An inspection robot cooperative inspection system and method

Abstract

The present application relates to the technical field of highway inspection robots, in particular to a kind of inspection robot cooperative inspection system and method, including the roadside fence robot deployed in roadside guardrail and the road fence robot deployed in central separation belt guardrail, both are equipped with laser projection module, and are connected with cloud dispatch platform communication.When visibility is lower than preset threshold, cloud platform controls robot to move along guardrail, and controls roadside robot to project yellow shoulder light band, road fence robot to project white lane light band and green arrow, two light bands jointly constitute dynamic moving light corridor, provide continuous, active light-emitting driving guidance for vehicle in low-visibility environment.The present application utilizes the space cooperation of roadside and road fence robot and active laser projection, significantly improves the path recognition ability of driver in fog, rain and other bad weather, effectively reduces the risk of vehicle lane deviation and rear-end collision.

Description

Technical Field

[0001] This invention relates to the field of highway inspection robot technology, and in particular to a collaborative inspection system and method for inspection robots. Background Technology

[0002] When encountering low-visibility weather conditions such as heavy fog, heavy rain, or dust storms on highways, drivers struggle to see lane markings and curbs, greatly increasing the risk of serious traffic accidents such as vehicles veering out of their lanes and chain-reaction collisions. Existing solutions primarily rely on fixed reflective guide markers and delineators along the roadside, as well as vehicle headlights. However, the visibility of these passive reflective devices drops drastically in dense fog or heavy rain (typically less than 30 meters) and cannot dynamically adapt to changes in visibility. Vehicle headlights have limited penetration, and drivers can only judge distances by the taillights of vehicles ahead, resulting in insufficient reaction time. In recent years, inspection robot technology that moves along guardrails has emerged, but existing solutions are mostly standalone operations, primarily handling routine tasks such as guardrail inspection and road surface defect identification. They lack a collaborative mechanism between roadside and median fence robots, failing to leverage the complementary spatial positions of both to proactively guide traffic. Therefore, there is an urgent need for an active optical navigation system that can overcome the limitations of passive reflectivity and enable collaborative work between roadside and median fence robots to improve driving safety in low-visibility conditions. Summary of the Invention

[0003] In view of this, the purpose of this invention is to propose a collaborative inspection system and method for inspection robots, so as to solve the problem that current inspection robots cannot effectively improve driving safety under low visibility conditions.

[0004] To achieve the above objectives, the present invention provides a collaborative inspection system for inspection robots, comprising: At least one roadside fence robot deployed on the roadside guardrail, and at least one road center fence robot deployed on the median guardrail. The roadside fence robot is equipped with a first laser projection module, which is used to project a first light strip onto the road surface; The road-side fence robot is equipped with a second laser projection module, which is used to project a second light strip onto the road surface; And a cloud-based dispatch platform, which is communicatively connected to both the roadside fence robot and the road center fence robot; The cloud-based scheduling platform is configured to: acquire visibility sensor data of each robot; when visibility is lower than a preset threshold, control the roadside fence robot and the road center fence robot to move along the guardrail; and control the first laser projection module to project a first light strip and the second laser projection module to project a second light strip. The first light strip and the second light strip together constitute a dynamic light corridor to guide vehicle passage.

[0005] Optionally, the roadside fence robot and / or the road center fence robot are also equipped with forward millimeter-wave radar to detect the motion status of vehicles ahead in the lane, including distance, relative speed, absolute speed and deceleration. The cloud-based scheduling platform is further configured to: calculate the rear-end collision risk level based on the motion state, and control the first laser projection module and / or the second laser projection module to project a light strip according to the light strip encoding method corresponding to the rear-end collision risk level, wherein the light strip encoding method includes changes in light strip color, flashing frequency, light strip length and / or projection symbol.

[0006] Optionally, the rear-end collision risk levels include at least a caution level, a warning level, and a danger level; The optical band encoding method is as follows: Warning level: The light band color transitions from yellow to orange, the flicker frequency is 1Hz, the light band is continuous, and the projection symbol is slow or... ; Warning level: The light strip is orange, the flashing frequency is 3Hz, the light strip is continuous but its length is shortened to 60% of the normal length, and the projected symbol is a brake or... ; Danger level: The light strip is red, the flashing frequency is 6Hz, the light strip is in the shape of a dashed line and its length is shortened to 30% of the normal length, and the real-time distance value is projected.

[0007] Optionally, the cloud-based scheduling platform is further configured to: when a robot detects a dangerous rear-end collision risk, assign a decreasing rear-end collision risk level to the upstream robot based on the distance between the robot's position and other upstream robots, and control the upstream robot to project a light strip according to the corresponding level of light strip encoding method to form a rearward warning chain.

[0008] Optionally, the roadside fence robot is also equipped with a UWB relative positioning module. The cloud scheduling platform is configured to: use the UWB relative positioning module to measure the lateral distance between the roadside fence robot and the adjacent roadside fence robot, and adjust the lateral projection position of the second light strip according to the lateral distance so that the spacing between the second light strip and the first light strip matches the actual lane width.

[0009] Optionally, the roadside fence robot and / or the road center fence robot are also equipped with a rearward millimeter-wave radar to detect the speed and distance of vehicles behind in the same lane; The cloud-based dispatch platform is also configured to: calculate the rear-end collision time between the vehicle and the vehicle behind based on the speed and distance of the vehicle behind and the motion state of the vehicle in front; and when the rear-end collision time is lower than a preset threshold, control the first laser projection module and / or the second laser projection module to project light strips according to the light strip coding method of the danger level.

[0010] Optionally, the cloud-based dispatch platform is further configured to: when a traffic accident is determined to have occurred based on the fusion of radar and / or camera data reported by each robot, control the roadside fence robot upstream of the accident point to switch the first light strip to a red high-frequency flashing light strip and project the accident text, control the road center fence robot upstream of the accident point to switch the second light strip to a red fork-shaped projection, and broadcast the accident information to following vehicles via V2X communication.

