Split type mobile signal device and traffic control method

Abstract

According to the split type mobile signal device provided by the invention, after the mobile body is placed on a target road section, traffic data is transmitted to the edge calculation module based on unmanned aerial vehicle lift-off scanning, and the edge calculation module generates a trolley deployment task to control an intelligent trolley to arrive at a specified position to complete specified work; meanwhile, the edge calculation module generates a timing signal to control the signal lamp module to manage and control the traffic of the target road section; the traffic state is detected in real time through the unmanned aerial vehicle in the traffic control process; in the whole process, the manual participation degree is effectively reduced, and the management efficiency for emergencies is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of traffic management and intelligent traffic control, in particular to a split type mobile signal device and a traffic management and control method. BACKGROUND

[0002] Traffic signal control facilities play an important role in urban traffic management. However, in actual work, many abnormal conditions will occur. For example, not all intersections in urban roads are equipped with traffic signal control facilities, or although the intersection is equipped with traffic signal control facilities, the facilities fail to use. At this time, if a sudden event occurs at the intersection, the existing solution mainly relies on manual setting of temporary traffic signs, cones or on-site traffic command by traffic police. This method has inherent defects such as long response time, low efficiency, high labor cost and high safety risk. Some cities are also promoting the use of mobile traffic signal lights. The existing products are usually single-function vehicle-mounted or towed signal lights, which can only provide basic red-yellow-green three-color signal indication, and lack real-time perception and optimization capability for traffic flow. The timing scheme of these simple mobile signal lights is often fixed or requires manual intervention by on-site personnel, making it difficult to dynamically and autonomously adjust according to real-time traffic flow, event type and impact range.

[0003] With the popularization and use of unmanned aerial vehicles, in recent years, some technical solutions have attempted to introduce unmanned aerial vehicles for traffic monitoring. Patent No. CN113096417A proposes a traffic signal light remote control method based on unmanned aerial vehicles, which collects intersection videos through unmanned aerial vehicles, and personnel issue voice commands, which are processed through voice recognition, convolutional neural networks and language models to generate signal light control instructions, achieving automatic adjustment of traffic signal lights. However, this technical solution is still manually controlled and cannot achieve localized and autonomous decision-making on site, and the efficiency is limited by the speed of human reaction and communication delay. Patent No. CN108922214A proposes an artificial intelligence processing system based on unmanned aerial vehicles and traffic signal lights, which works in coordination with unmanned aerial vehicles, telescopic devices, traffic devices, positioning devices, etc. to command traffic signal lights. However, the unmanned aerial vehicles used in this technical solution are mainly used for traffic signal light transportation and are not used for intelligent control and traffic flow diversion of signal lights.

[0004] In the aspect of intelligent decision-making, edge computing technology has been introduced into the field of intelligent transportation. The patent with publication number CN213276964U proposes an intelligent traffic signal system that combines a camera, an induction coil, and an edge computing module. The edge computing module can run a machine vision model to analyze traffic conditions and dynamically adjust traffic light timing. However, most existing edge computing intelligent traffic systems are deployed at fixed intersections or road segments. For non-pre-set, random, and sudden events that occur outside the coverage range of fixed units, these fixed intelligent systems cannot provide services.

[0005] In summary, the existing mobile traffic signal devices or intelligent traffic systems in the prior art are either simple mobile lights with single function and lack of perception, or are intelligent but limited to fixed deployment locations, or are intelligent but mainly controlled by artificial means. For intersections without traffic signal control facilities or intersections with malfunctioning traffic signal control facilities, once a sudden situation occurs, traffic control is mainly based on artificial deployment of various equipment and then artificial judgment of traffic conditions for management, which is not only low in efficiency but also highly dependent on the individual ability of traffic management personnel. SUMMARY

[0006] To solve the problem of relying mainly on traffic management personnel for artificial control efficiency when a sudden situation occurs at an intersection without traffic signal control facilities, the present application provides a split-type mobile signal device that can quickly respond to sudden road events and make autonomous decisions, reducing the degree of human involvement and effectively improving the management efficiency for sudden situations. The present application also discloses a traffic management and control method based on the split-type mobile signal device.

[0007] The scheme of the present application is as follows: a split-type mobile signal device, which includes a mobile body and a signal light module, characterized in that it further includes: a drone module, an edge computing module, and an intelligent car module; The mobile body has a mobile chassis with a multi-wheel structure for carrying all components; The signal light module is installed above the mobile body, and the signal light module includes at least four groups of standard-sized red, yellow, and green traffic signal lights; The drone module includes a drone and a drone parking bay; the drone parking bay is provided on the top of the mobile body; the drone module is equipped with monitoring equipment; The edge computing module is integrated within the mobile unit, incorporating a high-performance industrial-grade computing unit. This module includes pre-set real-time target detection models for vehicle identification, static information recognition for traffic data, and spatiotemporal hybrid road event detection algorithms. Based on the identification results and a pre-set strategy library, it generates real-time signal control and intelligent vehicle deployment schemes. The edge computing module also integrates a mobile communication module and a V2X communication module, supporting data transmission back to the cloud, remote control, and interconnection with field devices. Furthermore, it features a high-precision GNSS positioning system. The intelligent vehicle module includes: a drawer-type vehicle storage compartment and an intelligent vehicle. The vehicle storage compartment is located at the lower part of the mobile body above the mobile chassis. At least two intelligent vehicles are placed in the vehicle storage compartment, and each intelligent vehicle is equipped with a portable device. The portable device includes: an intelligent traffic cone, a warning sign, or a fire extinguisher. The traffic light module, the drone module, and the smart car module are all communicatively connected to the edge computing module.

