Preferably, in the technical solution, in step S01: the vehicle-mounted sensing terminal connects to the riding peripheral and the vehicle-mounted sensor via at least one of the following: ANT+ protocol, Bluetooth Low Energy BLE protocol, Wi-Fi or UWB communication method;

[0019] The preprocessing includes: denoising the sensor data using Gaussian filtering or median filtering; interpolating and compensating for missing data; mapping heterogeneous data to the same time axis for timestamp alignment; and performing individualized threshold interval mapping based on the cyclist's maximum heart rate and threshold power.

[0020] Preferably, in the technical solution, in step S03, the cycling risk score R is calculated using the following weighted model:

[0021] (1),

[0022] in, Indicates physiological risk factors, Indicates vehicle motion risk factors. Indicates attitude stability risk factor, Indicates environmental risk factors, Indicates the team's status risk factor. This indicates that external targets are close to risk factors. to Let be the initial weight coefficients of each factor, and satisfy:

[0023] ,

[0024] in, Let be the initial weighting coefficients for the i-th type of cycling risk factor;

[0025] When certain major sensor categories are not connected (e.g., no heart rate monitor is being worn or the vehicle lacks radar taillights), the system activates an automatic weight normalization adjustment mechanism. Assuming the j-th factor is unavailable, then let... The new weight coefficients of the remaining available factors Adjusted to:

[0026] ,

[0027] in, The initial weight coefficients are used for the k-th available cycling risk factors, thereby ensuring the robustness of the entire fusion judgment system and avoiding system failure due to incomplete hardware.

[0028] When the time to rear-end collision (TTC) is less than the preset collision time threshold, and the vehicle's riding risk score (R) is in the unsafe range, the risk event type is determined to be a compound proximity risk.

[0029] Preferably, in the technical solution, the dual or multiple criteria for determining the type of risk event in step S03 are:

[0030] Risk of sudden deceleration: the magnitude of the decrease in real-time speed and ,in, The preset threshold for the rate of decrease in speed. For the real-time deceleration of the vehicle, This is the preset deceleration threshold;

[0031] Fatigue risk: Heart rate remains above the individual fatigue heart rate threshold, and the variance of cadence increases beyond the preset stability threshold while the current real-time output power decreases.

[0032] Attitude instability risk: The frequency of vehicle attitude disturbance or the fluctuation range of the three-axis tilt angle exceeds the preset instability threshold;

[0033] Risk of falling behind: The relative distance between this vehicle and the vehicle in front in the platoon continues to increase and the growth rate exceeds the preset threshold;

[0034] Composite proximity risk: The time to rear-end collision (TTC) is less than the preset collision time threshold, and the riding risk score (R) of this vehicle is in the unsafe range.

[0035] Preferably, in the technical solution, in step S05, the local taillight warning and abnormal event reporting further includes a local network outage caching and retransmission mechanism: when the communication network is detected to be in an interrupted or weak connection state, the vehicle-mounted sensing terminal writes the original data, risk characteristics, riding risk level, and risk event type into a local high-reliability non-volatile memory for caching in real time, and attaches a danger level label or a normal level label to the risk event; when the communication network is detected to have resumed normal connection, the vehicle-mounted sensing terminal starts a cascaded retransmission mechanism: it prioritizes uploading high-frequency, high-precision data slices within a preset time period before and after the occurrence of the risk event with the danger level label to the remote monitoring platform; after the danger level label data is uploaded, the ordinary training data summary with the normal level label is asynchronously uploaded to the remote monitoring platform.

[0036] Preferably, in the technical solution, in step S06, the logic for issuing the convoy cascade warning command is as follows: when the risk event type reported by the on-board sensing terminal of a specific riding vehicle in the convoy is a sudden deceleration risk, attitude instability risk, or compound approach risk, and the corresponding riding risk level is dangerous, the remote monitoring platform automatically identifies at least one associated riding vehicle behind the specific riding vehicle and whose relative distance is less than the safe braking distance according to the pre-registered convoy topology, and simultaneously issues a cascade warning command to the on-board sensing terminal and intelligent taillight terminal of the associated riding vehicle, forcing its intelligent taillight terminal to switch to a high-frequency flashing warning state, and issuing a yielding prompt to the team members behind through the display or sound interface of the on-board sensing terminal.