[0011] Optionally, the cloud-based dispatch platform is also configured to: when receiving an approach signal from an emergency vehicle, control the roadside fence robot upstream of the accident point to switch the first light strip to an alternating green and white light strip, and control the roadside fence robot to project a right-pointing guide arrow to instruct other vehicles to give way.

[0012] Based on the same invention, this invention also provides a collaborative inspection method for inspection robots, comprising the following steps: S1: The cloud-based dispatch platform obtains visibility data reported by the roadside fence robot and the road center fence robot in real time; S2: When the visibility is lower than a preset threshold, the cloud scheduling platform controls the roadside fence robot and the road center fence robot to move along the guardrail, and controls the first laser projection module on the roadside fence robot to project a first light strip on the road surface and the second laser projection module on the road center fence robot to project a second light strip on the road surface. The first light strip and the second light strip together form a dynamic light corridor. S3: During the duration of the dynamic light corridor, the cloud scheduling platform dynamically adjusts the longitudinal spacing of the robots and the brightness and flicker frequency of the light strip according to the real-time visibility value.

[0013] Optional features include rear-end collision warning procedures: S4: The roadside fence robot and / or the road center fence robot detect the motion status of vehicles ahead in their lane using forward millimeter-wave radar, including distance, relative speed, absolute speed and deceleration, and report it to the cloud dispatch platform. S5: The cloud-based scheduling platform calculates the rear-end collision risk level based on the motion state, and controls the first laser projection module and / or the second laser projection module to project a light strip according to the light strip encoding method corresponding to the rear-end collision risk level. The light strip encoding method includes changes in light strip color, flashing frequency, light strip length, and / or projection symbol. S6: When a robot detects a dangerous rear-end collision risk, the cloud scheduling platform assigns a decreasing rear-end collision risk level to the upstream robot based on the distance between the robot's position and other upstream robots, and controls the upstream robot to project a light strip according to the corresponding level of light strip encoding method to form a rearward warning chain.

[0014] The cloud-based dispatch platform continuously receives data reported by the visibility sensors on each robot at a period of no more than 100ms. When any sensor measures visibility below a preset threshold (e.g., 100m) for more than 5 seconds, the cloud-based dispatch platform determines that it has entered a low visibility state and issues a light corridor navigation command to all roadside fence robots and road center fence robots in the affected road section. After receiving the command, each robot travels in the same direction at a uniform speed (default 20km / h). At the same time, the roadside fence robot controls its first laser projection module to project a continuous yellow light strip onto the road surface to the outer side to outline the road shoulder boundary. The road center fence robot controls its second laser projection module to project a white dashed light strip onto its own lane and a green arrow light strip onto the opposite lane. The area between the two light strips forms a driving channel that matches the lane width. The driver only needs to keep the vehicle between the two light strips to drive safely.

[0015] Compared to traditional passive reflective guide signs, this solution uses active laser projection, which can achieve a visible light strip distance of over 120m in low visibility conditions such as fog and rain. Through spatial coordination between roadside and median fence robots, it achieves continuous, dual-boundary optical guidance that moves along the road. Drivers can clearly identify the driving path without relying on lane markings, reducing the risk of vehicles deviating from their lanes and rear-end collisions in low visibility conditions, and improving driving safety in low visibility conditions. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of the low-visibility dynamic light corridor activation and control process according to an embodiment of the present invention; Figure 2 This is a flowchart of the rear-end collision risk detection and collaborative early warning chain according to an embodiment of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0019] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "include" or "comprising" mean that the element or object preceding the term covers the element or object listed after the term and its equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "up," "down," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0020] A collaborative inspection system using inspection robots includes: At least one roadside fence robot deployed on the roadside guardrail, and at least one road center fence robot deployed on the median guardrail. The roadside fence robot is equipped with a first laser projection module, which is used to project a first light strip onto the road surface; The road-side fence robot is equipped with a second laser projection module, which is used to project a second light strip onto the road surface; And a cloud-based dispatch platform, which is communicatively connected to both the roadside fence robot and the road center fence robot; The cloud-based scheduling platform is configured to: acquire visibility sensor data of each robot; when visibility is lower than a preset threshold, control the roadside fence robot and the road center fence robot to move along the guardrail; and control the first laser projection module to project a first light strip and the second laser projection module to project a second light strip. The first light strip and the second light strip together constitute a dynamic light corridor to guide vehicle passage.

[0021] The roadside fence robot is installed on the corrugated guardrails on both sides of the road. It adopts a bottom-supported walking mechanism, and achieves stable movement along the guardrail by engaging the drive wheels with the corrugated grooves on the lower edge of the guardrail and being attracted by permanent magnets. The median fence robot is installed on the guardrail of the central divider, and adopts a double-sided clamping walking mechanism. It moves by clamping the guardrail from both sides with two sets of drive wheels. The first and second laser projection modules are both composed of semiconductor lasers, collimating lenses, diffractive optical elements, and digital micromirrors, which can project a continuous light band with a width of 20cm on the ground. The visibility sensor is a transmission-type visibility meter with a measurement range of 10~2000m, installed on the top of the robot. The cloud scheduling platform communicates with each robot through a 4G / 5G network, with a communication cycle of no more than 100ms.

[0022] The cloud-based dispatch platform continuously receives data reported by the visibility sensors on each robot at a period of no more than 100ms. When any sensor measures visibility below a preset threshold (e.g., 100m) for more than 5 seconds, the cloud-based dispatch platform determines that it has entered a low visibility state and issues a light corridor navigation command to all roadside fence robots and road center fence robots in the affected road section. After receiving the command, each robot travels in the same direction at a uniform speed (default 20km / h). At the same time, the roadside fence robot controls its first laser projection module to project a continuous yellow light strip onto the road surface to the outer side to outline the road shoulder boundary. The road center fence robot controls its second laser projection module to project a white dashed light strip onto its own lane and a green arrow light strip onto the opposite lane. The area between the two light strips forms a driving channel that matches the lane width. The driver only needs to keep the vehicle between the two light strips to drive safely.

