CN-121521150-B - Multi-sensor data fusion path planning method for pipeline robot

CN121521150BCN 121521150 BCN121521150 BCN 121521150BCN-121521150-B

Abstract

The invention discloses a multi-sensor data fusion path planning method and system for a pipeline robot, which collect laser radar and inertial measurement data through an environment sensing module, a spatiotemporal calibration is performed at the semantic map building module to generate a three-dimensional semantic map containing obstacle properties. Global path planning module adoption And the local dynamic adjustment module establishes a rolling window based on the travelling speed, and utilizes a fast-expansion random tree algorithm to re-plan the dynamic obstacle. And the path execution conversion module carries out smooth optimization on the path through five-degree polynomial fitting and curvature change rate constraint and generates a bottom layer control instruction. According to the invention, through the multisource fusion and layering planning strategy, the problems of positioning drift and dynamic obstacle avoidance of the pipeline environment are effectively solved, and the travelling stability and safety of the robot are remarkably improved.

Inventors

LIU PENGJU
LIU JUNWU
FANG YINGCHUN
ZHAO DAN

Assignees

智铖造(北京)科技有限公司

Dates

Publication Date: 20260512
Application Date: 20251209

Claims (8)

1. The multi-sensor data fusion path planning method for the pipeline robot is characterized by comprising the following steps of: s1, constructing a path planning system of a pipeline robot, wherein the system comprises an environment sensing module, a semantic map building module, a global path planning module, a local dynamic adjustment module and a path execution conversion module; S2, configuring a pipeline environment semantic modeling method of multi-source data fusion based on the semantic map building module, and executing space-time calibration and semantic feature extraction of multi-sensor data; S3, based on the global path planning module and the local dynamic adjustment module, respectively constructing a global path planning algorithm fusing pipeline characteristics and a local adjustment mechanism of dynamic obstacle avoidance, and calculating a travel route conforming to the geometric constraint of the pipeline; s4, based on the path execution conversion module, a path smooth optimization model is established, and planned path coordinates are mapped into a bottom control instruction of the robot; S5, when the system operates, the environment sensing module collects pipeline environment data and transmits the pipeline environment data to the semantic map building module to generate a three-dimensional semantic map, the global path planning module generates an initial path based on the semantic map, the local dynamic adjustment module performs path re-planning when detecting a dynamic obstacle, and the path execution conversion module converts the optimized path into a robot control instruction; in step S3, the global path planning algorithm is configured as a Pipe-a algorithm, and dynamically adjusts heuristic functions for different Pipe segment types: For straight pipeline sections, calculating Euclidean distance as a heuristic function value; For a curved pipeline section, introducing a curvature penalty factor on the basis of Euclidean distance, wherein the curvature penalty factor is determined by a curve weight coefficient and a curve curvature; for a branch pipeline section, introducing a risk avoidance factor on the basis of Euclidean distance, wherein the risk avoidance factor is determined by a branch priority coefficient and a branch risk degree; The Pipe-a algorithm is constructed with a comprehensive cost function consisting of a weighted sum of the travel distance cost, the friction cost, and the steering cost.
2. The method for planning a multi-sensor data fusion path of a pipeline robot according to claim 1, wherein in step S2, the space-time calibration specifically comprises: In the time dimension, synchronizing the laser radar data and the inertial measurement data acquired by the environment sensing module by adopting a nearest neighbor matching method, and setting time synchronization error constraint conditions; and in the space dimension, mapping the laser radar data and the inertial measurement data into a unified coordinate system respectively, fusing the position data by using an extended Kalman filtering algorithm, and setting a consistency constraint condition of space registration.
3. The method for planning a multi-sensor data fusion path of a pipeline robot according to claim 1, wherein in step S2, the semantic feature extraction specifically comprises the steps of constructing a map according to three-level structures of pipeline segments, obstacles and constraint conditions: identifying a pipeline section type, wherein the pipeline section type comprises a straight pipeline, a bent pipeline and a branched pipeline, and the turning angle of the bent pipeline is recorded; extracting the attribute of the obstacle, marking the attribute as a static or dynamic type, and recording the size and position coordinates; meanwhile, the characteristic change of the obstacle is monitored, and when the new obstacle size is detected to exceed a preset threshold value, local map updating is triggered.
4. The method for planning a multi-sensor data fusion path of a pipeline robot according to claim 1, wherein the local adjustment mechanism of the dynamic obstacle avoidance specifically comprises: when a vision sensor in the environment sensing module detects a dynamic obstacle, a rolling window is established by taking the current position of the robot as the center, and the side length of the rolling window is calculated based on the current running speed of the robot, the preset braking time and the bilateral safety redundant distance; and adopting a fast expansion random tree algorithm to search local paths in the rolling window.
5. The method of claim 4, wherein the local adjustment mechanism further comprises smoothing the search result with a third-order Bezier curve to achieve the connection of the global path and the local path, and restricting the connection points to have continuous first and second derivatives and limiting the maximum steering angular velocity.
6. The method for planning a multi-sensor data fusion path of a pipeline robot according to claim 1, wherein in step S4, the path smoothing optimization model uses a five-degree polynomial fit to eliminate path sharp corners and sets a smoothness constraint of a curvature change rate, wherein the curvature change rate is defined as a derivative of a path curvature with respect to a path arc length.
7. The method for planning a multi-sensor data fusion path of a pipeline robot according to claim 1, wherein in step S4, the path execution conversion module is configured with a conversion accuracy verification formula, and the conversion accuracy verification formula is used for calculating a relative error between a theoretical command value calculated by a model and a command value actually issued to an execution mechanism, and verifying whether the relative error is smaller than a preset threshold.
8. A pipeline robot path planning system comprising an environment awareness module, a semantic map building module, a global path planning module, a local dynamic adjustment module, and a path execution transformation module, the system configured to perform the pipeline robot multi-sensor data fusion path planning method of any one of claims 1 to 7.