[0037] This invention also provides a multi-source collaborative early warning and control system for cycling safety, the system comprising:

[0038] Vehicle-mounted sensing terminal: fixed on the cycling vehicle, including a processor, memory, wireless sensing interface module and cellular communication module; the wireless sensing interface module is used to connect to cycling peripherals and vehicle-mounted sensors;

[0039] Intelligent taillight terminal: installed at the rear of the bicycle, it communicates with the vehicle-mounted sensing terminal via a wireless communication module. It includes a main control unit and an LED light-emitting component, and is used to receive the taillight control strategy sent by the vehicle-mounted sensing terminal and output the corresponding warning light message.

[0040] Risk assessment module: integrated into the processor of the vehicle-mounted sensing terminal, or distributed between the vehicle-mounted sensing terminal and the remote monitoring platform, used to preprocess, extract features and perform dynamic weight normalization fusion analysis on the collected heterogeneous data to generate cycling risk score, cycling risk level and risk event type;

[0041] Remote monitoring platform: Deployed on a cloud server, it receives data uploaded by one or more vehicle-mounted sensing terminals through cellular communication modules and wireless networks. It has a fleet collaborative analysis engine that is used to render the fleet training situation map in real time and to calculate and distribute fleet cascade early warning commands.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] 1. Achieved deep cross-domain integration of multi-dimensional heterogeneous data, eliminating blind spots in misjudgment: Breaking down the barriers between traditional safety hardware and competitive training hardware, it integrates physiological (heart rate, blood oxygen), power consumption (power, cadence), physical state (attitude, deceleration), as well as environmental and radar data. The introduced dynamic weight normalization algorithm can not only ensure adaptive operation when sensors are incomplete, but also completely eliminate false alarms caused by road bumps caused by a single accelerometer by using multi-dimensional criteria (such as "heart rate + power + cadence" to judge fatigue, and "deceleration + speed change" to judge emergency braking).

[0044] 2. A closed-loop cascaded collaborative early warning mechanism was constructed, connecting local to remote and individual to convoy levels. This mechanism not only achieved a local closed loop of "data anomaly → adaptive switching of taillight signals (e.g., switching from normal flashing to 8Hz ultra-strong pulse flashing)", but also enabled convoy-level cascaded early warning. When a lead rider or platoon leader encounters danger, the cloud can locate the following riders within milliseconds and forcibly activate their early warning terminals, effectively preventing catastrophic chain-reaction crashes during high-density road bike convoy training.

[0045] 3. An innovative cascaded network outage retransmission mechanism based on safety tags was designed to ensure the integrity of the review of extreme accidents: In response to the pain point of frequent network outages during outdoor cycling, the priority slice retransmission mechanism proposed in this invention can ensure that the most critical dangerous event snapshots are presented to the monitoring end first within the golden time of network recovery. This solves the technical problem of large-capacity raw telemetry data blocking the transmission channel in a weak network environment, and provides tamper-proof ironclad evidence for subsequent medical rescue positioning and accident technical analysis. Attached Figure Description

[0046] Figure 1 This is a schematic diagram of the overall system architecture and network topology provided in the embodiments of the present invention;

[0047] Figure 2This is an overall flowchart of the multi-source collaborative early warning and control method for cycling safety provided in an embodiment of the present invention;

[0048] Figure 3 This is a block diagram of the internal hardware circuit of the vehicle-mounted sensing terminal in an embodiment of the present invention;

[0049] Figure 4 This is a block diagram of the internal hardware circuit of the intelligent taillight terminal in an embodiment of the present invention;

[0050] Figure 5 This is a logic flow diagram of the risk determination module performing dynamic weight normalization and event determination in an embodiment of the present invention.

