CN-121977581-A - Path planning and gait optimizing method for complex terrain inspection robot

CN121977581ACN 121977581 ACN121977581 ACN 121977581ACN-121977581-A

Abstract

The invention relates to the technical field of robots and discloses a path planning and gait optimization method of a complex terrain inspection robot, wherein the cross-scale parallel perception of a macroscopic grid and microscopic foot drop points is realized for the first time through multi-scale terrain perception and uncertainty quantification, the fundamental contradiction of macroscopic trafficability and microscopic unreachable is fundamentally solved, and the microscopic foot drop point reachability and the terrain evolution law are integrated into path-gait joint optimization through constructing a multi-constraint collaborative planning and foot-ground interaction time-varying model, so that the full-period conflict of walking dynamic stability and operation static stability is fundamentally solved. Aiming at the coupling resonance problem of the contact operation of the foot robot, a cross-domain coupling dynamics model is established, edge-end millisecond real-time solving is realized by adopting a model order reduction and lightweight network, resonance is actively restrained, the operation precision is ensured to be +/-1N/0.5 mm, and a full-link uncertainty nonlinear conduction and risk grading closed-loop correction mechanism is introduced, so that the self-learning and self-healing of the system are realized.

Inventors

GAO BO
LI PEISONG
WANG WEI
YUAN SHOUBIN
ZHANG XIAOHUA
XU SHENGCHEN
KONG YIHUI
WANG YUXIN
LI YUNLONG
WANG MINZHEN

Assignees

国网吉林省电力有限公司延边供电公司
浙江钰伟智能科技有限公司

Dates

Publication Date: 20260505
Application Date: 20260407

Claims (10)