[0023] Compared to traditional passive reflective guide signs, this solution uses active laser projection, which can achieve a visible light strip distance of over 120m in low visibility conditions such as fog and rain. Through spatial coordination between roadside and median fence robots, it achieves continuous, dual-boundary optical guidance that moves along the road. Drivers can clearly identify the driving path without relying on lane markings, reducing the risk of vehicles deviating from their lanes and rear-end collisions in low visibility conditions, and improving driving safety in low visibility conditions.

[0024] In some embodiments, the roadside fence robot and / or the center-of-the-road fence robot are also equipped with forward-facing millimeter-wave radar for detecting the motion state of vehicles ahead in the lane, the motion state including distance, relative speed, absolute speed and deceleration; The cloud-based scheduling platform is further configured to: calculate the rear-end collision risk level based on the motion state, and control the first laser projection module and / or the second laser projection module to project a light strip according to the light strip encoding method corresponding to the rear-end collision risk level, wherein the light strip encoding method includes changes in light strip color, flashing frequency, light strip length and / or projection symbol.

[0025] The forward-facing millimeter-wave radar operates at a frequency of 77 GHz, has a detection range of 150 m, a detection angle of ±30°, and an update frequency of 20 Hz. It can output the distance, relative speed, azimuth angle, and radar cross-section of each target. The cloud-based scheduling platform is also equipped with a rear-end collision risk level calculation module. This module calculates the target's absolute speed based on the robot's own speed and the relative speed measured by the radar, and then calculates the target's deceleration by time difference. The deceleration calculation formula is a_dec = (v_abs(t) v_abs(t-Δt)) / Δt, where Δt=0.05s. The rear-end collision risk level is divided into three levels based on the absolute value of the deceleration: when the absolute value of the deceleration is greater than 2 m / s² and less than or equal to 4 m / s², it is the attention level; when it is greater than 4 m / s² and less than or equal to 6 m / s², it is the warning level; and when it is greater than 6 m / s², it is the danger level.

[0026] As the roadside fence robot and the center-of-road fence robot move along the guardrail, their forward-facing millimeter-wave radar continuously scans vehicles within a 150m range ahead of their lane, outputting the distance, relative speed, and azimuth of each target every 50ms. The robot adds its own speed to the relative speed to obtain the target's absolute speed, and divides the difference between the absolute speeds at two consecutive moments by the time interval to obtain the target's deceleration. When the absolute value of the deceleration exceeds 2m / s², the robot reports the target and its motion status to the cloud-based dispatch platform. The cloud-based dispatch platform determines the rear-end collision risk level (attention level, warning level, or danger level) based on the absolute value of the deceleration, and then generates corresponding light strip control commands to control the first laser projection module and / or the second laser projection module to project according to the light strip color, flashing frequency, light strip length, and projection symbol corresponding to that level.

[0027] Leveraging the ability of millimeter-wave radar to penetrate fog, rain, and smoke, this solution can accurately detect the movement of vehicles ahead even in conditions of extremely low visibility, solving the fundamental problem that drivers cannot visually observe the dynamics of vehicles in front. Simultaneously, by converting complex deceleration data into light strip codes (color, frequency, length, and symbols) that are intuitively perceptible to drivers, it achieves efficient and delay-free transmission of hazard information from machine to human. Drivers can receive road condition warnings without looking down at the dashboard, enhancing the ability to prevent rear-end collisions.

[0028] In some embodiments, the rear-end collision risk level includes at least a caution level, a warning level, and a danger level; The optical band encoding method is as follows: Warning level: The light band color transitions from yellow to orange, the flicker frequency is 1Hz, the light band is continuous, and the projection symbol is slow or... ; Warning level: The light strip is orange, the flashing frequency is 3Hz, the light strip is continuous but its length is shortened to 60% of the normal length, and the projected symbol is a brake or... ; Danger level: The light strip is red, the flashing frequency is 6Hz, the light strip is in the shape of a dashed line and its length is shortened to 30% of the normal length, and the real-time distance value is projected.

[0029] The specific light strip encoding corresponding to the attention level is as follows: The first or second laser projection module sets the light strip color to a gradient between yellow and orange (achieved by continuously changing the laser power), with a flashing frequency of 1Hz (i.e., 0.5 seconds on, 0.5 seconds off). The light strip remains continuous and uninterrupted, while the digital micromirror device projects slow-motion text or other text onto the ground at a frequency of 1Hz. The symbol has a font height of 0.3m. The specific light strip code corresponding to the warning level is as follows: the light strip color is orange (wavelength 590~595nm), the flashing frequency is 3Hz (on for 0.167 seconds, off for 0.167 seconds), the light strip is continuous but its length is shortened to 60% of the normal length (i.e., by controlling the laser module to project only 0.6 meters per meter), and simultaneously projecting brake letters or... The symbol; the light band code corresponding to the danger level is as follows: the light band color is red (wavelength 650nm), the flashing frequency is 6Hz (on for 0.083 seconds, off for 0.083 seconds), the light band is in the shape of a dashed line, that is, each bright band is 0.5m long and each dark band is 0.5m long, the total length is shortened to 30% of the normal length, and at the same time, the real-time distance value (e.g. 30m) is projected at a frequency of 2Hz. The distance value is rounded and dynamically generated by the digital micromirror device.

[0030] Once the cloud-based dispatch platform determines the rear-end collision risk level, it packages the corresponding light strip encoding parameters (color value, flashing period, duty cycle, dashed line pattern, and projected content) into control commands and sends them to the target robot. The robot's laser drive circuit controls the on / off state of the laser according to the flashing period and duty cycle to achieve the specified flashing frequency. For the control of the light strip length, the robot uses a high-speed switch of a digital micromirror device: at the attention level, the digital micromirror device remains on with a 100% duty cycle; at the warning level, the digital micromirror device switches on and off with a 60% duty cycle and a 1m cycle, so that only 0.6m of each 1m length is illuminated; at the danger level, the digital micromirror device operates with a 30% duty cycle and a superimposed 0.5m bright / 0.5m dark cycle. The projected symbols and numbers are projected onto the road surface with high brightness during the intervals of the light strip flashing (i.e., the periods when the light is off) after the digital micromirror device receives the graphic dot matrix data sent from the cloud.

