Expressway lane cleaning system based on air-ground amphibious unmanned aerial vehicle

Abstract

The invention provides an expressway lane cleaning system based on an air-ground amphibious unmanned aerial vehicle. The system comprises an air-ground amphibious cleaning unmanned aerial vehicle set, a detection unmanned aerial vehicle, an accompanying type mobile supply vehicle and a control center. The air-ground amphibious unmanned aerial vehicle has the characteristics of two operation modes of ground running and air flying, and a collaborative cleaning scheduling strategy is designed: during conventional operation, the amphibious unmanned aerial vehicle keeps the ground running mode to perform dust collection and water spraying so as to reduce energy consumption and maintain operation stability; when it is detected that the unmanned aerial vehicle gives an early warning to a rear vehicle, too much road dust or insufficient resources, the control center instructs the related unmanned aerial vehicle to be switched to an air flight mode in real time, and air avoidance, high-altitude dust suppression or rapid station returning supply operation is executed. The problems that a traditional cleaning vehicle hinders traffic and a common unmanned aerial vehicle is insufficient in endurance are effectively solved, and high efficiency and intelligence of lane cleaning are achieved.

Description

Technical Field

[0001] This invention relates to the fields of intelligent lane cleaning and drone technology, specifically a highway lane cleaning system based on an amphibious drone. Background Technology

[0002] As the main artery of national transportation, the cleanliness of highways directly affects driving safety and road lifespan. Currently, routine cleaning and maintenance of highways mainly relies on large manned sweepers or manual labor. However, traditional operating methods have significant drawbacks: first, large sweepers are slow and bulky, easily causing traffic congestion and even rear-end collisions; second, manual labor in emergency lanes or specific areas faces extremely high safety risks from high-speed traffic; and third, traditional equipment lacks flexibility and is difficult to use for rapid and precise cleaning in response to emergencies or specific road sections.

[0003] With the development of unmanned systems technology, drones and unmanned vehicles are increasingly being applied to road inspection and cleaning. While existing single ground cleaning robots can save manpower, their obstacle avoidance capabilities are weak when facing complex and dynamic road conditions on highways, making it difficult to operate safely without interfering with normal traffic. On the other hand, purely aerial cleaning drones are limited by battery life and payload capacity, making it difficult to perform long-duration, high-intensity physical cleaning tasks. Furthermore, existing multi-drone systems often lack efficient collaborative scheduling mechanisms, making it impossible to achieve autonomous replenishment and strategy adjustments when energy is insufficient, materials are depleted, or extreme dust weather occurs. Summary of the Invention

[0004] To address the problems of existing highway lane cleaning technologies, such as cleaning vehicles disrupting normal traffic flow, high safety risks associated with manual and ground-based operations, and discontinuous operation of single-mode cleaning equipment, this invention proposes a highway lane cleaning system based on amphibious unmanned aerial vehicles (UAVs). This system uses a control center to coordinate the control of a UAV group capable of switching between driving and flight modes, along with an accompanying mobile supply vehicle. Leveraging its amphibious characteristics, it ensures low-energy ground operations while enabling rapid aerial avoidance of oncoming vehicles. Furthermore, dynamic formation reorganization, replacement, and accompanying resupply mechanisms address the issues of endurance and resource constraints. Therefore, while ensuring smooth traffic flow, this system significantly improves the safety, continuity, and environmental adaptability of highway cleaning operations. The technical solution provided by this invention is as follows:

[0005] A highway lane cleaning system based on amphibious unmanned aerial vehicles (UAVs) includes an amphibious cleaning UAV group, a reconnaissance UAV, an accompanying mobile supply vehicle, and a control center. The amphibious cleaning UAV group comprises models with both vacuuming and water spraying functions, and has both ground-based and aerial operating modes, equipped with a mechanical transformation mechanism for physical transformation between these modes. The reconnaissance UAV is used for high-altitude environmental perception. The accompanying mobile supply vehicle travels in the emergency lane, maintaining the same speed as the UAV group, providing a mobile take-off and landing platform and resource resupply. The control center executes the following scheduling logic:

[0006] The instructions state that the amphibious cleaning drone crews should maintain ground driving mode during routine cleaning operations, and only switch to aerial flight mode when avoiding obstacles, suppressing dust, or making rapid resupply.

