Tunnel traffic accident emergency escape unmanned aerial vehicle cooperative guidance system and method

Abstract

The invention discloses a tunnel traffic accident emergency escape unmanned aerial vehicle cooperative guidance system and method, and belongs to the technical field of tunnel safety emergency. The system comprises an unmanned aerial vehicle nest unit which is internally provided with an ejection mechanism and accommodates a plurality of foldable unmanned aerial vehicles, a guiding unmanned aerial vehicle which carries an SLAM navigator, a thermal imager, a projection lamp and an emergency material putting device, and a communication relay unmanned aerial vehicle which constructs a Mesh + optical communication redundant link. According to the control method, path planning is dynamically optimized based on a reinforcement learning algorithm, and visual sense-tactile sense-auditory sense multichannel guidance is realized in combination with UWB positioning and the adaptive projection lamp. According to the invention, the sensing precision of the complex environment is improved through multi-sensor fusion positioning, the communication delay is reduced through cooperation of the dual-frequency Mesh network and optical communication, the evacuation efficiency is improved through interaction of dynamic projection and the AR terminal, the problems of real-time positioning, communication coverage and accurate guidance in a tunnel fire scene are effectively solved, and the success rate of emergency rescue is remarkably improved.

Description

Technical Field

[0001] This application relates to the field of tunnel safety emergency technology, and in particular to a collaborative guidance system and method for emergency escape from tunnel traffic accidents using unmanned aerial vehicles (UAVs). Background Technology

[0002] With the acceleration of urbanization, the scale and complexity of highway and railway tunnel construction have significantly increased, making tunnel safety a core challenge in public infrastructure management. Current tunnel emergency escape systems mainly rely on three technical solutions: fixed signage guidance, emergency lighting facilities, and manual rescue command. However, their limitations in extreme accident scenarios (such as large-scale traffic accidents and fires) are becoming increasingly apparent. Traditional tunnels commonly use reflective signs, fluorescent directional arrows, and readily available emergency lights as escape guidance. However, in accidents such as traffic accidents or fires, high-temperature smoke can cause visibility to plummet to less than 1 meter, rendering reflective materials completely ineffective in dense smoke. Furthermore, existing emergency lighting largely relies on independent battery power, but under extreme temperatures or structural damage, the electrical system may fail, limiting the lighting range to fixed locations and making it impossible to adjust guidance paths according to accident dynamics (such as the direction of toxic gas diffusion). Research indicates that in a major traffic fire accident in a tunnel in 2021, approximately 73% of trapped individuals mistakenly entered dangerous areas because they could not identify fixed signs. When an accident occurs, firefighters and medical personnel must enter the scene through the external entrance of the tunnel. Limited by the tunnel's length (some ultra-long tunnels exceed 10 kilometers) and the traffic congestion caused by the accident, the average response time exceeds 30 minutes, missing the "golden rescue period." Furthermore, the tunnel lacks adequate firefighting, oxygen production, and ventilation conditions. Rescuers must carry equipment through high-temperature, oxygen-deficient, or toxic gas environments, posing a serious threat to their own safety. Although robot-assisted rescue has developed in recent years, most equipment is limited to single functions (such as firefighting or detection) and lacks the ability to provide real-time dynamic guidance to trapped personnel. Some cutting-edge research has attempted to introduce unmanned aerial vehicles (UAVs) into tunnel rescue, but their designs are mostly geared towards open environments and are difficult to adapt to the special conditions of tunnels. For example, GPS signals are blocked inside tunnels; the visual / inertial navigation fusion algorithms relied upon by traditional UAVs have an error rate exceeding 40% in environments with dense smoke and dust, and are prone to collisions with obstacles; concrete structures significantly attenuate wireless signals, the communication distance between a single UAV and the control center is usually less than 500 meters, and they cannot penetrate multiple layers of collapsed structures; existing rescue UAVs are mostly focused on environmental monitoring (such as gas detection) or material delivery, lacking interactive guidance mechanisms for trapped personnel. While some solutions have attempted to integrate IoT and drone technology in recent years, numerous bottlenecks remain. These include: Lack of dynamic path updates: the system relies on static beacons or pre-stored maps, making it unable to dynamically update escape routes based on fire spread and structural damage; Insufficient multi-device coordination: the lack of a unified scheduling protocol among drones, sensors, and rescue terminals easily leads to command conflicts; and weak human-machine interaction: trapped individuals have a low response rate to complex commands (such as text prompts on mobile apps) in a state of panic.