[0031] This solution constructs a progressive early warning system from attention to warning and then to danger through three light band encodings with visually significant differences (yellow - orange gradient → orange → red, low frequency 1Hz → medium frequency 3Hz → high frequency 6Hz, continuous → shorter → dotted line, no symbol → warning symbol → real - time distance value); without the need for drivers to learn complex rules, they can instinctively perceive the escalation of the danger level - the red high - frequency dotted light band is significantly more urgent than the yellow low - frequency continuous light band, enabling different degrees of deceleration or avoidance measures to be taken according to the danger level, thus achieving the accurate transmission of risk classification.

[0032] In some embodiments, the cloud scheduling platform is further configured to: when a certain robot detects a dangerous - level rear - end collision risk, according to the distance between the position of this robot and other upstream robots, assign a decreasing rear - end collision risk level to the upstream robots, and control the upstream robots to project light bands in the corresponding light band encoding mode according to the level, forming a warning chain that is transmitted backward.

[0033] The other upstream robots refer to other roadside fence robots and median fence robots arranged along the same driving direction behind the robot that detects the dangerous - level risk; the rule for assigning the decreasing rear - end collision risk level is: let the mileage stake number of the robot that detects the dangerous - level risk be P0, and the mileage stake number of an upstream robot be Pi and Pi < P0, and the distance between the two is d = P0 - Pi; when d ≤ 50m, this upstream robot is assigned the dangerous level; when 50m < d ≤ 150m, the warning level is assigned; when 150m < d ≤ 300m, the attention level is assigned; when d > 300m, no early warning is assigned; the cloud scheduling platform updates the warning chain assignment every 0.5 seconds and sends the corresponding light band encoding instruction downward; after receiving the instruction, the upstream robot projects the light band in the corresponding light band encoding mode according to claim 3.

[0034] When a certain robot (designated as R0) detects that the deceleration of the vehicle in front reaches the dangerous level (i.e., greater than 6 m / s²) through the forward millimeter-wave radar, R0 immediately reports its detection result to the cloud scheduling platform, and the cloud platform records the mileage stake number P0 of R0; Subsequently, the cloud platform traverses the mileage stake numbers Pi of all other robots on this section of the road, filters out the robots with Pi < P0 (i.e., located behind R0), and calculates the distance difference d between each rear robot and R0 one by one; According to the interval that d falls into (≤50m, 50 - 150m, 150 - 300m), assign the corresponding warning level (dangerous level, warning level, attention level) to this robot, and generate the corresponding light band coding control instruction to send down; Each upstream robot projects the light band according to the assigned level, forming a warning chain that propagates backward from the dangerous point. The closer the robot is to the dangerous point, the higher the warning level and the stronger the light band warning effect; This warning chain is dynamically updated every 0.5 seconds. When the danger is lifted (the deceleration of the vehicle in front returns to normal) or the position of the dangerous point changes, the warning chain is adjusted or revoked accordingly.

[0035] Through the relay transmission of dangerous information, this solution extends the local dangerous event detected by a single robot backward to a range of 300m, enabling multiple vehicles behind to obtain warning information that matches their own distances in sequence before reaching the dangerous point; Compared with the solution where only one robot issues a warning, the warning chain mechanism effectively solves the problem that vehicles at a long distance cannot directly see the light band in front under low visibility conditions, avoids the occurrence of chain-reaction rear-end collisions, and realizes the leap from the warning ability of a single robot to the swarm intelligence of the entire robot cluster.

[0036] In some embodiments, the middle-of-road fence robot also carries a UWB relative positioning module, and the cloud scheduling platform is configured to: use the UWB relative positioning module to measure the lateral distance between the middle-of-road fence robot and the adjacent roadside fence robot, and adjust the lateral projection position of the second light band according to the lateral distance, so that the distance between the second light band and the first light band matches the actual lane width.

[0037] The UWB relative positioning module operates in the 3.5~6.5GHz frequency band, with a ranging accuracy of ±5cm and a ranging frequency of 10Hz. The roadside fence robot and the adjacent roadside fence robot exchange ranging signals in real time through the UWB module to measure the straight-line lateral distance between them. The cloud scheduling platform is equipped with a lane width matching algorithm. This algorithm calculates the theoretical lane width W_lane = (W_actual 0.5) / 2 based on the measured lateral distance W_actual and the preset shoulder width parameter (default 0.25m), and further calculates the lateral projection offset L_offset = W_lane 0.5m of the second light strip of the roadside fence robot. The second laser projection module of the roadside fence robot is equipped with an adjustable projection angle galvanometer. The cloud scheduling platform converts the offset into a rotation angle command for the galvanometer, so that the landing position of the white dashed light strip meets the requirement that the distance between it and the yellow light strip on the roadside is equal to the actual lane width.

[0038] In dynamic light corridor navigation mode, the road-center fence robot sends UWB ranging requests to the adjacent roadside fence robots at a frequency of 10Hz. The roadside fence robots respond with ranging requests, and the two calculate the precise lateral distance using the two-way time-of-flight method. The road-center fence robot reports this lateral distance to the cloud scheduling platform in real time. Based on this distance value and the standard shoulder width (0.25m on each side), the cloud scheduling platform calculates the actual lane width of the current road segment. Then, the cloud platform calculates the lateral position where the white dashed light strip should be projected (usually located about 0.5m to the left of the center of the lane) and converts it into the deflection angle of the galvanometer in the laser projection module of the road-center fence robot. The galvanometer adjusts the laser emission angle according to the instructions so that the white light strip falls precisely on the target position, thereby ensuring that the distance between the white light strip and the yellow shoulder light strip is always equal to the actual lane width (with an error controlled within ±0.1m). This adjustment process is carried out continuously in real time, and when the lateral spacing of the road changes due to construction, settlement, or other factors, the projection position of the light strip automatically adjusts accordingly.