[0007] When any cleaning drone is detected to have switched flight modes and left the formation for resupply due to resource depletion, the control center calculates the coverage gap width of the remaining formation and the actual travel speed of the remaining drones, and instructs the remaining drones to perform a zigzag reciprocating path on the ground at their actual travel speed to fill the road coverage gap.

[0008] Preferably, the amphibious unmanned aerial vehicle (UAV) group includes a flight-ground driving switching control subsystem, a dust extraction and water spraying cleaning subsystem, and a formation path planning subsystem. The flight-ground driving switching control subsystem is used to switch between a ground driving state with reduced energy consumption and a high-maneuverability air flight state according to the instructions of the control center. The supply vehicle includes a route planning system and a supply subsystem. The supply subsystem is equipped with a UAV recovery platform, a flexible docking interface based on visual servoing, a water tank, and a waste collection bin. The supply subsystem is used to dynamically supply the received UAVs in a relatively stationary state where the supply vehicle and the UAV are moving at the same speed, through visual recognition and robotic arm vibration compensation.

[0009] Preferably, during the task initialization phase, the control center determines the lane width based on the total lane width. Effective operating width of a single vacuum cleaning drone and safety interval According to the formula Calculate the required number of vacuum cleaning drones In the initial formation plan generated by the control center, the vacuuming drone formation is arranged in a "I" shape and operates in ground driving mode, the water spraying drone formation follows closely behind and operates in ground driving mode, and the detection drone maintains the aerial flight mode at the rear of the formation and transmits road condition data in real time.

[0010] When any cleaning drone switches flight mode and leaves the formation for resupply due to resource depletion, the control center calculates the actual speed of the remaining drones using the following formula. :

[0011]

[0012] in For the overall forward speed of the formation, To cover the gap width of the remaining formations, This represents the longitudinal step size for the reciprocating scan.

[0013] Preferably, the control center scheduling method also includes an air-to-ground mode switching avoidance strategy based on dynamic thresholds. This strategy requires the control center to calculate in real time the speed of oncoming vehicles transmitted back by the UAV. The safe reaction distance threshold is calculated according to the following formula. :

[0014]

[0015] in The current speed of the drone. To detect delay, The mechanical deformation time required for the UAV to perform mode switching. This refers to the vertical climb time. For a safe buffer distance;

[0016] When monitoring vehicle distance At that time, the control center will adjust the number of vehicles arriving. and preset vehicle threshold Response: If the number of vehicles is... The command instructs the cleaning drone to switch flight modes, ascend vertically, and hover to avoid an obstacle; if The command instructed the cleaning drone to switch flight modes, fly to the top of the mobile supply vehicle, and wait for recovery.

[0017] Preferably, the control center scheduling method also includes an adaptive flight dust suppression strategy based on dust height: when the detection drone detects the width of the road dust... When the threshold is exceeded, the control center determines the appropriate action based on the cone angle of the sprinkler drone's nozzles. and coverage redundancy coefficient Calculate the target's dust suppression altitude :

[0018]

[0019] The control center instructed the water-spraying drone to switch from ground driving mode to flight mode and ascend to the calculated altitude. Conduct wide-area spraying until the dust height returns to normal.

[0020] Preferably, the control center scheduling method also includes a circuit breaker response strategy based on the physical coverage capability of the formation: the control center counts in real time the number of drones with depleted power or resources. and the preset formation failure threshold Compare them.

[0021] like If the formation is determined to be unable to maintain effective coverage through speed compensation, the control center instructs all drones to switch from driving mode to flight mode and land on the mobile supply vehicle for centralized resupply and maintenance; if If the formation coverage dynamic compensation logic is executed, only the drones whose energy is depleted are instructed to leave the formation, and the remaining drones execute a "Z" shaped reciprocating path.

[0022] Preferably, the control center scheduling method also includes a dynamic formation reset strategy: after a stray drone completes charging, water filling, or garbage disposal on the mobile supply vehicle, the control center instructs it to take off and return to the formation; after the entire formation confirms its return to its position, the control center instructs all drones to switch back to ground driving mode and restore the initial "I"-shaped formation and standard operating speed. .