[0008] Its further features are: The vehicle storage compartment includes at least two layers of vehicle storage space, which are arranged from top to bottom within the interior of the vehicle storage compartment; the size of each layer of vehicle storage space is adapted to the size of the smart car. The trolley storage compartment also includes: a sliding door and a lifting mechanism; The lifting mechanism includes a lead screw, a guide nut, and a drive motor. The two lead screws are respectively set vertically on both sides of the entrance gate of the trolley storage compartment. The bottom end of each lead screw is rotatably connected to the base of the trolley storage compartment, and the top end is connected to the output end of the drive motor. The guide nut is threadedly fitted onto the outer circumference of the lead screw. A hinge is provided at the top of the sliding door, corresponding to the two guide nuts; one end of the sliding door is rotatably connected to the guide nuts via the hinges; the rotation axis of the hinges is set along the horizontal direction. The entrance to the trolley storage compartment is provided with a guide ramp, and the lower edge of the guide ramp is located on the outer side; Rollers are installed at the bottom of the sliding door; The storage compartment of the vehicle is also equipped with a magnetic charging base. The sliding door has two guide rail grooves, the distance between which is adapted to the spacing of the rollers at the bottom of the smart vehicle. The position of the guide rail grooves on the sliding door is adapted to the position of the magnetic charging base of the smart vehicle in the storage compartment.

[0009] A traffic control method based on a split-type mobile signal device, characterized by comprising the following steps: S1: Event triggering and device movement; Upon receiving a traffic incident alert, the split-type mobile signaling device is deployed to the area where the traffic incident occurs; The deployment points of the split-type mobile signal device in the area where the traffic incident occurred include: intersections or road sections; S2: Holographic scanning and real-time sensing; The split-type mobile signal device is activated, and the edge computing module controls the drone to take off. The drone scans the area where the traffic incident occurred based on its own deployed monitoring equipment, generating a point cloud topology map of the incident area. The edge computing module uses a spatiotemporal hybrid road event detection algorithm to identify the road segment and vehicle where the traffic incident occurred. S3: Dynamically generate intersection timing schemes: The edge computing platform dynamically generates signal timing schemes by identifying event types and their impact range in real time; the event types include: breakdowns, road flooding, or exit overflows; S4: Coordinated deployment of road management and intelligent vehicles; The intelligent vehicle quickly leaves the warehouse and autonomously travels to the designated location for deployment based on a spatiotemporal constraint path planning algorithm. S5: Real-time iterative optimization: During event handling, the drone continuously monitors and the edge computing platform adjusts the signal timing and the location of the smart car in real time; S6: After the incident is resolved, the drone and intelligent vehicle automatically return to their storage location, and the separate mobile signaling device is removed.

[0010] Its further features are: Step S2 includes the following steps in detail: S21: Generate point cloud topology; Representing the event region video stream using a spatiotemporal state matrix generates a point cloud topology map of the event region [f] mn ]; ; In the formula, f mn It is the HSV color space expression for the matrix points; S22: Static information recognition; The static information includes: lane lines θ1, water accumulation θ2, flames θ3, and debris θ4; A static information recognition model is constructed based on a neural network. After identifying lanes in the traffic incident area based on the static information recognition model, a static information topology layer is generated. This static information topology layer includes: a first lane static information topology layer [L]. mn ], Second topological layer in water accumulation state [W mn ], third topology layer in fire state [H mn], fourth topological layer of the spill [TH mn ]; ; In the formula, F(f) mn ;δ) is the static information recognition model, where δ is the model parameter; This represents identifying the object category θ at the matrix position (m,n). i The predicted probability relative to the background class θ0; The ArgMax operation is used to retrieve a probability vector. The category index with the highest probability is used as the predicted label for that pixel; If the predicted probability of lane line θ1 is higher than the predicted probability of θ0, then L mn =1, otherwise L mn =0; If the predicted probability of water accumulation θ2 is higher than the predicted probability of water accumulation θ0, then W mn =1, otherwise W mn =0; If the predicted probability of flame θ3 is higher than the predicted probability of flame θ0, then H mn =1, otherwise H mn =0; If the predicted probability of the spill θ4 is higher than the predicted probability of θ0, then TH mn =1, otherwise TH mn =0; S23: Dynamic information recognition; A real-time object detection model is constructed based on a convolutional neural network model. The input of the real-time object detection model is the point cloud topology [f] mn The output is the bounding boxes of all vehicles on the road; the output format of the real-time object detection model is [(m a ,n a ),(m b ,n b ),σ], where (m a ,n a (m) represents the coordinates of the top-left vertex 'a' of the vehicle bounding box. b ,n b ) represents the coordinate of the bottom right vertex b of the vehicle bounding box, and σ represents the vehicle ID; If a vehicle remains stationary for an extended period, it can be determined that the vehicle is in an abnormal state, and the fifth topology layer [Tr] is generated from the pixels included in the vehicle's bounding box. mn The abnormal conditions include: parking, breakdown, or accident; ; S24: Output lane overall status; Identify the topology matrix of each layer [W]mn ]、[H mn ]、[TH mn ] and [Tr mn In the middle, starting from L' m’n’ To the destination L MN For each element in the same lane, calculate the lane comprehensive state S corresponding to the k-th lane. k ; ; In the formula: S k =0 indicates that the k-th lane is open, S k >1 indicates that a traffic incident occurred in the k-th lane; S25: Event Area Traffic Smoothness Output: Based on the comprehensive status of each lane S k The unobstructedness F of the event region contained in the exit direction of the j-th phase is calculated. j ,; ; In the formula, K represents the number of lanes in the event area; This indicates that for min(S) k 1) Perform the invert operation; F j =0 indicates that the event area is closed and passage is stopped; 0 <F j <1 indicates that some lanes in the incident area are closed, causing slow traffic; F j =1 indicates that the event area is unobstructed; Step S3 specifically includes the following operations: S31: Calculate priority weights and evaluate the status; For the j-th phase of this intersection j Calculate its priority weight W j ; ; In the formula: L j P represents the congestion level of the lane controlled by the j-th phase. j : Upstream traffic pressure of the lane controlled by the j-th phase; F j : The unobstructedness of the event area contained in the j-th phase exit direction; α, β, γ: Adjustable weighting coefficients used to balance the importance of queue clearing, new demand, and downstream status. α and β are preset values, and γ is derived from the values ​​of α and β. ; In the formula: L o : Number of exit lanes for this phase; L i The number of lanes at the entrance of this phase; ; In the formula: A: The total number of approach lanes corresponding to the j-th phase at this intersection; q a Q: Queue length of lane a in the approach lane corresponding to phase j at this intersection; a The maximum queue length of the approach lane a corresponding to the j-th phase at this intersection; ; In the formula: Ts: The operating cycle time of the upstream intersection corresponding to the j-th phase of this intersection; B: The total number of merging phases of the upstream intersection corresponding to the j-th phase of this intersection; t b t represents the release time of the upstream merging phase b corresponding to the j-th phase at this intersection. If the upstream intersection is locked as the merging phase, then t b =T s P j =1; S32: Calculate the green light duration: Once the release phase is determined... j Then allocate a green light time G to it. j ; ; In the formula: J: Number of permitted traffic phases at this intersection; T S ': The phase with the highest priority weight upstream merging at the intersection; G min For the minimum green light time, G max This is the maximum green light time; S33 performs dynamic phase decision-making, as follows: F is prohibited from being released j Phases with a value of 0 are prohibited from entering the deployment area of ​​the intelligent vehicle. Select the current priority weight W j The highest phase is used for clearance; Any activated phase must last for at least one minimum green light time; To prevent long-cycle operation, the continuous release time of a single phase cannot exceed one maximum green light time; S34: Upstream and downstream signal coordination, specifically including the following: When there is severe congestion downstream in a certain direction at this intersection F j When =0, immediately send a coordination request to the adjacent upstream intersection to reduce the green light time for traffic heading in this direction; Based on the request from the downstream intersection, the downstream traffic flow F in the corresponding direction is adjusted in the local priority weight calculation.j Set to 0; S35: Continuous monitoring and adaptation; After one cycle ends, immediately return to S31, re-evaluate the priority weights of all phases, and begin the next decision cycle; Step S4 specifically includes the following steps: S41: Path planning; In the point cloud topology map [f mn In the diagram, warning points and work points P (m) are set for the trolley. p n p ), combined with the topology matrix of each layer ([W mn ]、[H mn ]、[TH mn ]、[Tr mn ]), calculate the travel range from the starting point P0(m0, n0) to the ending point P [R mn ]; in [R mn Within the specified range, based on the spatiotemporal constraint A* path planning algorithm, starting from the current point P... k Iterative search for the optimal point P k+1 Thus, the optimal driving path T is obtained. min ; ; ; ; In the formula: A * () represents the A* path planning algorithm based on spatiotemporal constraints; R mn =1 indicates that the point in the matrix is ​​passable, R mn =0 indicates that the point in the matrix is ​​not passable; L * (P (k+1) * (m k+1 ,n k+1 )) indicates that it is related to P k The set of adjacent points; L * (R k+1 (m k+1 ,n k+1 ) is L * (P k+1 * (m k+1 ,n k+1 )) corresponds to [R mn The set of possible coordinate values ​​is used to determine whether a point is passable. L(P k (m k ,n k)) represents the set of optimal points from the starting point P0 to the ending point P, corresponding to the optimal driving path T. min ; S42 formation marching and deployment; To ensure intersection safety, during the deployment of the intelligent vehicle, phases merging into the deployment area of ​​the intelligent vehicle are prohibited from entering; after the event area is cleared of vehicles, the intelligent vehicle proceeds to the warning point and work point according to the set path; S43: The intelligent vehicle performs control, rescue, and safety warning functions according to its assigned role.