[0051] Figure 6 This is a schematic diagram of the multi-vehicle collaboration and cascaded early warning process in a fleet training scenario according to an embodiment of the present invention. Detailed Implementation

[0052] The specific embodiments of the present invention will be described in detail below, but it should be understood that the scope of protection of the present invention is not limited to the specific embodiments.

[0053] Unless otherwise expressly stated, throughout the specification and claims, the term "comprising" or its variations such as "including" or "comprises" shall be understood to include the stated elements or components without excluding other elements or other components.

[0054] Example 1: Hardware Configuration of a Multi-Source Cooperative Early Warning and Control System for Cycling Safety

[0055] like Figure 1 , 3 As shown in Figure 4, this embodiment fully and thoroughly discloses the hardware topology, coupling, and electrical / wireless interconnection relationships between the various sub-functional modules in the multi-source collaborative early warning control system for cycling safety. The entire system is physically and logically divided into a vehicle-side perception and control layer, a rear-end warning execution layer, and a cloud-based collaborative monitoring layer.

[0056] The vehicle-mounted sensing terminal 200 is fixed in a prominent position on the handlebars or frame of the bicycle, serving as the centralized control hub for the vehicle. Its internal core utilizes a high-density integrated main control circuit board, on which a microprocessor 209 and a memory 210 electrically connected via a high-speed local bus are mounted. The microprocessor 209 preferably employs a microcontroller chip with a hardware floating-point unit (FPU) and a high-frequency ARM Cortex-M4 or higher architecture to support parallel real-time computation of multiple heterogeneous data streams. The memory 210 consists of high-speed SRAM and non-volatile flash memory. The microprocessor 209 also integrates the following core physical sub-modules on its peripheral hardware bus:

[0057] Wireless sensor interface module 201: It consists of a multi-mode single-chip RF circuit integrating ANT+, Bluetooth Low Energy (BLE), and Ultra Wideband (UWB) communication protocols, with an external onboard antenna. Wireless sensor interface module 201 establishes a bidirectional high-frequency signal connection with microprocessor 209 via SPI or UART bus. It is specifically responsible for establishing long-term local area network connections with various external RF cycling peripherals such as heart rate monitor 101, dual-sided power meter 102, cadence sensor 103, and speed sensor 104, continuously monitoring and capturing physiological electrical signals, torque power consumption signals, rotational speed pulse signals, and wheel speed signals.

[0058] Data acquisition module 202 and data preprocessing unit 203: The data acquisition module 202 is physically connected to the six-axis inertial attitude sensor 105 (including a three-axis accelerometer and a three-axis gyroscope) and the miniature ambient light sensor 106 integrated inside the vehicle-mounted sensing terminal 200, and performs high-order A / D conversion on the output analog signals. The data preprocessing unit 203 utilizes the computing core of the microprocessor 209 to open an independent FIFO data buffer queue, and performs Gaussian filtering / median filtering for noise reduction, linear interpolation compensation, missing value calibration, and millisecond-level alignment across hardware time axes on all heterogeneous raw data streams from the wireless sensing interface module 201 and the data acquisition module 202.

[0059] Feature extraction unit 204 and risk scoring unit 205 together constitute the core of the risk assessment module embedded in the vehicle. Feature extraction unit 204 runs as firmware in microprocessor 209 and is responsible for capturing the time and frequency domain features of data in real time using a preset time sliding window (e.g., a 3-second sliding window). Risk scoring unit 205 stores the mathematical model of the dynamic weight normalization algorithm described in claim 5. In the abnormal situation where some sensors are not connected or disconnected, it performs zeroing of the available factor weights and automatic re-normalization calculation, and outputs a dynamic riding risk score.

[0060] Control strategy generation unit 206: The input end is connected to risk scoring unit 205, and the output end is connected to wireless sensor interface module 201. Based on risk level and risk event criteria, it encodes refined driving instructions for the intelligent taillights in real time (including specific PWM light emission frequency, brightness level, and flashing light code).