[0023] Preferably, the control center scheduling method also includes a relative navigation strategy in GPS-denied environments: the accompanying mobile supply vehicle is equipped with an ultra-wideband UWB positioning base station array; when the detection drone detects that the formation has entered a tunnel or an area where satellite signals are lost, the control center instructs the air-to-ground amphibious cleaning drone group to switch from satellite positioning mode to relative navigation mode; in relative navigation mode, the drone uses its onboard UWB tag to communicate with the supply vehicle base station for ranging, calculates its relative coordinates to the mobile supply vehicle in real time, and combines the visual inertial odometry data from the onboard downward-looking camera to correct lateral drift, so as to maintain the formation accuracy in the absence of satellite signals.

[0024] Preferably, the control center is equipped with a road debris classification and differentiated handling module based on a convolutional neural network (CNN): the control center receives road images collected by the detection drone, and uses a lightweight CNN model to identify road debris into three categories: lightweight waste, heavy obstacles, and hazardous fluids; if identified as lightweight waste, the vacuuming drone is instructed to maintain its normal speed. Adsorption is performed; if a heavy obstacle is identified, the drone in the corresponding lane is instructed to execute a flight avoidance strategy, skip the area and mark the coordinates; if a dangerous fluid is identified, the vacuuming drone is instructed to avoid it, and the water spraying drone behind it is instructed to switch to high-pressure cleaning mode for targeted removal.

[0025] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0026] 1. This invention adopts amphibious UAV collaborative operation, which combines the high energy efficiency of ground driving with the high maneuverability of air flight. It not only reduces the energy consumption of long-term cleaning operations, but also supports flexible aerial obstacle avoidance and remote scheduling functions.

[0027] 2. The present invention is a highway cleaning and maintenance system based on multiple drones. It can dynamically adjust the formation plan according to real-time power supply, resource availability and traffic conditions, effectively avoiding the safety problems of traditional cleaning operations that are slow to move and easily cause congestion on highways.

[0028] 3. For users, this invention introduces a mobile platform for accompanying resupply vehicles, eliminating the need for pre-planning and laying of fixed charging or water supply facilities along the route. This helps reduce the construction cost of highway cleaning facilities and extends the single-operation mileage of the drone team.

[0029] 4. This invention has a high degree of environmental adaptability. It can automatically switch to the aerial spraying mode to enhance the dust suppression effect in extreme dust weather, and can implement different avoidance strategies according to the number of vehicles approaching from behind, so as to minimize the interference with normal traffic flow while ensuring cleaning efficiency. Attached Figure Description

[0030] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0031] Figure 1 This is a flowchart of the main workflow of the drone cleaning system proposed in this invention. Detailed Implementation

[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] To make the above-mentioned objectives, features and effects of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0034] Example 1: A highway lane cleaning system based on amphibious drones. The system consists of a formation of several amphibious cleaning drones, a mobile supply vehicle, a high-altitude detection drone, and a control center deployed in the cloud or locally.

[0035] To achieve dual-modal operation, each cleaning drone employs a specially designed composite aerodynamic layout structure. The main fuselage frame is manufactured using a single piece of high-strength, lightweight carbon fiber composite material. Six hexagonal symmetrical rotor arms are distributed around the fuselage, with servo motor-driven folding mechanisms at the connection points between the rotor arms and the fuselage. When the control center issues a "ground driving mode" command, the servo motors drive the rotor arms to fold downwards or inwards towards the fuselage to conform to the fuselage contours and reduce wind resistance. When receiving an "airborne flight mode" command, the rotor arms fold downwards or inwards during mechanical deformation. The interior quickly unfolds and locks, setting .

[0036] The fuselage integrates an independently suspended four-wheel drive chassis at its underside. The wheels are covered with highly wear-resistant solid rubber tires, and each wheel houses a miniature hub motor, directly controlled by the flight-to-ground driving switching subsystem. During routine cleaning operations, the vehicle is driven solely by the four hub motors at the bottom. Cruise speed, set At this time, the system's power output is only 10% to 15% of that in flight mode, greatly extending the operating range on a single charge. The vacuum-type drone is equipped with a wide-width suction nozzle and a negative pressure fan on its underside, with a width... The sprinkler drone is equipped with a high-pressure micro-mist nozzle array on its belly, with a nozzle cone angle of [missing information]. Adjustable, with a self-sealing quick-fill water inlet on the side, and the water spray width is the same as that of a vacuuming drone. The top of the fuselage integrates an RTK-GPS antenna, the four sides of the fuselage integrate ultra-wideband (UWB) positioning tags, and the bottom integrates a downward-looking camera for visual-inertial odometry (VIO) calculation.