[0011] This invention provides a split-type mobile signal device. After the mobile unit is placed on the target road segment, it uses a drone to scan the traffic data and transmits it to an edge computing module. The edge computing module generates a vehicle deployment task to control the intelligent vehicle to the designated location to complete the designated work. At the same time, the edge computing module generates timing signals to control the traffic light module to manage the traffic on the target road segment. During the traffic management process, the drone monitors the traffic status in real time. The entire process effectively reduces the degree of human intervention and improves the management efficiency for emergencies. Attached Figure Description

[0012] Fig. 1 This is a schematic diagram of the overall structure of a split-type mobile signaling device. Fig. 2 This is a structural diagram of the car's storage compartment; Fig. 3 This is a diagram illustrating how the smart car enters the car storage compartment. Detailed Implementation

[0013] like Figs. 1-3 As shown, this application includes a split-type mobile signaling device, which comprises a mobile body and a signal light module, a drone module, an edge computing module, and a smart car module. The split-type mobile signaling device in this application integrates high-precision aerial sensing, localized edge intelligent decision-making, and dynamic ground signal / signage deployment, enabling second-level response, autonomous decision-making, and efficient traffic management for sudden road incidents.

[0014] The mobile body has a multi-wheeled mobile chassis 8 for carrying all components. The chassis structure is preferably a six-wheeled or eight-wheeled all-terrain structure to ensure stability and mobility on urban roads, temporary construction areas, or complex emergency road sections.

[0015] The traffic light module 2 is installed above the mobile body to ensure that the traffic lights have a good field of vision height. The traffic light module 2 includes at least four sets of standard-sized red, yellow and green traffic lights for direct control of traffic flow at the deployment site.

[0016] The drone module 4 includes drone 4 and drone docking bay 3. When not in use, drone 4 is parked in drone docking bay 3. Drone 4 is equipped with monitoring equipment, including high-resolution visible light cameras, LiDAR, and other sensors, primarily for high-altitude 3D data acquisition. Drone docking bay 3 is located on top of the mobile unit and features an automatic opening and closing door, automatic and precise positioning of the landing platform, and an inductive automatic charging system, enabling fully automatic take-off, landing, and recharging of the drone. The specific drone and drone docking bay are based on existing commercial products.

[0017] The mobile unit is constructed from high-strength, wear-resistant materials and features an internal partitioned design to house the edge computing module, magnetic charging base, and intelligent identification vehicle module, while providing necessary shockproof, heat dissipation, and waterproof protection. The split-type mobile signaling device also integrates a power system; the power system is preferably a high-capacity lithium battery pack to provide continuous power to all modules, and also features an external AC power interface and an emergency power generation interface.

[0018] The edge computing module (not marked in the diagram) is integrated within the mobile unit, incorporating a high-performance industrial-grade computing unit. It receives real-time data streams transmitted from the drone. The edge computing module includes pre-set real-time target detection models for vehicle identification, static information recognition for traffic data, and spatiotemporal hybrid road event detection algorithms. Based on the recognition results and a pre-set strategy library, it generates real-time signal control and intelligent identification vehicle deployment schemes. The edge computing module also integrates 5G / 4G and V2X communication modules, enabling data transmission back to the cloud, remote control, and interconnection with field devices. Furthermore, it features a high-precision GNSS positioning system, facilitating deployment and management by central administrators.