[0061] Local cache module 207 and cellular communication module 208: The local cache module 207 includes a dedicated high-reliability, non-volatile, high-capacity NOR / NAND Flash memory chip, which communicates directly with the microprocessor 209 via a hardware QSPI interface. Internally, it features a streamlined file index management system, providing highly reliable and secure data writing during cellular network outages, and adding corresponding entity file retrieval tags for dangerous and normal events. The cellular communication module 208 preferably uses a highly robust 4G / 5G IoT Cat-1 or Cat-4 cellular communication chip, connected to an external high-gain active LTE antenna, supporting TCP / IP or UDP protocol stacks, and is responsible for establishing a full-duplex, bidirectional, long-term IoT connection between the vehicle-mounted sensing terminal 200 and the cloud platform.

[0062] The intelligent taillight terminal 300 is independently fixed to the underside of the seat bracket or rear rack of the bicycle via a locking bracket. Its internal hardware circuitry mainly includes a wireless communication module 301, a core control unit 302, an LED driver circuit 303, an LED light-emitting component 304, an environmental compensation unit 305, and a power management module 306.

[0063] The main control unit 302 establishes a high-speed point-to-point Bluetooth wireless connection with the wireless sensing interface module 201 of the vehicle sensing terminal 200 through the wireless communication module 301 (BLE / UWB RF chip) to receive taillight control strategies.

[0064] The LED driver circuit 303 uses a multi-channel high-frequency adjustable constant current driver chip, whose control pins are directly connected to the hardware PWM output interface of the main control unit 302 to eliminate flickering during high-brightness illumination. The LED light-emitting component 304 consists of an array of multiple high-power red LEDs with different lens polarization angles;

[0065] The environmental compensation unit 305 is electrically connected to the photodiode fixed on the surface of the taillight housing, and transmits the external ambient light intensity in real time, so that the main control unit 302 can perform peak clipping or gain control on the driving constant current value according to the day and night illuminance.

[0066] The power management module 306 integrates a high-energy-density lithium battery pack, a charging management IC, and a precision coulomb counter chip, and includes a microampere-level ultra-low-power standby wake-up circuit controlled by the main control unit 302.

[0067] In addition, the intelligent taillight terminal 300 also integrates a rear target perception module 107 (which includes a miniature 77GHz or 24GHz vehicle-mounted short-range millimeter-wave radar, facing directly behind the vehicle through an antenna cover). The signal processing end of the rear target perception module 107 establishes communication with the main control unit 302 to wirelessly transmit the collision time (TTC) data of the approaching object behind to the vehicle-mounted perception terminal 200 in real time.

[0068] The remote monitoring platform 400 is deployed on a distributed cloud server cluster, and its core runs a fleet collaborative analysis engine with high throughput and low latency response characteristics. The platform maintains long-term connection socket or MQTT sessions with a large number of vehicle-side cellular communication modules 208 through a multi-path load balancer and an external internet communication network 600. The internal hardware storage cluster of the monitoring platform 400 pre-registers static topology distance calculation submodules and queue member association databases for each fleet. The platform's output end interacts with the coach / management end 500 (which can be represented as a mobile tablet, a multi-screen display in the control room, or a mobile APP client) via a secure and encrypted internet interface, enabling cloud-based cascaded command calculation, global training situation rendering, and multi-vehicle closed-loop intervention control.

[0069] Example 2: Risk Assessment and Local Light Signal Adaptive Control in Personal Cycling Scenarios

[0070] like Figure 2 , 5 As shown, the core control logic of the vehicle-mounted sensing terminal 200 in a personal independent riding scenario is as follows:

[0071] After the system starts up (step S01), the processor acquires data through ANT+ and the built-in sensor; in this embodiment, the cyclist sets his maximum heart rate to 190 bpm, lactate threshold heart rate VT2 to 168 bpm, and threshold power (FTP) to 250 W.

[0072] During the ride, the risk assessment module performs the fusion analysis in step S03. Assuming the cyclist is not wearing external blood oxygen and body temperature monitoring devices, the system automatically triggers a weight reallocation.