[0037] The accompanying mobile supply vehicle travels along the highway emergency lane. Through an onboard high-precision RTK-GPS positioning module, an autonomous driving algorithm controls the vehicle to maintain a relatively stationary, same-speed accompanying state with the cleaning drone convoy. UWB positioning base station arrays are installed on the vehicle's surface and at the four corners of the roof to provide relative coordinate references for the drones in environments where satellite signals are lost. The roof features a flat landing deck, covered with anti-slip material and integrated with a matrix-style electromagnetic adsorption device. When a drone lands and touches the deck, the electromagnets are energized to attract it, preventing it from slipping due to vehicle bumps. A flexible docking supply system based on visual servoing is integrated below the deck. This flexible docking supply system incorporates a dynamic vibration compensation algorithm into the robotic arm's end effector. When the supply vehicle is in the emergency lane... While traveling, the visual sensor captures the self-sealing quick-connect interface on the side of the drone in real time. The robotic arm makes millisecond-level pose corrections based on the slight bumps caused by the vehicle's movement, thus achieving zero-leak water filling and circuit connection in a relatively static but absolutely moving state, completing a one-stop dynamic refueling process of "charging, adding water, and emptying trash".

[0038] The reconnaissance drone is equipped with a high-resolution optoelectronic pod, which transmits real-time road surface image streams. The control center has a built-in lightweight convolutional neural network for real-time waste classification and scene semantic segmentation of the road surface images.

[0039] like Figure 1 As shown, the scheduling method executed by the control center includes the following steps:

[0040] Step 1: Before the task begins, the control center first performs task parameter calculations. To ensure that the cleaning operation covers the entire road surface without leaving any blind spots, the calculations are based on the total width of the lanes. Effective operating width of a single vacuum cleaning drone and safety interval According to the formula Calculate the required number of vacuum cleaning drones In this embodiment, the total width of the three lanes in one direction is set. Single-machine operation width Safety interval The system's calculation process and the generated initial formation parameters are shown in Table 1.

[0041] Table 1 Cleaning Task Parameters and Formation Calculation Table

[0042]

[0043] Based on the calculation results, the control center instructed six vacuuming drones and six water-spraying drones to be released from the supply vehicle, unfold their wheels, enter ground driving mode, and line up horizontally in a "I" shape, advancing at a constant speed of 20 km / h.

[0044] Step 2: During the routine operation phase (time T0), the system continuously monitors the physical modes and resource status of each node. As shown in Table 2, at this time, all personnel are in a low-energy ground driving state, and only the detection drone is in a high-altitude flight state for environmental perception, and the system operates smoothly. The system not only continuously monitors the status of each node, but also runs AI intelligent recognition logic in parallel.

[0045] Table 2 Real-time Status Monitoring Table for Amphibious Unmanned Aerial Vehicle Groups (Time T0)

[0046]

[0047] The control center uses a CNN model to perform real-time reasoning on road debris. If the debris is determined to be lightweight waste such as fallen leaves, paper scraps, or dust, the vacuuming drone is instructed to pass through and absorb it normally. If the debris is determined to be heavy obstacles such as stones or tire linings, the drone in the corresponding lane is instructed to switch its flight mode to overcome the obstacle and mark the coordinates to notify manual handling. If the debris is determined to be fluid stains such as oil or mud, the vacuuming drone is instructed to avoid it, while the water spraying drone behind it is instructed to switch to "high-pressure fixed-point cleaning mode" to spray cleaning fluid for removal.

[0048] Step 3: The control center monitors the signal-to-noise ratio (SNR) of the RTK-GPS signal and the ambient light intensity in real time. When the SNR is detected to be below the safe threshold and there is a sudden change in light intensity, such as when entering a tunnel, all personnel are instructed to activate "tunnel navigation mode." The safe threshold for SNR is set at 35 dBHz. The UAV navigation logic switches from an absolute latitude and longitude coordinate system to a relative coordinate system with the mobile supply vehicle as the origin. The UAV uses onboard UWB tags and the base station array on the surface of the supply vehicle to perform time-of-flight ranging and calculate the relative coordinates. At the same time, the visual odometry of the onboard downward-looking camera is activated to correct lateral drift by recognizing road texture features, ensuring that the formation can maintain centimeter-level accuracy in a coordinated formation even in tunnels without satellite signals.