[0019] The intelligent vehicle module 5 includes: a drawer-type vehicle storage compartment 1 and an intelligent vehicle 5. The vehicle storage compartment 1 is located at the lower part of the mobile body above the mobile chassis 8. At least two intelligent vehicles 5 are placed in the vehicle storage compartment 1, and each intelligent vehicle 5 is equipped with an operating device. The operating device includes: a fire extinguisher 13, an operating arm 14, a cone (not marked in the figure), or a warning sign (not marked in the figure).

[0020] The traffic light module 2, the drone module 4, and the smart car module 5 are all connected to the edge computing module.

[0021] The car storage compartment 1 includes at least two layers of vehicle storage space, arranged from top to bottom within the interior of the car storage compartment 1; the dimensions of each layer of vehicle storage space are adapted to the dimensions of the smart car 5; such as Fig. 1 The car storage compartment 1 is divided into different vehicle storage spaces from top to bottom by partitions 7.

[0022] The trolley storage compartment 1 also includes a sliding door 9 and a lifting mechanism; the lifting mechanism includes a lead screw 6, a guide nut (not marked in the figure) and a drive motor (not marked in the figure), with two lead screws 6 respectively set vertically on both sides of the entrance door of the trolley storage compartment 1; the bottom end of each lead screw 6 is rotatably connected to the base 101 of the trolley storage compartment 1, and the top end is connected to the output end of the drive motor.

[0023] The guide nut is threaded and fitted around the outer periphery of the lead screw 6; a hinge (not marked in the figure) is set at the top of the lifting slide door 9 at the position corresponding to the two guide nuts; one end of the lifting slide door 9 is rotatably connected to the guide nut through the hinge; the rotation axis of the hinge is set along the horizontal direction.

[0024] The base 101 of the car storage compartment 1 is located at the edge of the entrance and a guide slope 102 is provided. The lower side of the guide slope 102 is located on the outside. Rollers 10 are installed at the bottom of the lifting sliding door 9 to reduce the probability of the sliding door 9 getting stuck during the lifting process.

[0025] When the sliding door 9 is closed, the guide nut is at its highest position, and the sliding door 9 is vertical, sealing the entrance to the trolley storage compartment 1. When the trolley storage compartment 1 needs to be opened, both drive motors start simultaneously, synchronously driving the two lead screws 6 to rotate, causing the guide nut to descend, which in turn causes the top of the sliding door 9 to descend. Because rollers 10 are installed at the bottom of the lifting sliding door 9, and a guide ramp 102 is set at the edge of the entrance on the base 101, the bottom of the sliding door 9 slides outward along the guide ramp 102. After the bottom of the sliding door 9 touches the ground, it forms a guide ramp for the intelligent trolley to leave the storage compartment. The edge computing module controls the descent height of the guide nut, ensuring that the top of the tilted sliding door 9 is level with the partition 7 or the base 101. This allows the intelligent trolley 5 in the vehicle storage space to leave the storage space and land along the tilted sliding door 9. Based on the cooperation of the sliding door 9 and the lifting mechanism, automatic and rapid exit and return of the trolley to the storage compartment can be achieved.

[0026] The storage compartment 1 for the vehicle is also equipped with a magnetic charging base 11. The magnetic charging interface is continuously powered to maintain the magnetic force, which is used to fix the vehicle and enable standby charging. Two guide grooves 12 are provided on the sliding door 9, the distance between which is adapted to the spacing of the rollers at the bottom of the intelligent vehicle 5. The position of the guide grooves 12 on the sliding door is adapted to the charging position of the intelligent vehicle in the storage compartment 1. The intelligent vehicle 5 has a built-in positioning model and computing chip, and an image acquisition device 15 is installed on the vehicle to collect environmental images. When the intelligent vehicle 5 enters the vehicle storage space along the guide grooves 12, it is positioned in front of the magnetic charging base 11, facilitating charging while the intelligent vehicle is parked.

[0027] The traffic control method based on the above-mentioned split-type mobile signal device includes the following steps.

[0028] S1: Event triggering and device movement; Upon receiving a traffic incident alert, the split-type mobile signaling device is deployed to the area where the traffic incident occurs; The deployment points for split-type mobile signal devices in the area where a traffic incident occurs include: intersections or road sections.

[0029] S2: Holographic scanning and real-time sensing; The split-type mobile signal device is activated, and the edge computing module controls the drone to take off. The drone scans the area where the traffic incident occurred based on its own deployed monitoring equipment, generating a point cloud topology map of the incident area. The edge computing module uses a spatiotemporal hybrid road event detection algorithm to identify the road segment and vehicle where the traffic incident occurred. Step S2 includes the following steps in detail.

[0030] S21: Generate point cloud topology; Representing the event region video stream using a spatiotemporal state matrix generates a point cloud topology map of the event region [f] mn ]; ; In the formula, f mn It is the HSV color space expression for the matrix point.

[0031] S22: Static information recognition; Static information includes: lane lines θ1, water accumulation θ2, flames θ3, and debris θ4; A static information recognition model is constructed based on a neural network; in this embodiment, a static information recognition model is constructed based on the U-Net model.

[0032] After identifying lanes in the traffic incident area based on the static information recognition model, a static information topology layer is generated. The static information topology layer includes: the first topology layer of lane static information [L] mn ], Second topological layer in water accumulation state [W mn ], third topology layer in fire state [H mn ], fourth topological layer of the spill [TH mn ]; ; In the formula, F(f) mn ;δ) is the static information recognition model, where δ is the model parameter; This represents identifying the object category θ at the matrix position (m,n). i The predicted probability relative to the background class θ0; The ArgMax operation is used to retrieve a probability vector. The category index with the highest probability is used as the predicted label for that pixel; If the predicted probability of lane line θ1 is higher than the predicted probability of θ0, then L mn =1, otherwise L mn =0; If the predicted probability of water accumulation θ2 is higher than the predicted probability of water accumulation θ0, then W mn =1, otherwise W mn =0; If the predicted probability of flame θ3 is higher than the predicted probability of flame θ0, then H mn =1, otherwise H mn =0; If the predicted probability of the spill θ4 is higher than the predicted probability of θ0, then TH mn =1, otherwise TH mn =0.