[0073] The initial weights are set as follows:

[0074] W1 (physiological) = 0.25, W2 (motor) = 0.20, W3 (posture) = 0.2, W4 (environment) = 0.05, W5 (team) = 0.10, W6 (external goal) = 0.2;

[0075] Since the current situation is individual cycling and some peripherals are unavailable, the data for the team status risk factor T is unavailable. Therefore, W5 = 0, and the weights of the remaining available factors are proportionally normalized and amplified according to the formula. For example, the adjusted physiological risk weight becomes:

[0076] ,

[0077] When the following specific scenario combinations occur, the system's event judgment and local light signal execution are as follows:

[0078] Scenario A (Encountering sudden emergency braking): The triaxial accelerometer in the IMU detects longitudinal deceleration. It reached 5.5 m / s 2 (Exceeding the preset first deceleration threshold of 3.5 m / s) 2 Simultaneously, the wheel speed sensor detects a sudden drop in speed from 40 km / h to 15 km / h within 0.5 seconds. At this moment, without relying on any cloud computing, the risk assessment module instantly determines the risk event type as "sudden deceleration risk" and sets the riding risk level to the highest level, "Danger." The onboard sensing terminal immediately sends a control strategy to the intelligent taillight terminal 300 via local BLE: the taillight instantly interrupts its current constant-on mode and switches to a "pulse high-brightness fast-flash mode" (frequency 8Hz, duty cycle 100% maximum rated power output) to provide the strongest visual warning to vehicles behind. This light signal mode is forcibly delayed for 3 seconds after braking to prevent rear-end collisions.

[0079] Scenario B (Extreme Fatigue or Exercise Overload): When a cyclist is climbing a hill for a long time, the heart rate collected by the heart rate monitor remains at 175 bpm (more than 1.2 × VT2 = 1.02 × 168 ≈ 171.3 bpm) for 10 minutes. At this time, the real-time output power fed back by the dual-sided power meter has dropped from 240W to 130W (a drop of more than 20%), and the cadence sensor shows that the cadence has become chaotic from a stable 90 rpm, with the variance increasing by 45%.

[0080] If all multi-dimensional criteria are met, the risk assessment module determines the risk event type as "fatigue risk" and the risk level as "Caution". At this time, the taillight control strategy generation unit combines the data from the ambient light sensor 106 (assuming the current environment is a low-visibility nighttime environment): in order to effectively warn those behind while avoiding glare damage to teammates behind, the taillight adaptively switches to a "breathing low-frequency light emission mode" (frequency 1Hz, brightness transitioning smoothly between 30% and 80% of the maximum value), using a specific light signal to announce to the outside world that the cyclist's physical strength has reached its limit and a safe social distance needs to be maintained.

[0081] Example 3: Cloud-based collaborative intervention and cascaded retransmission in a professional vehicle training scenario

[0082] like Figure 1 , 6 As shown, this embodiment demonstrates collaborative control in a professional road cycling team formation training scenario (such as a long single-file formation or a double-row alternating wind-breaking formation).

[0083] The convoy has 10 registered members, and each vehicle is equipped with the system of this invention. Each vehicle's onboard sensing terminal synchronizes a data stream with a unique device ID and a high-precision GPS timestamp to the remote monitoring platform 400 in the cloud at regular intervals (1Hz) via a cellular network (step S05). The convoy collaborative analysis engine within the remote monitoring platform calculates the relative spatial geometry and topology of all vehicles in real time.

[0084] When vehicle number 1 (the lead windbreaker) was going downhill and cornering, the vehicle skidded severely due to gravel on the road surface. Its onboard sensing terminal's IMU detected the change in yaw rate. The vehicle body tilt angle instantly reached 45°, triggering a "posture instability risk" (risk level: dangerous).