[0003] In summary, existing technologies are insufficient to meet the core requirements of real-time environmental perception, dynamic path planning, and multimodal reliable guidance in tunnel traffic accidents. There is an urgent need for an innovative escape system that integrates autonomous navigation, cluster collaboration, and human-computer interaction. Summary of the Invention

[0004] The purpose of this application is to overcome the shortcomings of existing technologies by providing a collaborative drone-based emergency escape system and method for tunnel traffic accidents. This system effectively solves the challenges of real-time positioning, communication coverage, and precise guidance in tunnel fire scenarios through deep integration of multi-source perception fusion, autonomous navigation algorithms, dynamic task allocation, and a highly robust communication architecture. This significantly improves the efficiency and safety of emergency rescue in tunnel fire scenarios and meets the technical requirements of modern tunnel rescue systems.

[0005] To achieve the above objectives, this application provides the following solution: A collaborative drone-guided emergency escape system and method for tunnel traffic accidents, characterized in that it includes: Distributed drone nesting units deployed within the tunnel, each containing multiple foldable emergency drones; An environmental sensing unit is used to detect temperature, gas concentration, and image information inside the tunnel. The control center generates escape routes based on environmental data and dispatches a swarm of drones. At least one guidance drone, equipped with a SLAM navigation module, thermal imager, projector, and emergency supplies delivery device; At least one communications relay drone is used to build a temporary wireless communication network.

[0006] A further improvement is that the guiding drone interacts with the trapped person's terminal via a UWB positioning module, the terminal including a mobile phone, smart bracelet, or AR glasses.

[0007] A further improvement is that the drone nest unit includes a catapult mechanism that can release the drone within 5 seconds and supports automatic homing and charging of the drone.

[0008] A further improvement is that the projection light can project dynamic escape path arrows on the ground and automatically adjust the brightness and color wavelength according to the smoke concentration.

[0009] A further improvement is that the control center uses a reinforcement learning algorithm to optimize the path planning of multiple drones, with the priority order being: escape route opening > survivor location > data transmission.

[0010] A method for controlling emergency escape in tunnels using drones in collaboration, characterized by the following steps: S1: Receives an accident trigger signal and activates the drone nesting unit in the corresponding area; S2: Communication relay drones take off to establish a mesh network and transmit real-time data inside the tunnel; S3: Guide the drone to scan and generate a 3D map of the tunnel, marking obstacles and heat sources; S4: The control center integrates multi-source data, generates a global escape route, and issues instructions; S5: Drones guide personnel to evacuate in an orderly manner through projection, voice, and terminal interaction, and simultaneously drop emergency supplies; S6: Continuously monitor and update escape routes.

[0011] A further improvement is that the route planning dynamically avoids the access lanes of rescue vehicles and updates the guidance direction using projection lights.

[0012] The beneficial effects of the technical solution in this application are as follows: Multi-source sensing and communication fusion enhances adaptability to complex environments. This application utilizes a fusion sensing system combining lidar, visual SLAM, and infrared thermal imaging to achieve high-precision positioning of fire sources, obstacles, personnel, and vehicles within tunnels. Employing a dual-frequency mesh network (2.4GHz + 5.8GHz) and redundant UWB / LED optical communication links significantly reduces communication latency in environments without satellite signals, expands the coverage radius, and provides strong technical support for safety management and emergency response in enclosed spaces such as tunnels.

[0013] Autonomous navigation and collaborative decision-making without satellite signals. This application utilizes laser SLAM and optical flow positioning technology to achieve autonomous obstacle avoidance and dynamic path planning in tunnels without GPS signals. It supports serpentine formation collaborative scanning and generates the optimal escape route based on global path planning using reinforcement learning algorithms, taking into account parameters such as fire spread speed and personnel movement speed, significantly improving the success rate of emergency rescue.

[0014] Dynamic guidance and multimodal interaction enhance evacuation efficiency and accuracy. The guiding drone uses adaptive brightness green lasers to project escape arrows, improving smoke penetration by 40% compared to traditional red light, and creating a continuous guidance zone 10 meters in front of personnel, ensuring uninterrupted visual guidance in dense smoke environments. UWB positioning triggers mobile phone vibration commands and an AR virtual escape wall (blue light strip), combined with multilingual voice broadcasts, achieving multi-channel guidance through visual, tactile, and auditory senses, significantly improving evacuation efficiency. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a system block diagram of a drone-assisted emergency escape system and method for tunnel traffic accidents. Figure 2 This is a forward link diagram of a drone-coordinated tunnel traffic accident emergency escape guidance system; Figure 3 A flowchart of a drone-assisted emergency escape system and method for tunnel traffic accidents; Figure 4 This is a schematic block diagram of a three-level communication network for a collaborative guidance system and method for emergency escape from tunnel traffic accidents using unmanned aerial vehicles (UAVs). Detailed Implementation