[0033] S23: Dynamic information recognition; A real-time object detection model is constructed based on a convolutional neural network model. In this embodiment, a real-time object detection model is constructed based on YOLOv8.

[0034] The input to the real-time object detection model is the point cloud topology [f mn The output is the bounding boxes of all vehicles on the road; the output format of the real-time object detection model is [(m a ,n a ),(m b ,n b ),σ], where (m a ,n a (m) represents the coordinates of the top-left vertex 'a' of the vehicle bounding box. b ,n b ) represents the coordinate of the bottom right vertex b of the vehicle bounding box, and σ represents the vehicle ID; If a vehicle remains stationary for an extended period, it can be determined that the vehicle is in an abnormal state, and the fifth topology layer [Tr] is generated from the pixels included in the vehicle's bounding box. mn Abnormal conditions include: parking, breakdown, or accident; .

[0035] S24: Output lane overall status; Identify the topology matrix of each layer [W] mn ]、[H mn ]、[TH mn ] and [Tr mn In the middle, starting from L' m’n’ To the destination L MN For each element in the same lane, calculate the lane comprehensive state S corresponding to the k-th lane. k ; ; In the formula: S k =0 indicates that the k-th lane is open, S k >1 indicates that a traffic incident occurred in the k-th lane.

[0036] S25: Event Area Traffic Smoothness Output: Based on the comprehensive status of each lane S k The unobstructedness F of the event region contained in the exit direction of the j-th phase is calculated. j ,; ; In the formula, K represents the number of lanes in the event area; This indicates that for min(S) k 1) Perform the invert operation; F j =0 indicates that the event area is closed and passage is stopped; 0 <F j <1 indicates that some lanes in the incident area are closed, causing slow traffic; F j =1 indicates that the event area is unobstructed.

[0037] S3: Dynamically generate intersection timing schemes: The edge computing platform dynamically generates signal timing schemes by identifying event types and impact ranges in real time; event types include: breakdowns, road flooding, or exit overflows.

[0038] In step S3, the method for generating the signal timing scheme specifically includes the following operations.

[0039] S31: Calculate priority weights and evaluate the status; For the j-th phase of this intersection j Calculate its priority weight W j ; ; In the formula: L j P represents the congestion level of the lane controlled by the j-th phase. j : Upstream traffic pressure of the lane controlled by the j-th phase; F j : The unobstructedness of the event area contained in the j-th phase exit direction; α, β, γ: are adjustable weighting coefficients used to balance the importance of queue clearing, new demand, and downstream status. α and β are set according to historical data. In this embodiment, α and β are set to 1, and γ is derived from the values ​​of α and β. ; In the formula: L o : Number of exit lanes for this phase; L i The number of lanes at the entrance of this phase; ; In the formula: A: The total number of approach lanes corresponding to the j-th phase at this intersection; q a Q: Queue length of lane a in the approach lane corresponding to phase j at this intersection; a The maximum queue length of the approach lane a corresponding to the j-th phase at this intersection; ; In the formula: Ts: The cycle time of the upstream intersection corresponding to the j-th phase of this intersection; B: The total number of merging phases from the upstream intersection corresponding to the j-th phase of this intersection; t b t represents the release time of the upstream merging phase b corresponding to the j-th phase at this intersection. If the upstream intersection is locked as the merging phase, then t b =T s P j =1; Normally, in a phase, the merging phase from the upstream intersection includes the permitted time and the prohibited time. However, in some special cases, such as when special vehicles are passing through, the direction of travel of special vehicles will be set to be in a permitted state for the entire phase time, that is, locked as a merging phase. In this case, t b =T s .

[0040] S32: Calculate the green light duration: Once the release phase is determined... j Then allocate a green light time G to it. j ; ; In the formula: J: Number of permitted traffic phases at this intersection; T S ': The phase with the highest priority weight upstream merging at the intersection; G min For the minimum green light time, G max This is the maximum green light time.

[0041] S33: Perform dynamic phase decision-making, as follows: F is prohibited from being released j Phases with a value of 0 are prohibited from entering the deployment area of ​​the intelligent vehicle. Select the current priority weight W j The highest phase is used for clearance; Any activated phase must last for at least a minimum green light time G. min ; To prevent long-cycle operation, the continuous release time for a single phase cannot exceed one maximum green light time G. max .

[0042] In this embodiment, Gmin The value is 15 seconds, G max The value is 90 seconds.

[0043] S34: Upstream and downstream signal coordination, specifically including the following: When there is severe congestion downstream in a certain direction at this intersection F j When =0, immediately send a coordination request to the adjacent upstream intersection to reduce the green light time for traffic heading in this direction; Based on the request from the downstream intersection, the downstream traffic flow F in the corresponding direction is adjusted in the local priority weight calculation. j Setting it to 0 significantly reduces the priority of this phase, thus achieving current limiting.

[0044] S35: Continuous monitoring and adaptation; After one cycle ends, the system immediately returns to S31 to reassess the priority weights of all phases and begin the next decision cycle. The entire process is dynamic and adaptive.

[0045] S4: Coordinated deployment of road management and intelligent vehicles; The intelligent vehicle quickly leaves the warehouse and autonomously travels to the designated location for deployment based on a spatiotemporal constraint path planning algorithm.

[0046] The spatiotemporal constraint path planning algorithm in step S4 specifically includes the following steps: S41: Path planning; In the point cloud topology map [f mn In the diagram, warning points and work points P(m) are set for the trolley. p n p ), combined with the topology matrix of each layer ([W mn ]、[H mn ]、[TH mn ]、[Tr mn ]), calculate the travel range from the starting point P0(m0, n0) to the ending point P [R mn ]; in [R mn Within the specified range, based on the spatiotemporal constraint A* path planning algorithm, starting from the current point P... k Iterative search for the optimal point P k+1 Thus, the optimal driving path T is obtained. min ; ; ; ; In the formula: A * () represents the A* path planning algorithm based on spatiotemporal constraints; R mn =1 indicates that the point in the matrix is ​​passable, Rmn =0 indicates that the point in the matrix is ​​not passable; L * (P (k+1) * (m k+1 ,n k+1 )) indicates that it is related to P k The set of adjacent points; L * (R k+1 (m k+1 ,n k+1 ) is L * (P k+1 * (m k+1 ,n k+1 )) corresponds to [R mn The set of possible coordinate values ​​is used to determine whether a point is passable. L(P k (m k ,n k )) represents the set of optimal points from the starting point P0 to the ending point P, corresponding to the optimal driving path T. min ; S42 formation marching and deployment; To ensure intersection safety, during the deployment of the intelligent vehicle, phases merging into the deployment area of ​​the intelligent vehicle are prohibited from entering; after the event area is cleared of vehicles, the intelligent vehicle proceeds to the warning point and work point according to the set path; S43: The intelligent vehicle performs control, rescue, and safety warning functions according to its assigned role.