[0085] At this moment, with the formation spacing only 30cm and the speed reaching 50km / h, the team members of vehicle number 1 had no time to verbally shout. Simultaneously, the onboard sensing terminal of vehicle number 1 triggered the taillight hazard warning locally and transmitted a UDP disaster telemetry message containing the "crash / instability hazard tag" to the remote monitoring platform 400 within microseconds.

[0086] Upon receiving the message, the cloud-based remote monitoring platform 400 automatically identified vehicles 2, 3, and 4 as "absolutely associated high-risk vehicles" within 10ms based on the current queue topology. Breaking away from conventional polling mechanisms, the platform proactively broadcast a "fleet cascading warning instruction" to the onboard sensing terminals of vehicles 2, 3, and 4. Upon receiving the instruction, the onboard sensing terminals of vehicles 2, 3, and 4 immediately activated their built-in buzzers to warn of a potential crash ahead, while their intelligent taillights automatically switched to a phased, forced flashing mode to alert team members 5 through 10 further behind to quickly change lanes and avoid the vehicle.

[0087] The specific working process of the network outage recovery mechanism is as follows: When the convoy rode to a deep valley in the mountains, the 4G / 5G network was completely interrupted (the cellular module returned an RSSI below -110dBm). During the network outage, vehicle number 3 unfortunately suffered a tire blowout and braked suddenly, resulting in a crash.

[0088] The onboard sensing terminal 200 of vehicle No. 3 detected a sudden combined event of rapid deceleration and attitude instability. Due to the inability to connect to the network, the local cache module 207 immediately packaged the raw high-frequency sensing data (100Hz sampling rate) of the six-axis IMU for the 30 seconds before and after the event, along with the heart rate / power mutation curve, and stored them separately in a dedicated high-reliability sector of Flash memory, assigning them a "danger level" label. The data from stable riding during other normal time periods were assigned a "normal level" label as ordinary training data.

[0089] When the convoy rode out of the canyon and the cellular network connection was restored (the base station signal-to-noise ratio was detected to meet RSSI > -95dBm and remain stable for 3 seconds), the processor 209 of the vehicle-mounted sensing terminal immediately blocked the reporting channel of ordinary training data and initiated the cascaded retransmission protocol: First, the 30-second high-precision slice of accident data with the "danger level label" in Flash was pushed to the remote monitoring platform 400 with the highest priority. The management software 500 on the coach's end popped up the precise three-axis acceleration waveform, braking deceleration curve, and heart rate and pulse characteristics of car number 3 at the moment of the crash as soon as the network was restored, achieving a 100% complete and lossless review of the extremely dangerous accident under network outage conditions. After confirming that the danger label data was fully verified (ACK reply code was normal), the vehicle-mounted sensing terminal started a low-priority sparse thinning algorithm in the background to asynchronously retransmit the ordinary training summary (such as average heart rate and cumulative mileage) of the remaining tens of minutes during the network outage, saving bandwidth resources to the greatest extent in weak network environments.

[0090] The foregoing description of specific exemplary embodiments of the invention is for illustrative and explanatory purposes. These descriptions are not intended to limit the invention to the precise forms disclosed, and it will be apparent that many changes and variations can be made in accordance with the foregoing teachings. The exemplary embodiments were chosen and described in order to explain the specific principles of the invention and its practical application, thereby enabling those skilled in the art to implement and utilize various different exemplary embodiments of the invention, as well as various different choices and variations. The scope of the invention is intended to be defined by the claims and their equivalents.