[0039] This solution addresses the technical problem of light corridor failure caused by changes in road geometry parameters. Through real-time UWB ranging and dynamic adjustment of projection angle, the dynamic light corridor can adaptively match the actual lane width of any road segment. Even on curves, ramps, or construction sections with uneven guardrail spacing, the light corridor still maintains accurate guidance. Compared to solutions that pre-store fixed lane width data, this solution does not rely on real-time updates of high-precision maps, reducing the system's dependence on map data and improving the robustness and universality of light corridor guidance.

[0040] In some embodiments, the roadside fence robot and / or the center-of-the-road fence robot are also equipped with a rearward millimeter-wave radar for detecting the speed and distance of vehicles behind in the same lane; The cloud-based dispatch platform is also configured to: calculate the rear-end collision time between the vehicle and the vehicle behind based on the speed and distance of the vehicle behind and the motion state of the vehicle in front; and when the rear-end collision time is lower than a preset threshold, control the first laser projection module and / or the second laser projection module to project light strips according to the light strip coding method of the danger level.

[0041] The rearward millimeter-wave radar operates at a frequency of 77GHz, with a detection range of 80m, a detection angle of ±15°, and an update frequency of 20Hz. It is used to detect the speed and distance of vehicles behind in the same lane. The cloud-based dispatch platform is equipped with a rear-end collision time calculation module. The formula for calculating the rear-end collision time TTC is: TTC = d_rear / (v_rear v_front), where d_rear is the distance between the robot and the following vehicle as measured by the rearward radar, v_rear is the absolute speed of the following vehicle as measured by the rearward radar, and v_front is the absolute speed of the preceding vehicle as measured by the forward radar (or the robot's own speed; when there is no preceding vehicle behind the robot, the robot's own speed is used). The preset threshold is 3 seconds. When the TTC is less than 3 seconds, the cloud-based dispatch platform determines that there is a risk of a rear-end collision and controls the first laser projection module and / or the second laser projection module of the corresponding robot to project a light strip according to the light strip encoding method of the danger level in claim 3 (red, 6Hz flashing, dashed line, projection real-time distance). This light strip faces the oncoming vehicle.

[0042] In low-visibility light corridor navigation mode, the rearward millimeter-wave radar of the roadside fence robot and the center-of-the-road fence robot continuously scans vehicles within an 80m range behind their respective lanes, outputting the distance and speed of vehicles behind in real time. Simultaneously, the robot's forward millimeter-wave radar (or its own speed sensor) provides speed information for vehicles ahead. The cloud-based scheduling platform receives these two sets of data at a frequency of 20Hz and uses them to calculate the time to collision (TTC) using the formula: TTC = The distance to the following vehicle is calculated as (the speed of the following vehicle is the speed of the vehicle in front). When the TTC (Time To Till) is less than 3 seconds, it means that the following vehicle will collide with the vehicle in front (or this robot) within 3 seconds at the current speed. The cloud immediately determines that there is a high risk of rear-end collision and issues a hazard-level light strip coding instruction to the corresponding robot. The robot then switches the light strip of its laser projection module to a red high-frequency (6Hz) dashed light strip and projects the real-time distance value between the following vehicle and the vehicle in front (e.g., 20m) at a frequency of 2Hz. This light strip and projection information are directly facing the oncoming vehicle, prompting the driver of the following vehicle to take immediate emergency braking measures. When the TTC rises back to more than 3 seconds, the system automatically exits the hazard-level warning and returns to the normal light corridor mode corresponding to the current visibility.

[0043] This solution actively monitors the approaching speed of vehicles behind using rearward millimeter-wave radar, and accurately calculates the time of rear-end collision by combining the speed data of the vehicle in front. This fills the technological gap where traditional forward warning systems are powerless against the risk of rear-end collisions. When the driver of the following vehicle cannot see the vehicle in front due to dense fog or fails to slow down in time due to fatigue, the robot actively projects a highly penetrating red warning light strip behind, forcibly attracting the attention of the driver of the following vehicle and effectively avoiding chain accidents caused by rear-end collisions.

[0044] In some embodiments, the cloud-based dispatch platform is further configured to: when a traffic accident is determined to have occurred based on the fusion of radar and / or camera data reported by each robot, control the roadside fence robot upstream of the accident point to switch the first light strip to a red high-frequency flashing light strip and project accident text, control the road center fence robot upstream of the accident point to switch the second light strip to a red fork-shaped projection, and broadcast the accident information to following vehicles via V2X communication.

[0045] The specific method for determining traffic accidents through radar and / or camera data fusion is as follows: The millimeter-wave radar of the roadside fence robot detects a stationary obstacle (speed < 2 km / h) in its lane for more than 5 seconds; simultaneously, the camera of the roadside fence robot identifies features such as vehicle overturning, airbag deployment, and people getting out of the vehicle through a deep learning model (YOLOv8); the bidirectional camera of the center-of-the-road fence robot detects the same features in the opposite lane; after receiving intermediate or higher-level alarms reported by at least two different robots (or two different sensor channels of the same robot) within a spatial distance of 50 m and a time difference of 3 seconds, the cloud dispatch platform performs data fusion and determination. The system identifies a traffic accident and records the coordinates of the accident center point (using the centroid of multiple reported coordinates). After the determination, the cloud-based dispatch platform controls all roadside fence robots within 200m upstream of the accident point to switch the first light strip to a red high-frequency flashing light strip (6Hz, 100% brightness) and project the accident text. It also controls all roadside fence robots within 200m upstream of the accident point to switch the second light strip to a red fork-shaped projection (each fork is 1.5m × 1m in size, spaced 10m apart). Simultaneously, it broadcasts the accident information to connected vehicles within a 2km radius via V2X communication and pushes the accident location and on-site data to the traffic management center and navigation map platform.