[0049] Step 4: To address the safety hazard posed by oncoming vehicles at high speeds from behind on the highway, this system implements a dynamic threshold-based avoidance strategy. The detection drone scans the rear airspace in real time and calculates the relative speeds of oncoming vehicles. The control center combines sensor detection with delay. UAV mode switching time and vertical climb time The safe reaction distance threshold is dynamically calculated based on kinematic formulas. :

[0050]

[0051] Assuming the speed of the vehicle approaching from behind drone speed .set up , , Safety buffer The calculation yields: When the radar measures the vehicle distance At that time, the system based on the number of vehicles Compared with preset vehicle threshold Respond to the size relationship, assuming :

[0052] like The drone is instructed to switch flight mode, ascend vertically to a height of 5 meters, hover, and return to the ground after the vehicles have passed. If there is a continuous convoy behind, then... If the drone is not instructed to switch its flight mode, it will fly horizontally to the top of the mobile supply vehicle for recovery, and take advantage of the gap to carry out opportunistic resupply.

[0053] In the formula This is a unique physical parameter of the present invention. It quantifies the uncontrollable time window generated by the amphibious UAV during mechanical deformation, thereby upgrading the avoidance strategy from a simple "distance judgment" to "spatiotemporal prediction based on mechanical characteristics".

[0054] Step 5: To address the coverage gap caused by some UAVs leaving the formation for resupply due to resource depletion, this invention implements a path reconstruction strategy based on speed compensation.

[0055] It is important to note that the collaborative scheduling logic of this invention is fundamentally different from the "data packet routing scheduling" in traditional communication networks. The scheduling in this system is based on physical resources, such as remaining water volume and battery power, and involves physical spatial reconstruction. For example, when the system executes "candidate detection" or "Z-shaped reciprocating path planning" commands, the object of scheduling is a physical UAV with physical mass, motion inertia, and mechanical deformation time, rather than a massless information flow. The control center calculates... At this time, it is necessary to introduce an aerodynamic energy consumption model and the deformation time of the robotic arm. This is a technical feature that pure data scheduling systems do not possess, serving as a constraint.

[0056] At time T1, the system detected a water-spraying drone located in the middle of the formation. With its battery depleted, it was instructed to switch to flight mode, leave the formation, and head to the supply vehicle. At this point, the ground formation appeared with a width... Physical gaps. To ensure no area is missed and no drone falls behind, the control center instructed the remaining drones to maintain their overall longitudinal advance speed. Without changing the parameters, perform a left-right reciprocating "Z" shaped scan. The longitudinal step size of the reciprocating scan is based on the preset longitudinal step size. The control center calculates the target speed based on the compensation formula. :

[0057]

[0058] The control center then issued an instruction to increase the speed of the hub motors of the remaining UAVs to 25 km / h. The state changes and parameter comparisons at time T1 are shown in Table 3.

[0059] Table 3 Formation State Table Based on Z-Shaped Path Planning (Time T1)

[0060]

[0061] As shown in Table 3, the system achieves dynamic reconstruction of physical paths through precise mathematical model calculations, ensuring the continuity of cleaning operations.

[0062] Step 6: The control center monitors road dust conditions in real time using a detection drone. Assuming that a wide dust band is detected after the vacuuming operation, the width of the dust area transmitted back is... This means that dust has spread to adjacent lanes. To effectively suppress dust within this area, the control center calculates the target flight water spraying height using the following formula. :

[0063]

[0064] In this embodiment, the parameters are set as follows: the spray cone angle of the wide-angle nozzles mounted on the water-spraying drone. The set coverage redundancy coefficient The water spraying area should be 20% wider than the dust area to prevent escape. Substitute the values ​​into the calculation: Therefore, the control center sent a command to the water-spraying drone, instructing it to switch from ground driving mode to flight mode, ascend vertically to a height of 2.1 meters and hover, while simultaneously activating maximum flow spraying to cover a 3.5-meter-wide dust area using gravity settling, until the sensors detected the dust. After returning to normal, the drone was instructed to land and resume ground operation.

[0065] Step 7: The system counts the number of drones that have exhausted resources in real time. And set a failure threshold. The value here is set to 3.

