Search

CN-121982266-A - Vehicle automatic identification and self-adaptive profiling following cleaning system based on three-dimensional contour scanning

CN121982266ACN 121982266 ACN121982266 ACN 121982266ACN-121982266-A

Abstract

The invention belongs to the technical field of vehicle cleaning, in particular to a vehicle automatic identification and self-adaptive profiling following cleaning system based on three-dimensional profile scanning, which comprises an edge-cloud cooperative real-time sensing and modeling subsystem which is deployed in a cleaning channel and runs at an edge computing node and is used for fusing multi-mode data streams, and a vehicle semantic fine digital model attached with material, texture and microscopic profile characteristics is generated on line and updated in real time in an incremental manner by taking a previous frame reconstruction model as a reference.

Inventors

  • CHEN YONG

Assignees

  • 深圳市秒秒物联科技有限公司

Dates

Publication Date
20260505
Application Date
20251225

Claims (10)

  1. 1. The vehicle automatic identification and self-adaptive profiling following cleaning system based on three-dimensional contour scanning is characterized by comprising the following components: an edge-cloud collaborative real-time perception and modeling subsystem deployed in a cleaning channel, comprising: the multi-mode edge perception array is formed by integrating a multi-dimensional laser radar, a High Dynamic Range (HDR) vision sensor and a millimeter wave radar on a cradle head capable of dynamically adjusting the gesture; the incremental real-time three-dimensional reconstruction engine is operated at an edge computing node and used for fusing the multi-mode data stream, and generating and updating a vehicle semantical fine digital model attached with material, texture and microcosmic outline characteristics on line in an incremental manner by taking a previous frame reconstruction model as a reference; A predictive decision and adaptive planning subsystem in signal communication with the perception and modeling subsystem, comprising: The predictive cleaning unit based on physical simulation is internally provided with a multi-physical-field simulation model comprising fluid dynamics, contact mechanics and material abrasion; The meta-action strategy optimizer is used for deconstructing a traditional continuous cleaning path into a series of atomized cleaning meta-actions based on simulation prediction results; An intelligent self-contained collaborative execution and online learning subsystem comprising: The multi-agent collaborative cleaning mechanical arm group comprises more than two seven-degree-of-freedom mechanical arms with force/position mixed control capability, wherein each mechanical arm end effector is respectively integrated with a high-pressure water knife, a flexible contact brush and an air knife drying module, each mechanical arm is used as an independent agent, task allocation and motion coordination are carried out through a distributed consensus algorithm based on the optimal cleaning action sequence, and interference-free synchronous, collaborative or relay operation is realized.
  2. 2. The automatic vehicle identification and self-adaptive profiling following cleaning system based on three-dimensional profile scanning of claim 1, wherein a cradle head in the multi-mode edge perception array is provided with an active tracking mechanism based on vehicle entrance gesture pre-judgment, and the mechanism dynamically adjusts the pitch angle and the azimuth angle of each sensor before the vehicle enters a cleaning channel by fusing low-delay vision prescreening and millimeter wave radar speed measurement information at an entrance so as to ensure that the optimal observation view angle and data acquisition integrity are maintained in the whole vehicle traveling process.
  3. 3. The automatic vehicle identification and self-adaptive profiling follow-up cleaning system based on three-dimensional contour scanning as claimed in claim 1, wherein the incremental real-time three-dimensional reconstruction engine adopts a semantic-guided voxel-nerve radiation field hybrid representation method, combines the geometric priori provided by laser point cloud with the material reflection characteristic extracted by HDR image to realize pixel-level semantic annotation of microscopic pollution states such as vehicle paint scratch, slurry attachment area and glass water stain, and synchronously updates the pixel-level semantic annotation into a digital twin model to serve as a cleaning priority weight basis.
  4. 4. The vehicle automatic identification and self-adaptive profiling following cleaning system based on three-dimensional contour scanning of claim 1, wherein the physical simulation-based predictive cleaning unit is configured with a stain-material coupling response model which dynamically calls corresponding fluid impact threshold values, contact friction coefficients and paint wear sensitivity parameters according to stain types, adhesion thicknesses and substrate material properties marked in the semantic fine digital model, and calculates cleaning efficacy and surface damage risks of each candidate cleaning element action on a local area in real time in multi-physical field simulation; The meta-action strategy optimizer further adopts a constraint satisfaction problem solving framework based on the weighted evaluation result of the cleaning efficiency and the damage risk, generates a minimum redundant meta-action sequence covering the whole vehicle surface on the premise of meeting a preset paint safety threshold, and distributes execution priority, action duration and a terminal tool parameter set for each meta-action.