[0046] In dynamic light corridor navigation mode, each robot continuously runs an accident detection algorithm. When a robot detects a stationary obstacle in its lane for more than 5 seconds using its forward radar, and its infrared camera identifies accident-related features (such as abnormal vehicle posture or people getting out of the vehicle) through a deep learning model, the robot reports a medium-level alarm to the cloud dispatch platform, along with the coordinates of the suspected accident point. Upon receiving the first alarm, the cloud dispatch platform initiates a timing and spatial search. If it receives the same alarm from another robot (or another sensor channel of the same robot) within 3 seconds and 50 meters, it then... The accident was determined to be a genuine traffic accident. The cloud platform then calculated all robots within a 200m range upstream of the accident point and issued accident warning commands to these robots. These robots immediately performed light strip switching operations: the roadside fence robot switched its original yellow light strip to a red high-frequency flashing light strip and projected accident text using digital micromirrors; the road center fence robot switched its white dashed light strip to a red cross-shaped projection. At the same time, all controlled robots broadcast accident warning messages to the outside world through the V2X module, and the cloud platform pushed accident information to the traffic management center and navigation map platform through the API interface.

[0047] This solution reduces the false alarm rate of traffic accidents by using a fusion decision mechanism based on multi-source heterogeneous sensors, avoiding invalid warnings caused by sensor misdetection (such as spilled material being mistaken for a parked vehicle). After an accident occurs, the system can complete the light band switching and information broadcasting within seconds without manual intervention. Compared with traditional manual alarms, this solution shortens the accident warning delay and effectively reduces the risk of secondary accidents caused by following vehicles entering the accident area. At the same time, V2X broadcasting and navigation map push provide following vehicles with accurate accident locations and detour suggestions, improving the overall traffic efficiency of the road network.

[0048] In some embodiments, the cloud-based dispatch platform is further configured to: when receiving an approach signal from an emergency vehicle, control the roadside fence robot upstream of the accident point to switch the first light strip to an alternating green and white light strip, and control the roadside fence robot to project a right-pointing guide arrow to instruct other vehicles to give way.

[0049] The emergency vehicle approach signal is received via V2X communication. This message includes the vehicle type (ambulance, fire truck, police car), location, speed, and direction of travel. After receiving the emergency vehicle approach signal, the cloud dispatch platform determines the distance between the emergency vehicle and the accident site. When the distance is less than 5km, the escort mode is activated. In the escort mode, the cloud dispatch platform controls all roadside fence robots between 300m upstream of the accident site and the accident site to switch the first light strip to an alternating green and white light strip (0.5 seconds green and 0.5 seconds white per second), and controls all roadside fence robots in the same road segment to project right-pointing guide arrows (→→→ symbols, each arrow spaced 20m apart) to instruct other vehicles to change lanes to the right shoulder to avoid the emergency vehicle. The escort mode continues until the emergency vehicle passes 100m behind the accident site or automatically terminates after the emergency vehicle sends a mission completion signal.

[0050] When an emergency vehicle approaches the accident scene, its onboard V2X device broadcasts an EmergencyVehicle Alert message every 100ms. Upon receiving this message, the V2X modules of the roadside fence robots and the roadside fence robots parse the emergency vehicle's position, speed, and direction of travel, and forward this information to the cloud dispatch platform. The cloud dispatch platform compares the emergency vehicle's position with the identified accident location. If the emergency vehicle is less than 5km from the accident location, it activates escort mode. The cloud platform determines the road segment from 300m upstream of the accident location to the accident location itself and sends instructions to all roadside fence robots within this segment, requiring them to switch their light strips to an alternating green and white pattern (changing color once per second) to illuminate the roadside fence. All roadside fence robots within the section send instructions to project right-pointing guide arrows. When drivers of other vehicles see the alternating green and white light strips and the right-pointing arrow projection on the road, they naturally understand that they should change lanes to the right shoulder to make way for emergency vehicles, clearing the left lane or emergency lane. The escort mode continues to operate until the cloud platform detects that the emergency vehicle has passed the accident point and is more than 100 meters behind it, or receives a mission completion signal from the emergency vehicle. The system then reverts to the accident warning mode (red light strip + red cross projection) or, depending on visibility, to the normal light corridor mode.

[0051] This solution utilizes the dynamic changes in light strip colors and symbols to create a visual, unmanned green channel for emergency vehicles. Drivers of other vehicles can intuitively understand the direction and timing of avoidance through the visual guidance of the road light strips (alternating green and white + right arrow). Compared to traditional methods that rely on sirens and shouts, this solution remains effective in low-visibility conditions such as dense fog and does not generate additional noise interference. At the same time, this solution is entirely based on light strip coding, eliminating the need for additional road signs or traffic lights. It has low implementation costs, wide coverage, and can shorten the time for emergency vehicles to reach the accident scene, improving emergency rescue efficiency.

[0052] To further implement the present invention, such as Figure 1 As shown, the present invention also provides a collaborative inspection method for inspection robots, comprising the following steps: S1: The cloud-based dispatch platform obtains visibility data reported by the roadside fence robot and the road center fence robot in real time; S2: When the visibility is lower than a preset threshold, the cloud scheduling platform controls the roadside fence robot and the road center fence robot to move along the guardrail, and controls the first laser projection module on the roadside fence robot to project a first light strip on the road surface and the second laser projection module on the road center fence robot to project a second light strip on the road surface. The first light strip and the second light strip together form a dynamic light corridor. S3: During the duration of the dynamic light corridor, the cloud scheduling platform dynamically adjusts the longitudinal spacing of the robots and the brightness and flicker frequency of the light strip according to the real-time visibility value.

[0053] In step S1, the cloud-based scheduling platform polls the visibility sensor data of all roadside fence robots and center fence robots every 100ms. The data reported by each robot includes the visibility value (unit: m), the robot's current mileage marker, and a timestamp. In step S2, when the visibility value reported by any robot is ≤100m within 5 consecutive seconds, the cloud-based scheduling platform triggers the light corridor navigation mode and issues formation instructions to all robots in the affected road section (extending 500m before and after the area with visibility ≤100m), including the target speed (default 20km / h), the target longitudinal spacing (determined according to Table 1), and the light strip parameters (color, brightness, flashing frequency). After receiving the instructions, each robot adjusts the drive motor speed through the PID controller to track the target speed, exchanges position information with adjacent robots through V2X to maintain the target spacing, and simultaneously starts the laser projection module to project the light strip according to the instructions. In step S3, the cloud-based scheduling platform continuously monitors the real-time visibility value reported by each robot. When the visibility changes, it dynamically adjusts the spacing and light strip parameters in Table 1 and reissues them.