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] This embodiment relates to a drone-assisted emergency escape system and method for tunnel traffic accidents. It includes: Distributed drone nesting units deployed within the tunnel, each containing multiple foldable emergency drones; An environmental sensing unit is used to detect temperature, gas concentration, and image information inside the tunnel. The control center generates escape routes based on environmental data and dispatches a swarm of drones. At least one guidance drone, equipped with a SLAM navigation module, thermal imager, projector, and emergency supplies delivery device; At least one communications relay drone is used to build a temporary wireless communication network.

[0020] The drone is guided to interact with the trapped person's terminal via a UWB positioning module. The terminal can include a mobile phone, smart bracelet, or AR glasses.

[0021] The drone nest unit includes a catapult mechanism that can release the drone within 5 seconds and supports automatic homing and charging of the drone.

[0022] The projector can project dynamic escape path arrows onto the ground and automatically adjust its brightness and color wavelength according to the smoke concentration.

[0023] The control center uses reinforcement learning algorithms to optimize the path planning of multiple drones, with the priority order being: escape route opening > survivor location > data transmission.

[0024] A method for controlling emergency escape in tunnels using drones in collaboration includes the following steps: S1: Receives an accident trigger signal and activates the drone nesting unit in the corresponding area; S2: Communication relay drones take off to establish a mesh network and transmit real-time data inside the tunnel; S3: Guide the drone to scan and generate a 3D map of the tunnel, marking obstacles and heat sources; S4: The control center integrates multi-source data, generates a global escape route, and issues instructions; S5: Drones guide personnel to evacuate in an orderly manner through projection, voice, and terminal interaction, and simultaneously drop emergency supplies; S6: Continuously monitor and update escape routes.

[0025] The route planning dynamically avoids access lanes for rescue vehicles, and the guidance direction is updated via projection lights. In this embodiment, a 3.2-kilometer-long, two-way, four-lane highway tunnel is used as the application scenario. The tunnel's traffic accident emergency escape guidance system employs a five-layer collaborative architecture to achieve integrated control of sensing, computation, and communication. For example... Figure 2 As shown, the system integrates the high efficiency of centralized processing with the fault tolerance of distributed architecture, and achieves millisecond-level emergency response through sensor-computer collaboration, meeting the reliable control requirements of complex environments in tunnel scenarios.

[0026] At the sensor layer, multi-source heterogeneous sensors (LiDAR, cameras, millimeter-wave radar) achieve holographic environmental perception. Raw data is spatiotemporally aligned and fused through a centralized architecture. Meanwhile, thermal imagers accurately locate the fire source and survivor coordinates, providing low-latency input for subsequent decision-making. The edge computing layer uses a distributed architecture for feature layer fusion, extracting key information such as obstacle outlines and motion trajectories to improve system fault tolerance. The central processing layer uses a reinforcement learning engine to fuse historical cases and environmental data, dynamically allocate computing resources, and generate globally optimal paths, achieving real-time decision-making with an escape route opening priority higher than 80%. The communication control layer uses a Mesh+5G hybrid network and narrowband IoT dual-channel transmission to ensure end-to-end latency of emergency commands <50ms. The execution terminal layer deploys guidance drones equipped with projection lights and AR navigation to form multimodal interaction, with dynamic path correction response speed three times faster than traditional solutions. The system forms a closed-loop control through a forward link: sensor → edge computing → central decision-making → terminal execution and feedback link: terminal status → path correction → algorithm update. Combining the efficiency of centralized processing with the reliability of distributed architecture, the system achieves a 99.9% emergency response success rate in the complex environment of tunnels.

[0027] The system and hardware configuration deployment are as follows: Eight drone nesting units were deployed, embedded every 400 meters along the tunnel ceiling. Each unit housed two foldable guide drones and one communication relay drone. The folded drones measured 30cm x 20cm x 15cm. The ejection mechanism used a combination of a compression spring and an electromagnetic lock, unlocking within 0.5 seconds of triggering and launching the drones to a hovering position 5 meters above the ground within 5 seconds. Each unit integrated a wireless charging module, allowing the drones to fully charge within 30 minutes of returning to their nests, achieving a charging efficiency of 90%.