  5. 5. The vehicle automatic identification and adaptive profiling follow-up cleaning system based on three-dimensional profile scanning as set forth in claim 1, wherein each of the plurality of robot arms in the multi-agent collaborative cleaning robot arm set is configured with a self-aware-execute closed loop unit comprising: the miniature six-dimensional force/moment sensor and the high-frame-rate near-field vision module are integrated in the end effector and are used for capturing the distribution of the contact force between the tool and the vehicle body and the change of the local surface state in real time in the cleaning process; And the self-adaptive compliant controller based on impedance-admittance hybrid control dynamically adjusts the rigidity and damping parameters of the tail end of the mechanical arm under a task coordinate system according to the contact force feedback so as to maintain a preset acting force track under the impact disturbance of high-pressure water jet or the deformation of the flexible brush.
  6. 6. The vehicle automatic identification and adaptive profiling follow-up cleaning system based on three-dimensional profile scanning as set forth in claim 1, further comprising a safety awareness and emergency intervention subsystem integrated between the edge computing node and the robotic arm controller configured with: The multi-level collision early warning module is used for carrying out forward collision detection based on a real-time digital twin model and a mechanical arm kinematic envelope curve, and triggering a deceleration or suspension instruction within 200ms before potential interference occurs by combining the penetration sensing capability of the millimeter wave radar on a non-rigid barrier; and (3) a paint damage emergency response mechanism, when the self-perception-execution closed loop unit detects local contact force mutation or abnormal surface reflectivity reduction, immediately stopping the current element action and starting high-precision local rescanning, and if the paint damage risk is confirmed to exceed a threshold value, switching to a low-impact air knife drying mode to finish the residual area treatment.
  7. 7. The vehicle automatic identification and self-adaptive profiling following cleaning system based on three-dimensional contour scanning of claim 1, wherein the predictive decision and self-adaptive planning subsystem is further coupled with an environmental disturbance compensation model, the model collects temperature and humidity, wind speed and water pressure fluctuation data in a cleaning channel in real time, and inputs the data as boundary conditions into a multi-physical-field simulation engine, dynamically corrects fluid jet diffusion angles, brush friction coefficients and drying efficiency parameters, and ensures the robustness of cleaning strategies under different climates and water supply conditions; The meta-action strategy optimizer synchronously outputs an environment sensitivity label when generating an action sequence, and is used for guiding the mechanical arm to preferentially select non-contact cleaning meta-actions with strong anti-interference performance under a high-disturbance scene.
  8. 8. The automatic vehicle identification and adaptive profiling follow-up cleaning system based on three-dimensional contour scanning as claimed in claim 1, wherein the automatic vehicle identification and adaptive profiling follow-up cleaning system based on three-dimensional contour scanning supports a man-machine hybrid enhanced cleaning mode, The AR interactive interface also displays a predicted cleaning effect thermodynamic diagram and a paint safety margin indication in a real-time superposition mode, and assists an operator in making an intervention decision.
  9. 9. The automatic vehicle identification and self-adaptive profiling following cleaning system based on three-dimensional profile scanning of claim 1, wherein the intelligent cooperative execution and online learning subsystem is provided with an energy efficiency perception scheduler, and the scheduler is used for carrying out time elastic scheduling on non-emergency cleaning tasks based on an energy consumption model of each cleaning element action and combining a power grid peak-valley electricity price signal and a water resource availability index; In a multi-vehicle continuous operation scene, the scheduler performs space-time staggered arrangement on the cleaning action sequences of adjacent vehicles through an edge side task queue optimization algorithm so as to reduce peak power requirements and improve equipment utilization rate.
  10. 10. The vehicle automatic identification and self-adaptive profiling following cleaning system based on three-dimensional contour scanning of claim 1, wherein the edge-cloud collaborative real-time perception and modeling subsystem supports a cross-site model federal learning architecture, a plurality of cleaning sites deployed at different geographic positions update to a cloud aggregation server through encryption and uploading gradients of a local incremental reconstruction model, and the cloud generates a globally optimized semantic segmentation and stain identification model on the premise of protecting data privacy and periodically transmits the model to each edge node; meanwhile, the digital twin model adopts light-weight containerization packaging, supports one-key migration to a novel vehicle or a temporary mobile cleaning platform, and realizes quick replication and generalized deployment of system capacity.