[0054] The cloud-based scheduling platform sends data requests to all online robots at 100ms intervals. Each robot replies with its current visibility value, location, and status. The cloud platform maintains a road segment-visibility mapping table. When the minimum visibility of any road segment is detected to be below 100m for 5 seconds, the platform marks that road segment as a light corridor navigation activation zone and issues an activation command to all robots within that zone. After receiving the activation command, the robot first calculates its target position in the formation (sorted according to mileage markers), and then adjusts the motor speed through a PID controller to achieve a uniform target speed. The robot operates at a speed of 20 km / h. Simultaneously, it exchanges positions with adjacent robots every 100 ms via V2X, and uses a consensus algorithm to fine-tune the speed to keep the error between the actual distance and the target distance within ±2m. After that, the robot starts the laser projection module and begins to project a light strip according to the light strip parameters (color, brightness, flicker frequency) corresponding to the current visibility. The cloud scheduling platform continuously receives visibility data. When the visibility of a certain road section improves to more than 150m and lasts for 30 seconds, the platform issues an exit command, and the robot stops projecting the light strip and enters standby or low-power mode.

[0055] This method provides a complete, programmable collaborative inspection control process, realizing fully automated closed-loop management from data acquisition, condition judgment, formation control to light strip projection. Through periodic polling and dynamic adjustment mechanisms, the system can respond to changes in visibility in real time and automatically switch working modes when weather conditions fluctuate, without manual intervention. At the same time, the PID regulation and consistency algorithm in formation control ensure the synchronization of multi-robot movement and the stability of spacing, providing an underlying guarantee for the continuity and accuracy of the light corridor.

[0056] In some embodiments, such as Figure 2 As shown, it also includes rear-end collision warning steps: S4: The roadside fence robot and / or the road center fence robot detect the motion status of vehicles ahead in their lane using forward millimeter-wave radar, including distance, relative speed, absolute speed and deceleration, and report it to the cloud dispatch platform. S5: The cloud-based scheduling platform calculates the rear-end collision risk level based on the motion state, and controls the first laser projection module and / or the second laser projection module to project a light strip according to the light strip encoding method corresponding to the rear-end collision risk level. The light strip encoding method includes changes in light strip color, flashing frequency, light strip length, and / or projection symbol. S6: When a robot detects a dangerous rear-end collision risk, the cloud scheduling platform assigns a decreasing rear-end collision risk level to the upstream robot based on the distance between the robot's position and other upstream robots, and controls the upstream robot to project a light strip according to the corresponding level of light strip encoding method to form a rearward warning chain.

[0057] In step S4, the forward millimeter-wave radars of the roadside fence robot and the road center fence robot detect the distance and relative speed of vehicles within 150 m in front of the current lane at a frequency of 20 Hz, calculate the absolute speed of the target by combining the speed of the robot itself (provided by Beidou / IMU, with an accuracy of ±0.5 km / h), and then calculate the target deceleration through time difference (Δt = 0.05 s). Report all detected targets and motion states to the cloud scheduling platform; in step S5, the cloud scheduling platform determines the rear-end collision risk level according to the absolute value of the deceleration according to the classification standard in claim 3 (>2 and ≤4 is the attention level, >4 and ≤6 is the warning level, >6 is the danger level), and controls the corresponding robot to project the light band according to the light band coding method of this level; in step S6, when a certain robot detects a danger-level rear-end collision risk (that is, the deceleration of the vehicle in front >6 m / s²), the cloud scheduling platform records the mileage stake number P0 of this robot, traverses the mileage stake numbers Pi of all the robots behind (Pi < P0), calculates the distance d = P0 - Pi, and assigns a decreasing early warning level to the robots behind according to the rule of assigning the danger level when d ≤ 50 m, the warning level when 50 < d ≤ 150 m, and the attention level when 150 < d ≤ 300 m, and controls the robots behind to project the light band according to the corresponding level, forming an early warning chain that propagates backward; the entire rear-end collision early warning process runs in parallel with the dynamic light corridor navigation process without interference.

[0058] During the duration of the dynamic light corridor navigation mode, the rear-end collision early warning process is executed in parallel as a background task; the forward millimeter-wave radars of each robot continuously collect data of the vehicles in front, calculate the target deceleration every 50 ms and compare it with the threshold; when the deceleration exceeds 2 m / s², the robot reports the target information to the cloud; after the cloud scheduling platform determines the risk level according to the deceleration, it immediately sends the light band coding instruction corresponding to this level to the robot, and the robot immediately adjusts the light band projection method to transmit the early warning information to the vehicles behind; at the same time, if the risk level reaches the danger level (>6 m / s²), the cloud platform starts the early warning chain distribution algorithm: calculate the mileage stake number of this robot, screen out all the robots located behind it, and assign the danger level, warning level or attention level respectively according to the distance, and send the light band coding instructions one by one; after receiving the instructions, the robots behind modify their own light band projection methods according to the assigned levels, and the closer the robot is to the danger point, the higher the light band warning intensity (red high-frequency dotted line), and the farther the distance, the lower the light band warning intensity (yellow-orange gradient low frequency); when the danger is lifted (the deceleration of the vehicle in front returns to normal), the cloud platform cancels all early warning instructions, and each robot resumes the normal light corridor mode corresponding to the current visibility.

[0059] This method seamlessly integrates rear-end collision warning functionality into the dynamic optical corridor navigation process, achieving integrated operation of navigation and warning. Through a parallel processing architecture, forward vehicle detection, risk classification, optical band encoding, and warning chain propagation can be completed within milliseconds, ensuring real-time warning. The warning chain mechanism enables the impact of a single hazardous event to propagate more than 300 meters along the road, forming a domino-like relay warning effect, effectively preventing chain-reaction rear-end collisions under low visibility conditions. At the same time, this method is entirely based on the existing optical corridor navigation hardware platform, requiring no additional physical devices, and achieving significant functional enhancements solely through software algorithms, resulting in high cost-effectiveness.