[0047] For example: when a smart car carrying cones or warning signs arrives at a warning point, it turns on the warning flashing lights; when a smart car carrying a fire extinguisher arrives at a work site, it automatically detects the fire source, turns on the fire extinguisher, and carries out firefighting operations; when a bomb disposal smart car arrives at a work site, it grabs flammable and explosive materials and tows them away from the scene.

[0048] S5: Real-time iterative optimization: During event handling, the drone continuously monitors, and the edge computing platform adjusts the signal timing and the location of the smart car in real time.

[0049] S6: After the incident is resolved, the drone and intelligent vehicle automatically return to their storage location, and the separate mobile signaling device is removed.

[0050] The split-type mobile signaling device of this application is activated after being deployed to the event area. The drone module takes off to scan, fusing point cloud data with video streams and transmitting the data in real time to the edge computing module. The edge computing module uses an improved AI model to identify event types in real time, generating dynamic signal timing schemes and vehicle deployment tasks. The signal light module immediately executes the generated dynamic timing scheme; the intelligent signage vehicle module quickly exits the compartment via a sliding rail mechanism, plans its path based on a spatiotemporal constraint A* algorithm, and autonomously proceeds to the designated location to complete designated tasks, such as deploying cones and warning signs, performing firefighting and bomb disposal operations, or completing warning, control, and rescue tasks.

[0051] The entire process utilizes V2X communication to achieve low-latency collaborative control of the drone, traffic lights, and the vehicle. During incident handling, the drone continuously monitors the system, and the system adjusts the signal timing and vehicle position in real time. After the incident is resolved, the vehicle automatically returns to its storage location, and the device is withdrawn.

[0052] Example 1: Emergency response to sudden incidents on road sections.

[0053] S1 device deployment: When the city traffic management department detects an emergency such as vehicle breakdown, accident, or waterlogging closure on a certain road section through other systems (such as geomagnetism, monitoring), the maintenance personnel immediately deploy a split-type mobile signal device to the upstream intersection without signal control of the emergency.

[0054] S2 signal control activation: After the device comes to a complete stop, the signal light module begins to direct traffic flow.

[0055] S3 Aerial Scanning and Event Confirmation: The drone module automatically ascends to a preset monitoring altitude, quickly scans the area's traffic conditions, and determines the scope and extent of the impact.

[0056] S4 Decision Generation and Signal Control: The edge computing module receives video streams from drones in real time and optimizes traffic light control strategies based on event and traffic condition recognition results to prevent too many vehicles from entering downstream event sections and causing traffic congestion.

[0057] S5 Smart Car Deployment: The system activates the smart identification car module. The electromagnetic lock on the car's storage compartment unlocks, the sliding door extends, and the smart identification car (equipped with cones and warning signs) exits the compartment within 30 seconds.

[0058] S6 Path Planning and Marking Deployment: The vehicle uses built-in high-precision GNSS and spatiotemporal constraint algorithms, combined with real-time imagery provided by drones, to plan the best path to avoid normally driving vehicles. It then autonomously drives to a predetermined safe distance behind a broken-down vehicle or emergency location to deploy cones and LED warning signs, forming a physical warning area and completing the tasks of closing off the obstructed lane and issuing warnings.

[0059] S7 Collaborative Control and Dynamic Optimization: The drone continuously monitors traffic flow and emergency rescue progress in the air, and the edge computing platform analyzes the degree of traffic congestion in real time. If there is a significant overflow of traffic flow, the operation plan of the traffic light module can be dynamically adjusted, and the vehicle's positioning and task synchronization can be maintained through the V2X protocol.

[0060] Example 2: Intersection overflow control and traffic flow management.

[0061] S1 Manual Deployment: The city traffic management center anticipates or senses a potential or already occurring exit overflow event at a signalless or out-of-signal intersection. Maintenance personnel deploy a split-type mobile signal device to the center of that intersection.

[0062] S2 Signal Takeover: After the signal light module of the device is in place, it can quickly take over or coordinate the command of the existing signal control scheme of the intersection through wireless communication or reserved interface.

[0063] Traffic status perception and recognition at S3 intersection: The drone module automatically ascends to a preset altitude to monitor the traffic operation status in all directions of the intersection.

[0064] S4 Intersection Overflow Status Recognition: The edge computing platform analyzes the video data transmitted in real time by drones and runs a traffic status recognition model. The model identifies an abnormally slow downstream traffic flow, causing vehicles to wait within the intersection and resulting in vehicle accumulation exceeding the exit lane of the intersection, i.e., an intersection overflow event.

[0065] S5 Decision Generation: The edge computing module generates signal control strategies in real time based on the current intersection operation status: for overflowing exit traffic entering the intersection, shorten the green light time or directly implement "red light" control to reduce the number of vehicles entering the intersection, prioritize clearing overflowing vehicles, and avoid the intersection from being completely locked.

[0066] S6 Signal Control Execution: The traffic light module immediately executes the dynamic timing scheme.

[0067] S7 Continuous Monitoring and Dynamic Optimization: The drone continuously monitors the intersection status, and the edge computing module calculates the intersection's operating status in real time.

[0068] S8 Intersection Recovery: When the edge computing module determines that the overflowing vehicles have been basically cleared and the vehicle occupancy rate in the intersection is lower than the threshold, the system will dynamically adjust the timing and gradually restore the green light time of the affected inbound traffic flow, so that the traffic flow at the intersection can return to normal operation and achieve efficient traffic management.

[0069] This patented split-type mobile signal device, through its highly integrated design, achieves a closed loop of perception, decision-making, and execution, greatly improving the intelligence level and response speed of traffic management in emergencies.