[0066] like The aforementioned "Z"-shaped compensation strategy is implemented, allowing for single-unit rotational resupply. If... The formation was deemed unable to maintain effective coverage. The all-team shutdown mechanism was triggered, instructing all drones to switch flight modes and be collectively retrieved to the supply vehicle for centralized maintenance. Once the stray drones were resupplyed by the supply vehicle, they took off again, rejoined the formation, and filled the gaps. The entire formation immediately ceased its zigzag maneuvers, resumed straight-line travel, and reduced its speed. It restores the state to the lowest energy consumption.

[0067] Step 8: When the formation reaches the destination, the control center instructs all drones to fly back to the supply vehicle for recovery, uploads the operation log, and updates the mission status to "completed".

[0068] To verify the significant energy efficiency advantages of this system, this embodiment calculated and compared the energy consumption of a single cleaning task. The parameters were set as follows: lane length was... Average forward speed of drones The total homework time Assuming the ground driving power of this type of amphibious UAV... Airborne power In this mission, a total of 4 obstacle avoidance or obstacle crossing maneuvers were triggered, with each flight taking a certain amount of time. The rest of the time is spent traveling on the ground, therefore the flight time is... Travel time: According to the energy consumption calculation model Calculated If a traditional drone of the same specifications is used for full-flight operations, , ,but Calculation results show that, under the same operating conditions, the energy consumption of a single unit in this embodiment of the invention is only about 19.1% of that of the traditional all-flight operation mode, verifying the high energy efficiency of the system in long-distance highway cleaning tasks.

[0069] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A highway lane cleaning system based on an amphibious unmanned aerial vehicle (UAV), characterized in that, The system includes an amphibious cleaning drone unit, a reconnaissance drone, a mobile supply vehicle, and a control center. The amphibious cleaning drone unit includes models with both vacuuming and water spraying functions, and is capable of both ground-based and aerial operation modes, equipped with a mechanical transformation mechanism for physical transformation between the two modes. The reconnaissance drone is used for high-altitude environmental perception. The mobile supply vehicle travels in the emergency lane, maintaining the same speed as the drone unit, and provides a mobile take-off, landing, and resource resupply platform. The control center executes the following scheduling logic: The instructions state that the amphibious cleaning drone crews should maintain ground driving mode during routine cleaning operations, and only switch to aerial flight mode when avoiding obstacles, suppressing dust, or making rapid resupply. When any cleaning drone is detected to have switched flight modes and left the formation for resupply due to resource depletion, the control center calculates the coverage gap width of the remaining formation and the actual travel speed of the remaining drones, and instructs the remaining drones to perform a zigzag reciprocating path on the ground at their actual travel speed to fill the road coverage gap.

2. A highway lane cleaning system based on an amphibious unmanned aerial vehicle (UAV) according to claim 1, characterized in that, The amphibious unmanned aerial vehicle (UAV) group includes a flight-ground driving switching control subsystem, a dust extraction and water spraying cleaning subsystem, and a formation path planning subsystem. The flight-ground driving switching control subsystem is used to switch between a ground driving state with reduced energy consumption and a high-maneuverability air flight state according to the instructions of the control center. The supply vehicle includes a route planning system and a supply subsystem. The supply subsystem is equipped with a UAV recovery platform, a flexible docking interface based on visual servoing, a water tank, and a waste collection bin. The supply subsystem is used to dynamically supply the received UAVs in a relatively stationary state where the supply vehicle and the UAV are moving at the same speed, through visual recognition and robotic arm vibration compensation.

3. A highway lane cleaning system based on an amphibious unmanned aerial vehicle (UAV) according to claim 1, characterized in that, During the task initialization phase, the control center determines the lane width based on the total width of the lanes. Effective operating width of a single vacuum cleaning drone and safety interval According to the formula Calculate the required number of vacuum cleaning drones In the initial formation plan generated by the control center, the vacuuming drone formation is arranged in a "I" shape and operates in ground driving mode, the water spraying drone formation follows closely behind and operates in ground driving mode, and the detection drone maintains the aerial flight mode at the rear of the formation and transmits road condition data in real time. When any cleaning drone switches flight mode and leaves the formation for resupply due to resource depletion, the control center calculates the actual speed of the remaining drones using the following formula. : ； in For the overall forward speed of the formation, To cover the gap width of the remaining formations, This represents the longitudinal step size for the reciprocating scan.