[0060] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of the invention (including the claims) is limited to these examples; within the framework of the invention, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of the different aspects of the invention as described above, which are not provided in the details for the sake of brevity.

[0061] This invention is intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A collaborative inspection system for inspection robots, characterized in that, include: At least one roadside fence robot deployed on the roadside guardrail, and at least one road center fence robot deployed on the median guardrail. The roadside fence robot is equipped with a first laser projection module, which is used to project a first light strip onto the road surface; The road-side fence robot is equipped with a second laser projection module, which is used to project a second light strip onto the road surface; And a cloud-based dispatch platform, which is communicatively connected to both the roadside fence robot and the road center fence robot; The cloud-based scheduling platform is configured to: acquire visibility sensor data of each robot; when visibility is lower than a preset threshold, control the roadside fence robot and the road center fence robot to move along the guardrail; and control the first laser projection module to project a first light strip and the second laser projection module to project a second light strip. The first light strip and the second light strip together constitute a dynamic light corridor to guide vehicle passage.

2. The system according to claim 1, characterized in that, The roadside fence robot and / or the road center fence robot are also equipped with forward millimeter-wave radar to detect the motion status of vehicles ahead in the lane, including distance, relative speed, absolute speed and deceleration. The cloud-based scheduling platform is further configured to: calculate the rear-end collision risk level based on the motion state, and control the first laser projection module and / or the second laser projection module to project a light strip according to the light strip encoding method corresponding to the rear-end collision risk level, wherein the light strip encoding method includes changes in light strip color, flashing frequency, light strip length and / or projection symbol.

3. The system according to claim 2, characterized in that, The rear-end collision risk levels include at least the caution level, the warning level, and the danger level; The optical band encoding method is as follows: Warning level: The light band color transitions from yellow to orange, the flicker frequency is 1Hz, the light band is continuous, and the projection symbol is slow or... ; Warning level: The light strip is orange, the flashing frequency is 3Hz, the light strip is continuous but its length is shortened to 60% of the normal length, and the projected symbol is a brake or... ; Danger level: The light strip is red, the flashing frequency is 6Hz, the light strip is in the shape of a dashed line and its length is shortened to 30% of the normal length, and the real-time distance value is projected.

4. The system according to claim 3, characterized in that, The cloud-based scheduling platform is also configured to: when a robot detects a dangerous rear-end collision risk, assign a decreasing rear-end collision risk level to the upstream robot based on the distance between the robot's position and other upstream robots, and control the upstream robot to project a light strip according to the corresponding level of light strip coding method to form a rearward warning chain.

5. The system according to claim 1, characterized in that, The roadside fence robot is also equipped with a UWB relative positioning module. The cloud scheduling platform is configured to: use the UWB relative positioning module to measure the lateral distance between the roadside fence robot and the adjacent roadside fence robot, and adjust the lateral projection position of the second light strip according to the lateral distance so that the spacing between the second light strip and the first light strip matches the actual lane width.

6. The system according to claim 3, characterized in that, The roadside fence robot and / or the road center fence robot are also equipped with a rearward millimeter-wave radar for detecting the speed and distance of vehicles behind in this lane; The cloud-based dispatch platform is also configured to: calculate the rear-end collision time between the vehicle and the vehicle behind based on the speed and distance of the vehicle behind and the motion state of the vehicle in front; and when the rear-end collision time is lower than a preset threshold, control the first laser projection module and / or the second laser projection module to project light strips according to the light strip coding method of the danger level.

7. The system according to claim 1, characterized in that, The cloud-based dispatch platform is also configured to: when a traffic accident is determined to have occurred based on the fusion of radar and / or camera data reported by each robot, control the roadside fence robot upstream of the accident point to switch the first light strip to a red high-frequency flashing light strip and project the accident text, control the road center fence robot upstream of the accident point to switch the second light strip to a red fork-shaped projection, and broadcast the accident information to the following vehicles via V2X communication.

8. The system according to claim 7, characterized in that, The cloud-based dispatch platform is also configured to: when it receives an approaching emergency vehicle signal, control the roadside fence robot upstream of the accident point to switch the first light strip to an alternating green and white light strip, and control the roadside fence robot to project a right-pointing guide arrow to instruct other vehicles to give way.

9. A collaborative inspection method for inspection robots, applied to the system described in any one of claims 1 to 8, characterized in that, Includes the following steps: S1: The cloud-based dispatch platform obtains visibility data reported by the roadside fence robot and the road center fence robot in real time; S2: When the visibility is lower than a preset threshold, the cloud scheduling platform controls the roadside fence robot and the road center fence robot to move along the guardrail, and controls the first laser projection module on the roadside fence robot to project a first light strip on the road surface and the second laser projection module on the road center fence robot to project a second light strip on the road surface. The first light strip and the second light strip together form a dynamic light corridor. S3: During the duration of the dynamic light corridor, the cloud scheduling platform dynamically adjusts the longitudinal spacing of the robots and the brightness and flicker frequency of the light strip according to the real-time visibility value.

10. The method according to claim 9, characterized in that, It also includes rear-end collision warning steps: S4: The roadside fence robot and / or the road center fence robot detect the motion status of vehicles ahead in their lane using forward millimeter-wave radar, including distance, relative speed, absolute speed and deceleration, and report it to the cloud dispatch platform. S5: The cloud-based scheduling platform calculates the rear-end collision risk level based on the motion state, and controls the first laser projection module and / or the second laser projection module to project a light strip according to the light strip encoding method corresponding to the rear-end collision risk level. The light strip encoding method includes changes in light strip color, flashing frequency, light strip length, and / or projection symbol. S6: When a robot detects a dangerous rear-end collision risk, the cloud scheduling platform assigns a decreasing rear-end collision risk level to the upstream robot based on the distance between the robot's position and other upstream robots, and controls the upstream robot to project a light strip according to the corresponding level of light strip encoding method to form a rearward warning chain.

Images