[0070] By employing the proposed technical solution, a closed-loop architecture of "airborne perception - edge decision-making - ground execution" enables rapid response to traffic emergencies, significantly improving the intelligence level and response speed of traffic management during emergencies. Simultaneously, the use of high-performance edge computing devices and spatiotemporal hybrid algorithms supports localized, autonomous dynamic decision-making, dynamically adjusting signal timing based on real-time traffic flow and event type, thus solving the problems of fixed timing and lack of intelligence in traditional mobile traffic lights. This application also integrates an intelligent signage vehicle module, enabling rapid deployment of physical warning signs such as cones and warning signs, which, in conjunction with traffic light control, achieve three-dimensional, multi-level traffic guidance and warning.

Claims

1. A split-type mobile signaling device, comprising a mobile body and a signal light module, characterized in that, It also includes: a drone module, an edge computing module, and a smart car module; The mobile body has a multi-wheeled mobile chassis used to carry all components; The traffic light module is installed on top of the mobile body, and the traffic light module includes at least four sets of standard-sized red, yellow and green traffic lights; The drone module includes a drone and a drone docking bay; the drone docking bay is located on the top of the mobile body; the drone module is equipped with monitoring equipment; The edge computing module is integrated within the mobile unit, incorporating a high-performance industrial-grade computing unit. This module includes pre-set real-time target detection models for vehicle identification, static information recognition for traffic data, and spatiotemporal hybrid road event detection algorithms. Based on the identification results and a pre-set strategy library, it generates real-time signal control and intelligent vehicle deployment schemes. The edge computing module also integrates a mobile communication module and a V2X communication module, supporting data transmission back to the cloud, remote control, and interconnection with field devices. Furthermore, it features a high-precision GNSS positioning system. The intelligent vehicle module includes: a drawer-type vehicle storage compartment and an intelligent vehicle. The vehicle storage compartment is located at the lower part of the mobile body above the mobile chassis. At least two intelligent vehicles are placed in the vehicle storage compartment, and each intelligent vehicle is equipped with a portable device. The portable device includes: an intelligent traffic cone, a warning sign, or a fire extinguisher. The traffic light module, the drone module, and the smart car module are all communicatively connected to the edge computing module.

2. The split-type mobile signal device according to claim 1, characterized in that: The vehicle storage compartment includes at least two layers of vehicle storage space, which are arranged from top to bottom within the interior of the vehicle storage compartment; the size of each layer of vehicle storage space is adapted to the size of the smart car.

3. The split-type mobile signaling device according to claim 2, characterized in that: The trolley storage compartment also includes: a sliding door and a lifting mechanism; The lifting mechanism includes a lead screw, a guide nut, and a drive motor. The two lead screws are respectively set vertically on both sides of the entrance gate of the trolley storage compartment. The bottom end of each lead screw is rotatably connected to the base of the trolley storage compartment, and the top end is connected to the output end of the drive motor. The guide nut is threadedly fitted onto the outer circumference of the lead screw. A hinge is provided at the top of the sliding door, corresponding to the two guide nuts; one end of the sliding door is rotatably connected to the guide nuts via the hinges; the rotation axis of the hinges is set along the horizontal direction. The entrance to the trolley storage compartment is provided with a guide ramp, and the lower edge of the guide ramp is located on the outer side; Rollers are installed at the bottom of the sliding door.

4. The split-type mobile signal device according to claim 3, characterized in that: The storage compartment of the vehicle is also equipped with a magnetic charging base, and the sliding door is provided with two guide rail grooves. The distance between the two guide rail grooves is adapted to the distance between the rollers at the bottom of the intelligent vehicle. The guide rail groove is positioned on the sliding door to match the position of the magnetic charging base of the smart car in the car storage compartment.

5. A traffic control method based on a split-type mobile signal device, characterized in that, It includes the following steps: S1: Event triggering and device movement; Upon receiving a traffic incident alert, the split-type mobile signaling device is deployed to the area where the traffic incident occurs; The deployment points of the split-type mobile signal device in the area where the traffic incident occurred include: intersections or road sections; S2: Holographic scanning and real-time sensing; The split-type mobile signal device is activated, and the edge computing module controls the drone to take off. The drone scans the area where the traffic incident occurred based on its own deployed monitoring equipment, generating a point cloud topology map of the incident area. The edge computing module uses a spatiotemporal hybrid road event detection algorithm to identify the road segment and vehicle where the traffic incident occurred. S3: Dynamically generate intersection timing schemes: The edge computing platform dynamically generates signal timing schemes by identifying event types and their impact range in real time; the event types include: breakdowns, road flooding, or exit overflows; S4: Coordinated deployment of road management and intelligent vehicles; The intelligent vehicle quickly leaves the warehouse and autonomously travels to the designated location for deployment based on a spatiotemporal constraint path planning algorithm. S5: Real-time iterative optimization: During event handling, the drone continuously monitors and the edge computing platform adjusts the signal timing and the location of the smart car in real time; S6: After the incident is resolved, the drone and intelligent vehicle automatically return to their storage location, and the separate mobile signaling device is removed.