4. A highway lane cleaning system based on an amphibious unmanned aerial vehicle (UAV) according to claim 3, characterized in that, The control center scheduling method also includes an air-to-ground mode switching avoidance strategy based on dynamic thresholds. This strategy requires the control center to calculate in real time the speed of oncoming vehicles transmitted back by the UAV. The safe reaction distance threshold is calculated according to the following formula. : ； in The current speed of the drone. To detect delay, The mechanical deformation time required for the UAV to perform mode switching. This refers to the vertical climb time. For a safe buffer distance; When monitoring vehicle distance At that time, the control center will adjust the number of vehicles arriving. and preset vehicle threshold Response: If the number of vehicles is... The command instructs the cleaning drone to switch flight modes, ascend vertically, and hover to avoid an obstacle; if The command instructed the cleaning drone to switch flight modes, fly to the top of the mobile supply vehicle, and wait for recovery.

5. A highway lane cleaning system based on an amphibious unmanned aerial vehicle (UAV) according to claim 3, characterized in that, The control center scheduling method also includes an adaptive flight dust suppression strategy based on dust height: when the detection drone detects the width of the dust on the road surface... When the threshold is exceeded, the control center determines the appropriate action based on the cone angle of the sprinkler drone's nozzles. and coverage redundancy coefficient Calculate the target's dust suppression altitude : ； The control center instructed the water-spraying drone to switch from ground driving mode to flight mode and ascend to the calculated altitude. Conduct wide-area spraying until the dust height returns to normal.

6. A highway lane cleaning system based on an amphibious unmanned aerial vehicle (UAV) according to claim 3, characterized in that, The control center scheduling method also includes a circuit breaker response strategy based on the physical coverage capability of the formation: the control center counts the number of drones with depleted power or resources in real time. and the preset formation failure threshold Perform a comparison; like If the formation is determined to be unable to maintain effective coverage through speed compensation, the control center instructs all UAVs to switch from driving mode to flight mode and land on the mobile supply vehicle for centralized supply and maintenance. like If the formation coverage dynamic compensation logic is executed, only the drones whose energy is depleted are instructed to leave the formation, and the remaining drones execute a "Z" shaped reciprocating path.

7. A highway lane cleaning system based on an amphibious unmanned aerial vehicle (UAV) according to claim 3, characterized in that, The control center's scheduling method also includes a dynamic formation reset strategy: after a stray drone completes charging, water filling, or garbage disposal on a mobile supply vehicle, the control center instructs it to take off and return to the formation; after the entire formation confirms its return to its position, the control center instructs all drones to switch back to ground driving mode and restore the initial "I"-shaped formation and standard operating speed. .

8. A highway lane cleaning system based on an amphibious unmanned aerial vehicle (UAV) according to claim 3, characterized in that, The control center's scheduling method also includes a relative navigation strategy in GPS-denied environments: the accompanying mobile supply vehicle is equipped with an ultra-wideband (UWB) positioning base station array; when the detection drone detects that the formation has entered a tunnel or an area where satellite signals are lost, the control center instructs the air-to-ground amphibious cleaning drone group to switch from satellite positioning mode to relative navigation mode; in relative navigation mode, the drone uses its onboard UWB tag to communicate with the supply vehicle's base station for ranging, calculates its relative coordinates to the mobile supply vehicle in real time, and combines the visual inertial odometry data from the onboard downward-looking camera to correct lateral drift, so as to maintain the formation accuracy in the absence of satellite signals.

9. A highway lane cleaning system based on an amphibious unmanned aerial vehicle (UAV) according to claim 1, characterized in that, The control center is equipped with a road debris classification and differentiated handling module based on convolutional neural networks (CNNs): the control center receives road images collected by the detection drone and uses a lightweight CNN model to identify road debris into three categories: lightweight waste, heavy obstacles, and hazardous fluids; if identified as lightweight waste, the vacuuming drone is instructed to maintain its normal speed. Adsorption is performed; if a heavy obstacle is identified, the drone in the corresponding lane is instructed to execute a flight avoidance strategy, skip the area and mark the coordinates; if a dangerous fluid is identified, the vacuuming drone is instructed to avoid it, and the water spraying drone behind it is instructed to switch to high-pressure cleaning mode for targeted removal.

Images