[0028] Environmental sensing units are deployed. A set of sensor nodes is installed every 50 meters along the tunnel sidewall. Each set includes an infrared thermal imaging camera (640×480 resolution, 30fps), a multi-functional gas sensor (detecting CO, NO2, and PM2.5 concentrations), and a vibration sensor (sensitivity 0.1g, capable of detecting vibration shock waves). Sensor data is transmitted to the control center via a redundant fiber optic network with a latency of less than 50ms.

[0029] Control Center. Deployed at the tunnel management station, the hardware utilizes NVIDIA Jetson AGX Xavier edge computing devices, equipped with a pre-trained reinforcement learning model (based on the PyTorch framework, with training data sourced from 100,000 tunnel fire simulation scenarios), supporting 32 TOPS of real-time computing power. Multi-drone path planning is implemented, prioritizing: escape route opening > survivor location > data transmission. When a truck rear-ends another vehicle and catches fire in the middle of the tunnel (1.5 kilometers from the entrance): The temperature sensor detected a sudden rise in local temperature to 80°C and a CO concentration exceeding 500 ppm. The vibration sensor captured a vehicle collision signal (peak acceleration of 3.5g), triggering an accident.

[0030] S1: The control center determines the accident level to be "severe fire" and immediately activates the three drone nesting units N4, N5 and N6 closest to the accident point to start multi-drone collaborative networking.

[0031] Communication relay drones are prioritized for takeoff to establish a three-tiered communication network: Layer 1: Relay drones are directly connected to drone nesting units, with the highest bandwidth priority; The second layer: Relay drones interconnect via the 5.8GHz frequency band to extend coverage; The third layer: guides drones to connect to the network and controls them via relay nodes.

[0032] S2: The communication relay drone activates a dual-frequency mesh network (2.4GHz + 5.8GHz), covering a radius of 800 meters, transmitting real-time 360° panoramic video of the accident site (H.265 encoding, bandwidth usage <10Mbps), and sending emergency text messages to the trapped personnel's mobile phones: "Fire in the middle of the tunnel, please wait for drone guidance for evacuation." S3: Guide the UAV to activate the lidar (wavelength 905nm, scanning frequency 20Hz) and visual SLAM module, and enter the accident area in a serpentine formation (the distance between adjacent UAVs is 10 meters, and the scanning range overlaps by 30%) to build a 3D point cloud map of the tunnel with a mapping accuracy error of <1%.

[0033] The thermal imager identified three areas where people were gathered (areas with a temperature of 37±2℃, marked as P1-P3), and combined with data from the gas sensor, predicted the fire spread rate to be 2 m / s.

[0034] S4: The control center runs a reinforcement learning algorithm (Q-Learning model). Input parameters include the direction of fire spread, the minimum width of the escape route (1.2m), and the personnel movement speed (0.8m / s). It generates a globally optimal escape route: two escape routes (Route A / B) are opened from P1-P3 towards the exit, prioritizing the removal of obstructing vehicles on Route A. S5: Dynamic Projection Guidance. Multiple guidance drones project arrows along the escape route in segments, with adjacent projection areas overlapping by 50% to ensure guidance continuity in smoky environments. The projection lights switch to green laser (wavelength 520nm, smoke penetration is 40% higher than red light), projecting a 1.5-meter-wide arrow 10 meters in front of personnel. The brightness is adaptively adjusted according to the smoke concentration (500-1500 lumens), with a flashing frequency of 2Hz, indicating "Proceed 300 meters in the direction of the arrow to the exit".

[0035] Terminal Interaction. The system links with the trapped person's mobile phone (an iPhone / Android device that supports UWB) via a UWB positioning module, triggering continuous vibration alerts on the phone (3 short vibrations followed by 1 long vibration to indicate "moving forward"). The AR glasses then display a virtual escape wall (a semi-transparent blue light strip) superimposed on the real-world scene.

[0036] Voice and resource delivery. The drone swarm coordinates resource delivery according to the task allocation protocol: dropping 4 high-temperature resistant silicone breathing masks (filter canisters support CO and particulate matter protection) to area P2. If the payload of a single drone is insufficient, the task is automatically split and executed by a nearby drone.

[0037] The drones share real-time locations (updating at 10Hz), and the speed obstacle method is used to predict trajectory conflicts. The priority rule is: escape route opening > resource drop > data collection. When a drone's battery level drops below 20% or it encounters an obstacle, the control center initiates task reassignment (e.g., if drone 1 is obstructed, drone 2 takes over its projection task). The control center receives real-time location information from fire trucks (GPS + tunnel internal beacons) and reserves a 3.5-meter-wide passage for rescue vehicles in the route planning. When a fire truck enters the southern section of the tunnel, the projector immediately switches the path arrow A to yellow, and a voice prompt says "Keep to the right and proceed slowly" to avoid people and vehicles colliding. After all personnel have evacuated in an orderly manner, the drone continuously broadcasts the voice command "Rescue complete, do not return." The communication relay drone maintains the Mesh network until the open flames are extinguished, transmitting real-time data on fire temperature and structural stability. The drone is then guided to automatically return to its nest for charging, and the system resets to standby mode.