Claims

1. A multi-source cooperative early warning control method for cycling safety, characterized in that, The method includes the following steps: S01. Multi-source data acquisition and preprocessing: By connecting to cycling peripherals and onboard sensors through an onboard sensing terminal installed on the cycling vehicle, at least three types of heterogeneous data are collected and received from the following data: cyclist physiological data, vehicle motion state data, posture data, environmental data, team data, and external target data; and the heterogeneous data are preprocessed to obtain preprocessed cycling data. S02, Multi-source risk feature extraction: Based on the preprocessed cycling data, extract risk features related to cycling safety within a preset time window; S03, Cycling Risk Fusion and Determination: The risk characteristics are input into the risk determination module for fusion analysis. A cycling risk score R is generated based on the dynamic weight normalization algorithm. The cycling risk level is divided according to the cycling risk score R. At the same time, the risk event type is determined by combining the dynamic changes of various risk characteristics. S04. Taillight warning strategy generation: Based on the riding risk level and the risk event type, dynamically calculate and generate a taillight control strategy that includes light emission frequency, brightness and light signal mode. S05, Local taillight warning and abnormal event reporting: The taillight control strategy is sent to the intelligent taillight terminal, so that the intelligent taillight terminal outputs a warning light message that matches the riding risk level and risk event type. When the communication network is normal, risk events, including risk event type, event timestamp, and event fragment data, are reported to the remote monitoring platform through the communication module. S06. Remote Collaborative Monitoring and Intervention: The remote monitoring platform receives cycling data and risk events uploaded by one or more vehicle-mounted sensing terminals, generates a monitoring status of individual cycling status or team training status, and issues intervention suggestions to the corresponding vehicle-mounted sensing terminals based on the cycling risk level and risk event type, or issues team cascade warning instructions to the vehicle-mounted sensing terminals of related cycling vehicles in the same formation.

2. The multi-source cooperative warning control method for riding safety according to claim 1, characterized in that, In step S01: the cyclist's physiological data includes at least one of heart rate, heart rate change rate, blood oxygen saturation, body temperature, or heart rate recovery rate after exercise; The vehicle motion status data includes at least one of the following: real-time speed, acceleration, deceleration, current cadence, real-time output power, cumulative riding distance, or current continuous riding duration; The attitude data includes at least one of the following: three-axis tilt angle, yaw rate, vehicle body vibration intensity, or attitude disturbance frequency. The environmental data includes at least one of ambient light intensity, ambient temperature, current weather conditions, or current road slope. The team data includes at least one of the following: the registered order of team members in the same group, the relative distance between vehicles in front and behind, the formation density, or the overall speed distribution of the formation; The external target data includes at least one of the following: relative speed, azimuth angle, relative distance, or collision risk approach trend of the target approaching from behind or to the side of the vehicle.

3. The multi-source collaborative early warning and control method for cycling safety according to claim 1, characterized in that, In step S01: the vehicle-mounted sensing terminal connects to the riding peripheral and the vehicle-mounted sensor via at least one of the following methods: ANT+ protocol, Bluetooth Low Energy BLE protocol, Wi-Fi or UWB communication. The preprocessing includes: denoising the sensor data using Gaussian filtering or median filtering; interpolating and compensating for missing data; mapping heterogeneous data to the same time axis for timestamp alignment; and performing individualized threshold interval mapping based on the cyclist's maximum heart rate and threshold power.

4. The multi-source collaborative early warning and control method for cycling safety according to claim 1, characterized in that, In step S02, the risk characteristics include at least three of the following: heart rate load characteristics, power load characteristics, cadence stability characteristics, speed mutation rate characteristics, maximum deceleration characteristics, vehicle posture disturbance characteristics, ambient low brightness characteristics, relative speed difference between team members characteristics, and rear target collision time (TTC) characteristics.

5. The multi-source collaborative early warning and control method for cycling safety according to claim 1, characterized in that, In step S03, the cycling risk score R is calculated using the following weighted model: （1）， in, Indicates physiological risk factors, Indicates vehicle motion risk factors. Indicates attitude stability risk factor, Indicates environmental risk factors, Indicates the team's status risk factor. This indicates that external targets are close to risk factors. to Let be the initial weight coefficients of each factor, and satisfy: ； When one or more types of heterogeneous data are not connected or are currently unavailable, the risk assessment module resets the weights corresponding to the unavailable factors to 0 and re-normalizes and redistributes the weight coefficients of the remaining available factors so that the sum of the weights of the available factors after re-distribution is still 1.