Description

Multi-sensor data fusion path planning method for pipeline robot Technical Field The invention relates to the technical field of pipeline robots, in particular to a multi-sensor data fusion path planning method for a pipeline robot. Background The pipeline is used as a core infrastructure for energy transportation, municipal drainage and industrial production, and the safe operation and maintenance of the pipeline have important significance for the stable operation of society and economy. With the wide application of pipeline robots in detection, maintenance and other scenes, the path planning capability of the pipeline robots becomes a key for determining the working efficiency and the safety. However, the pipeline environment has the characteristics of sealing, topological complexity (such as multiple branches, curves and reducing sections), dynamic interference (such as temporary obstacles and water flow impact) and the like, and the prior art has obvious limitations. Conventional path planning algorithms such asDijkstra et al apply maturity in a structured environment, but do not consider the mechanical constraints of a pipe bend, conventionally drawing a sharp bend path that is not executable; Although the algorithm can search paths rapidly, path oscillation easily occurs in the long and narrow pipeline, so that planning efficiency is reduced. The intelligent optimization algorithm, such as a genetic algorithm, improves the path optimality through group optimization, but has the advantages of delayed response to dynamic obstacles, long re-planning time consumption, high training cost, weak generalization capability and obvious performance attenuation after changing pipeline parameters although the reinforcement learning method shows adaptability in an unknown environment. Therefore, there is a need for a path planning method for a pipeline robot that combines environmental adaptability, planning efficiency and robustness. Disclosure of Invention The invention provides a multi-sensor data fusion path planning method and system for a pipeline robot, which are used for solving the problems of low path planning precision, slow obstacle avoidance response and unsmooth motion control of the traditional pipeline robot in complex geometric structures and dynamic environments. The invention provides a multi-sensor data fusion path planning method of a pipeline robot, which comprises the steps of constructing a path planning system of the pipeline robot, wherein the system comprises an environment sensing module, a semantic map building module, a global path planning module, a local dynamic adjustment module and a path execution conversion module, configuring a multi-source data fusion pipeline environment semantic modeling method based on the semantic map building module, executing space-time calibration and semantic feature extraction of multi-sensor data, respectively constructing a global path planning algorithm for fusing pipeline characteristics and a dynamic obstacle avoidance local adjustment mechanism based on the global path planning module and the local dynamic adjustment module, calculating a travel path conforming to pipeline geometric constraints, and establishing a path smooth optimization model based on the path execution conversion module, and mapping planned path coordinates to bottom control instructions of the robot. In the space dimension, the laser radar data and the inertia measurement data are respectively mapped into a unified coordinate system, the position data are fused by using an extended Kalman filtering algorithm, and a consistency constraint condition of space registration is set so as to eliminate space errors caused by the difference of the installation positions of the sensors. Based on the calibrated data, semantic features are extracted according to the three-level structure of the pipeline section, the obstacle and the constraint condition, and a map is constructed. The method comprises the steps of identifying pipeline section types, wherein the pipeline section types comprise straight pipelines, bent pipelines and branched pipelines, recording turning angles for the bent pipelines, extracting barrier attributes, marking the barrier attributes as static or dynamic types, recording size and position coordinates, generating traffic constraint, calculating and recording maximum turning angles and minimum curvature radius, monitoring barrier feature changes, and triggering local map updating when the new barrier size is detected to exceed a preset threshold value. In the path planning stage, the global path planning algorithm is configured toThe method comprises the steps of calculating Euclidean distance as a heuristic function value for straight pipeline sections, introducing curvature penalty factors based on the Euclidean distance for curved pipeline sections, wherein the curvature penalty factors are determined by curve weight coefficients and curve curvature, and introd