Description

Vehicle automatic identification and self-adaptive profiling following cleaning system based on three-dimensional contour scanning Technical Field The invention belongs to the technical field of vehicle cleaning, and particularly relates to a vehicle automatic identification and self-adaptive profiling following cleaning system based on three-dimensional profile scanning. Background The vehicle cleaning equipment is a special device for automatically or semi-automatically cleaning attachments such as dirt, dust, oil stains and the like on the surface of the vehicle, and is widely applied to scenes such as automobile service stations, intelligent vehicle washing centers, transportation hubs and the like. The vehicle cleaning device achieves primary cleaning through a high-pressure spray system. The high-pressure water pump pressurizes the water flow of the mixed cleaning agent, then forms dense water columns through the nozzles, and peels off sediment and loose dirt on the surface of the vehicle body with strong impact force, the pre-washing link usually comprises foam spraying, so that oily dirt is fully emulsified and decomposed, and the core cleaning depends on physical contact brushing or non-contact high-pressure fine washing. The contact type device uses a rotary hairbrush group (top, side and rolling brush) to match with cleaning liquid to wrap the vehicle body for brushing, and uses a multi-angle high-pressure fan-shaped water knife to flush gaps of the vehicle body accurately in a non-contact mode, and the two modes can use clear water for secondary flushing to thoroughly remove residual cleaning agent. In the prior art, although various vehicle cleaning devices exist in the current market, the vehicle cleaning devices still have significant defects in aspects of intelligence, adaptability, safety and the like, and the vehicle cleaning devices can be summarized into the following three points: firstly, the main stream equipment at present adopts the structural design of fixed volume and preset spray washing path, is only suitable for standard-size cars, and cannot be effectively adapted to the models of commercial vehicles, medium-large SUVs and the like with larger vehicle body size and complex outline, so that the cleaning blind areas are more and the coverage rate is low; the existing equipment generally depends on a fixed pressure nozzle or a rotary brush to perform rough flushing, lacks the sensing and response capability of stain type, adhesion strength and material characteristics on the surface of a vehicle body, is difficult to realize deep and differential cleaning, and has poor cleaning effect on areas such as stubborn stains, corner gaps and the like; Thirdly, the energy consumption is high, the potential safety hazard is large, and the environmental adaptability is weak. Most automatic or self-service car washing equipment depends on single-phase 220V or three-phase 380V strong electric drive, a high-power water supply and air supply system is heavy in size, high in operation noise and high in electricity utilization safety risks such as electric leakage and short circuit, meanwhile, the equipment generally lacks a water and energy saving mechanism, water resources and energy waste are serious, the equipment does not accord with the development trend of green low carbon, is limited by infrastructure conditions and is difficult to flexibly deploy in places with limited space or insufficient power configuration, and therefore, the automatic vehicle identification and self-adaptive profiling following cleaning system based on three-dimensional profile scanning is provided for the problems. Disclosure of Invention In order to overcome the defects of the prior art and solve at least one technical problem in the background art, the invention provides a vehicle automatic identification and self-adaptive profiling following cleaning system based on three-dimensional profile scanning. The technical scheme adopted by the invention for solving the technical problems is that the vehicle automatic identification and self-adaptive profiling follow-up cleaning system based on three-dimensional profile scanning comprises: an edge-cloud collaborative real-time perception and modeling subsystem deployed in a cleaning channel, comprising: the multi-mode edge perception array is formed by integrating a multi-dimensional laser radar, a High Dynamic Range (HDR) vision sensor and a millimeter wave radar on a cradle head capable of dynamically adjusting the gesture; the incremental real-time three-dimensional reconstruction engine is operated at an edge computing node and used for fusing the multi-mode data stream, and generating and updating a vehicle semantical fine digital model attached with material, texture and microcosmic outline characteristics on line in an incremental manner by taking a previous frame reconstruction model as a reference; A predictive decision and adaptive planning subsystem in signal