[0038] Further optimized solutions in this embodiment include: to reduce costs, the lidar can be replaced with a binocular vision + ToF sensor, and the projection lamp can be replaced with an LED array; when extending the ultra-long tunnel, the density of drone nests can be increased to one group every 200 meters, and the communication drones can support satellite relay.

[0039] In this embodiment, the enhancement of the collaborative logic is explained as follows: ① Hierarchical task allocation: The control center dynamically allocates tasks through reinforcement learning. For example, it prioritizes scheduling the drone closest to the fire source to perform heat source scanning, while remote drones are responsible for opening up the outer perimeter.

[0040] ② Offline Emergency Mode: When the communication relay drone is damaged, the remaining drones activate the Ad-hoc protocol to maintain guidance continuity through optical communication and local decision-making.

[0041] ③ Resource Collaboration and Optimization: The material delivery mission adopts a distributed negotiation mechanism. The drones replenish supplies nearby based on the remaining payload, and the mission completion status is verified through the Mesh network.

[0042] It is worth noting that this embodiment only shows one deployment form of the tunnel traffic accident emergency escape drone collaborative guidance system and method, but it does not mean that the tunnel traffic accident emergency escape drone collaborative guidance system and method provided in this application has only this one form. The specific structural form can be adjusted according to the specific design.

[0043] In the description of this application, it should be understood that the terms "front", "rear", "upper", "lower", "outer", "inner", "horizontal", "top", "bottom", "surface", "bottom layer", "top layer", "upper part", "lower part", "bottom", "top", "inner", "surface", "center", "right side", "middle part", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing this application, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this application.

[0044] The embodiments described above are merely preferred embodiments of this application and are not intended to limit the scope of this application. Any modifications and improvements made by those skilled in the art to the technical solutions of this application without departing from the spirit of this application shall fall within the protection scope defined by the claims of this application.

Claims

1. A drone-assisted emergency escape system and method for tunnel traffic accidents, characterized in that, include: Distributed drone nesting units deployed within the tunnel, each containing multiple foldable emergency drones; An environmental sensing unit is used to detect temperature, gas concentration, and image information inside the tunnel. The control center generates escape routes based on environmental data and dispatches a swarm of drones. At least one guidance drone, equipped with a SLAM navigation module, thermal imager, projector, and emergency supplies delivery device; At least one communications relay drone is used to build a temporary wireless communication network.

2. The system according to claim 1, characterized in that: The guiding drone interacts with the trapped person's terminal via a UWB positioning module, the terminal including a mobile phone, smart bracelet or AR glasses.

3. The tunnel traffic accident emergency escape drone-assisted guidance system and method according to claim 1, characterized in that, The drone nest unit includes a catapult mechanism that can release the drone within 5 seconds and supports automatic homing and charging of the drone.

4. The tunnel traffic accident emergency escape drone collaborative guidance system and method according to claim 1, characterized in that, The projection light can project dynamic escape path arrows on the ground and automatically adjust the brightness and color wavelength according to the smoke concentration.

5. The tunnel traffic accident emergency escape drone collaborative guidance system and method according to claim 1, characterized in that, The control center uses reinforcement learning algorithms to optimize multi-UAV path planning, with the priority order being: escape route opening > survivor location > data transmission.

6. A method for emergency escape control in tunnels using unmanned aerial vehicles (UAVs), characterized in that, Includes the following steps: S1: Receives an accident trigger signal and activates the drone nesting unit in the corresponding area; S2: Communication relay drones take off to establish a mesh network and transmit real-time data inside the tunnel; S3: Guide the drone to scan and generate a 3D map of the tunnel, marking obstacles and heat sources; S4: The control center integrates multi-source data, generates a global escape route, and issues instructions; S5: Drones guide personnel to evacuate in an orderly manner through projection, voice, and terminal interaction, and simultaneously drop emergency supplies; S6: Continuously monitor and update escape routes.

7. A method for emergency escape control in a tunnel using unmanned aerial vehicles (UAVs) in collaboration with claim 6, characterized in that, The system dynamically avoids access lanes for rescue vehicles during route planning and updates the guidance direction using projection lights.

Images