6. The traffic control method based on a split-type mobile signal device according to claim 5, characterized in that: Step S2 includes the following steps in detail: S21: Generate point cloud topology; Representing the event region video stream using a spatiotemporal state matrix generates a point cloud topology map of the event region [f] mn ]; ； In the formula, f mn It is the HSV color space expression for the matrix points; S22: Static information recognition; The static information includes: lane lines θ1, water accumulation θ2, flames θ3, and debris θ4; A static information recognition model is constructed based on a neural network. After identifying lanes in the traffic incident area based on the static information recognition model, a static information topology layer is generated, which includes: a first lane static information topology layer [L]. mn ], Second topological layer in water accumulation state [W mn ], third topology layer in fire state [H mn ], fourth topological layer of the spill [TH mn ]; ； In the formula, F(f) mn ;δ) is the static information recognition model, where δ is the model parameter; This represents identifying the object category θ at the matrix position (m,n). i The predicted probability relative to the background class θ0; The ArgMax operation is used to retrieve a probability vector. The category index with the highest probability is used as the predicted label for that pixel; If the predicted probability of lane line θ1 is higher than the predicted probability of θ0, then L mn =1, otherwise L mn =0; If the predicted probability of water accumulation θ2 is higher than the predicted probability of water accumulation θ0, then W mn =1, otherwise W mn =0; If the predicted probability of flame θ3 is higher than the predicted probability of flame θ0, then H mn =1, otherwise H mn =0; If the predicted probability of the spill θ4 is higher than the predicted probability of θ0, then TH mn =1, otherwise TH mn =0; S23: Dynamic information recognition; A real-time object detection model is constructed based on a convolutional neural network model. The input of the real-time object detection model is the point cloud topology [f] mn The output is the bounding boxes of all vehicles on the road; the output format of the real-time object detection model is [(m a ,n a ),(m b ,n b ),σ], where (m a ,n a (m) represents the coordinates of the top-left vertex 'a' of the vehicle bounding box. b ,n b ) represents the coordinate of the bottom right vertex b of the vehicle bounding box, and σ represents the vehicle ID; If a vehicle remains stationary for an extended period, it can be determined that the vehicle is in an abnormal state, and the fifth topology layer [Tr] is generated from the pixels included in the vehicle's bounding box. mn The abnormal conditions include: parking, breakdown, or accident; ； S24: Output lane overall status; Identify the topology matrix of each layer [W] mn ]、[H mn ]、[TH mn ] and [Tr mn In the middle, starting from L' m’n’ To the destination L MN For each element in the same lane, calculate the lane comprehensive state S corresponding to the k-th lane. k ; ； In the formula: S k =0 indicates that the k-th lane is open, S k >1 indicates that a traffic incident occurred in the k-th lane; S25: Event Area Traffic Smoothness Output: Based on the comprehensive status of each lane S k The unobstructedness F of the event region contained in the exit direction of the j-th phase is calculated. j ,; ； In the formula, K represents the number of lanes in the event area; This indicates that for min(S) k 1) Perform the invert operation; F j =0 indicates that the event area is closed and passage is stopped; 0 <F j <1 indicates that some lanes in the incident area are closed, causing slow traffic; F j =1 indicates that the event area is unobstructed.

7. The traffic control method based on a split-type mobile signal device according to claim 5, characterized in that: Step S3 specifically includes the following operations: S31: Calculate priority weights and evaluate the status; For the j-th phase of this intersection j Calculate its priority weight W j ; ； In the formula: L j P represents the congestion level of the lane controlled by the j-th phase. j : Upstream traffic pressure of the lane controlled by the j-th phase; F j : The unobstructedness of the event area contained in the j-th phase exit direction; α, β, γ: Adjustable weighting coefficients used to balance the importance of queue clearing, new demand, and downstream status. α and β are preset values, and γ is derived from the values ​​of α and β. ； In the formula: L o : Number of exit lanes for this phase; L i The number of lanes at the entrance of this phase; ； In the formula: A: The total number of approach lanes corresponding to the j-th phase at this intersection; q a Q: Queue length of lane a in the approach lane corresponding to phase j at this intersection; a The maximum queue length of the approach lane a corresponding to the j-th phase at this intersection; ； In the formula: Ts: The operating cycle time of the upstream intersection corresponding to the j-th phase of this intersection; B: The total number of upstream merging phases corresponding to the j-th phase of this intersection; t b t represents the release time of the upstream merging phase b corresponding to the j-th phase at this intersection. If the upstream intersection is locked as the merging phase, then t b =T s P j =1; S32: Calculate the green light duration: Once the release phase is determined... j Then allocate a green light time G to it. j ; ； In the formula: J: Number of permitted traffic phases at this intersection; T S ': The phase with the highest priority weight upstream merging at the intersection; G min For the minimum green light time, G max This is the maximum green light time; S33 performs dynamic phase decision-making, as follows: F is prohibited from being released j Phases with a value of 0 are prohibited from entering the deployment area of ​​the intelligent vehicle. Select the current priority weight W j The highest phase is used for clearance; Any activated phase must last for at least one minimum green light time; To prevent long-cycle operation, the continuous release time of a single phase cannot exceed one maximum green light time; S34: Upstream and downstream signal coordination, specifically including the following: When there is severe congestion downstream in a certain direction at this intersection F j When =0, immediately send a coordination request to the adjacent upstream intersection to reduce the green light time for traffic heading in this direction; Based on the request from the downstream intersection, the downstream traffic flow F in the corresponding direction is adjusted in the local priority weight calculation. j Set to 0; S35: Continuous monitoring and adaptation; After one cycle ends, immediately return to S31 to re-evaluate the priority weights of all phases and begin the next decision cycle.

8. The traffic control method based on a split-type mobile signal device according to claim 5, characterized in that: Step S4 specifically includes the following steps: S41: Path planning; In the point cloud topology map [f mn In the diagram, warning points and work points P (m) are set for the trolley. p n p ), combined with the topology matrix of each layer ([W mn ]、[H mn ]、[TH mn ]、[Tr mn ]), calculate the travel range from the starting point P0(m0, n0) to the ending point P [R mn ]; in [R mn Within the specified range, based on the spatiotemporal constraint A* path planning algorithm, starting from the current point P... k Iterative search for the optimal point P k+1 Thus, the optimal driving path T is obtained. min ; ； ； ； In the formula: A * () represents the A* path planning algorithm based on spatiotemporal constraints; R mn =1 indicates that the point in the matrix is ​​passable, R mn =0 indicates that the point in the matrix is ​​not passable; L * (P (k+1) * (m k+1 ,n k+1 )) indicates that it is related to P k The set of adjacent points; L * (R k+1 (m k+1 ,n k+1 ) is L * (P k+1 * (m k+1 ,n k+1 )) corresponds to [R mn The set of possible coordinate values ​​is used to determine whether a point is passable. L(P k (m k ,n k )) represents the set of optimal points from the starting point P0 to the ending point P, corresponding to the optimal driving path T. min ; S42 formation marching and deployment; To ensure intersection safety, during the deployment of the intelligent vehicle, phases merging into the deployment area of ​​the intelligent vehicle are prohibited from entering; after the event area is cleared of vehicles, the intelligent vehicle proceeds to the warning point and work point according to the set path; S43: The intelligent vehicle performs control, rescue, and safety warning functions according to its assigned role.

Images