6. The multi-source collaborative early warning and control method for cycling safety according to claim 5, characterized in that, The risk event type determined in step S03 includes: When the real-time speed decrease exceeds a preset speed decrease threshold, and the vehicle's real-time deceleration exceeds a preset deceleration threshold, the risk event type is determined to be a sudden deceleration risk. When the heart rate is consistently higher than the individual fatigue heart rate threshold, and the variance of cadence increases beyond the preset stability threshold while the current real-time output power decreases, the risk event type is determined to be fatigue risk. When the frequency of vehicle body attitude disturbance or the fluctuation range of three-axis tilt angle exceeds the preset instability threshold, the risk event type is determined to be attitude instability risk; When the relative distance between this vehicle and the vehicle in front in the platoon continues to increase and the growth rate exceeds a preset threshold, the risk event type is determined to be the risk of falling behind. When the time to rear-end collision (TTC) is less than the preset collision time threshold, and the vehicle's riding risk score (R) is in the unsafe range, the risk event type is determined to be a compound proximity risk.

7. The multi-source collaborative early warning and control method for cycling safety according to claim 1, characterized in that, In step S05, the local taillight warning and abnormal event reporting also includes a local network outage caching and retransmission mechanism: When the communication network is detected to be interrupted or in a weak connection state, the vehicle-mounted sensing terminal writes the raw data, risk characteristics, riding risk level and risk event type into the local high-reliability non-volatile memory in real time for caching, and attaches a danger level label or a normal level label to the risk event. When the communication network is detected to have resumed normal connection, the vehicle-mounted sensing terminal initiates a cascaded retransmission mechanism: it prioritizes uploading high-frequency, high-precision data slices within a preset time period before and after the occurrence of a risk event with a hazard level label to the remote monitoring platform; after the hazard level label data is uploaded, it asynchronously uploads ordinary training data summaries with ordinary level labels to the remote monitoring platform.

8. The multi-source collaborative early warning and control method for cycling safety according to claim 1, characterized in that, In step S06, the logic for issuing the convoy cascade warning command is as follows: when the risk event type reported by the on-board sensing terminal of a specific riding vehicle in the convoy is a sudden deceleration risk, attitude instability risk, or compound approach risk, and the corresponding riding risk level is dangerous, the remote monitoring platform automatically identifies at least one associated riding vehicle behind the specific riding vehicle and whose relative distance is less than the safe braking distance according to the pre-registered convoy topology. The platform then simultaneously issues a cascade warning command to the on-board sensing terminal and the smart taillight terminal of the associated riding vehicle, forcing its smart taillight terminal to switch to a high-frequency flashing warning state, and issuing a yielding prompt to the riders behind through the display or sound interface of the on-board sensing terminal.

9. A multi-source collaborative early warning control system for cycling safety, characterized in that, The system is used to execute the multi-source cooperative early warning control method for cycling safety as described in any one of claims 1 to 8, the system comprising: Vehicle-mounted sensing terminal (200): fixed on the cycling vehicle, including processor (209), memory (210), wireless sensing interface module (201) and cellular communication module (208); the wireless sensing interface module (201) is used to connect cycling peripherals and vehicle-mounted sensors; Intelligent taillight terminal (300): installed at the rear of the vehicle, and connected to the vehicle-mounted sensing terminal (200) via a wireless communication module (301). It includes a main control unit (302) and an LED light-emitting component (304), and is used to receive the taillight control strategy sent by the vehicle-mounted sensing terminal (200) and output the corresponding warning light message. Risk assessment module: integrated into the processor of the vehicle-mounted sensing terminal (200), or distributed between the vehicle-mounted sensing terminal and the remote monitoring platform, used to preprocess, extract features and perform dynamic weight normalization fusion analysis on the collected heterogeneous data to generate cycling risk score, cycling risk level and risk event type; Remote monitoring platform (400): Deployed on a cloud server, it receives data uploaded by one or more vehicle-mounted sensing terminals (200) through a cellular communication module (208) and a wireless network. It has a fleet collaborative analysis engine for rendering the fleet training situation map in real time and for calculating and distributing fleet cascaded early warning instructions.

Images