1. The path planning and gait optimizing method for the complex terrain inspection robot is characterized by comprising the following steps of: S1, initializing a two-dimensional sensing scale, and acquiring multi-scale topographic data of a robot operation area, wherein the multi-scale topographic data comprises a macroscopic grid map and a microscopic topographic elevation model; s2, constructing a multi-constraint collaborative planning objective function fused with microscopic foot drop accessibility constraint, terrain evolution constraint, robot kinematics constraint and inspection task constraint based on the multi-scale terrain data, and generating a global path containing a foot drop sequence; S3, based on the microscopic terrain elevation model and the foot drop point sequence, establishing a foot-to-ground interaction time-varying characteristic model, quantifying the transformation effect of single-step treading on the terrain and the accumulated deformation rule of multi-step treading, and respectively optimizing gait parameters for a walking stage and an operation stage; S4, constructing a cross-domain coupling dynamics model integrating robot body dynamics, foot-end contact mechanics and mechanical arm operation dynamics, performing reduced-order processing on the model, building an edge-end real-time solving frame through a lightweight convolutional neural network, and generating a cooperative impedance control instruction to inhibit coupling resonance in the walking and operation process; S5, based on feedback data in the operation executing process, constructing a full-link uncertainty closed-loop correction and linkage failure pre-judging mechanism, and iteratively optimizing the global path, gait parameters and control instructions to generate a motion operation integrated executing scheme; And S6, a visual defect detection module and an operation mechanical arm control unit are linked, the operation parameters of the gait optimization and dynamics control module are dynamically adjusted according to the operation precision requirement, and the unmanned operation integrating tower foundation defect identification and ground resistance detection is completed.
2. The method for planning and optimizing the gait of the complex terrain inspection robot path according to claim 1, wherein the specific configuration of the two-dimensional perception scale in step S1 is as follows: the macro grid scale is set to 0.1 to 0.5m, matching the global path planning decision granularity; The microscopic foot drop scale is set to be 1cm multiplied by 1cm grade resolution, and the foot drop planning decision granularity at the foot end is matched; Feature extraction dimensions of semantic labels, gradients, elevations and obstacle distribution are matched for macroscopic grid dimensions, and uncertainty quantization rules of semantic segmentation confidence and topography reconstruction errors are respectively and correspondingly set for feature extraction dimensions of microscopic foot drop dimensions, such as surface roughness, contact stiffness and elevation fluctuation.
3. The method for path planning and gait optimization of the complex terrain inspection robot according to claim 2, wherein the specific process of constructing the macro grid map and reconstructing the micro terrain in the step S1 is as follows: Determining a map coverage area by taking the current position of the robot as a starting point and a tower foundation target point to be inspected as an ending point, dividing the coverage area into a plurality of standard grid units, calculating grid elevation and gradient parameters to finish preliminary trafficability classification, and generating a macroscopic grid map by matching semantic tags with classification confidence levels for grids through a lightweight semantic segmentation network; Based on a global path initial planning result, extracting a path along a grid as a reconstruction target area, completing three-dimensional implicit reconstruction of the target area through a lightweight neural radiation field technology, generating a microscopic topography elevation model, extracting foot-to-ground interaction key features of each microscopic subdivision unit, matching foot drop safety evaluation indexes, counting the occupation ratio of safety foot drop points in a macroscopic grid, generating a microscopic reachability evaluation result, and synchronizing the microscopic reachability evaluation result to a macroscopic grid map.
4. The method for path planning and gait optimization of the complex terrain inspection robot according to claim 1, wherein the core constraints of the multi-constraint collaborative planning objective function in step S2 are specifically: the microscopic foot drop accessibility constraint is used for ensuring that the continuous safe foot drop occupancy rate of the planned path along the line meets a preset threshold; The full-period terrain evolution constraint is used for pre-judging foot-ground interaction terrain transformation effect and avoiding the easily deformed area as a tower foundation operation parking point; The robot kinematic limit constraint is used for matching the physical limit of the maximum passing gradient, the obstacle crossing height and the minimum turning radius of the robot; and the inspection task constraint is used for ensuring that the planned path fully covers all the tower foundation operation points to be inspected.
5. The method for path planning and gait optimization of the complex terrain inspection robot according to claim 4, wherein the specific process of global path initial planning and reachability verification in step S2 is as follows: Generating an initial global path covering all inspection target points based on an improved RRT algorithm and a macroscopic grid map, dividing the initial path into a plurality of continuous path segments, and matching microscopic topographic data with microscopic reachability evaluation results; Calculating the duty ratio of the continuous safe footdrop number aiming at each path segment, judging that the path segment is effective and generating an alternative footdrop sequence when the duty ratio is higher than a preset accessibility threshold, and triggering path iterative correction when the duty ratio is lower than the preset threshold until the corrected path segment meets the microscopic footdrop accessibility constraint; combining the operation space of the mechanical arm, the visual detection field and the stability constraint of the landed point topography, solving the optimal operation landed pose of the robot, and generating an operation point location accurate planning path.
6. The method for path planning and gait optimization of the complex terrain inspection robot according to claim 1, wherein the specific process of the foot-to-ground interaction time-varying characteristic model and gait optimization in the step S3 is as follows: Based on a microscopic terrain elevation model, a foot drop point sequence and robot body parameters, quantifying the mapping relation of foot contact force, contact angle and terrain deformation and contact stiffness change under single-step pedaling, establishing a terrain accumulation deformation evolution rule under multi-step continuous pedaling, and pre-judging the terrain state change trend under asynchronous state parameters; Aiming at the walking stage, taking dynamic passing stability, obstacle crossing capability and energy consumption efficiency as optimization targets, matching corresponding step length, step height and supporting phase duty ratio core gait parameters for different terrain scenes in a mountain area, and completing terrain self-adaptive real-time gait optimization; Aiming at the working stage, the static stability of the machine body and the consistency of the rigidity of the foot end ground are taken as optimization targets, working gait parameters are matched, the distribution positions of the front gait sequence and the foot end are optimized, and the working parking gait pre-optimization is completed.
7. The method for planning and optimizing gait of the complex terrain inspection robot path according to claim 1, wherein the specific construction process of the cross-domain coupling dynamics model and the edge-end real-time solving framework in step S4 is as follows: Based on the structural parameters of the robot body, the foot contact characteristics and the mechanical arm kinematics parameters, constructing a cross-domain coupling dynamics model integrating the body floating base multi-body dynamics, the foot contact dynamics and the mechanical arm operation rigid body dynamics, quantifying the two-way coupling mapping relation of the mechanical arm operation reaction force, the foot contact rigidity change and the body posture disturbance, and revealing the generation mechanism and the triggering condition of the coupling resonance; performing reduced order processing on the high-dimensional coupling dynamics model, reserving the coupling resonance related core dynamics characteristics, and eliminating redundant high-dimensional parameters; and constructing a lightweight convolutional neural network, constructing an edge real-time solving framework based on the coupling resonance suppression control laws under different working conditions of simulation and real-time working condition data pre-training, and realizing the output of the optimal control quantity within 10 ms.
8. The method for path planning and gait optimization of the complex terrain inspection robot according to claim 7, wherein the specific execution of the cooperative impedance control instruction in step S4 is as follows: When the robot executes walking action, gait tracking precision and body gesture stability are used as control targets, the impedance parameters of all joints and foot contact force are dynamically adjusted, and body fluctuation caused by terrain disturbance is restrained; when the robot executes contact type operation, an operation action instruction of the mechanical arm is synchronously received, the influence of operation disturbance on the robot body is prejudged in advance, the gait support phase, the foot grounding rigidity and the joint impedance are synchronously adjusted, and the passive fluctuation of the robot body caused by operation reaction force is counteracted; The method comprises the steps of monitoring the fluctuation of the body gesture, the change of the foot ground rigidity and the contact force error of the tail end of the mechanical arm in real time, and when the coupling resonance trend is detected, adjusting the joint impedance, changing the supporting phase, finely adjusting the operation track of the mechanical arm to change the natural frequency of the system, interrupting the positive feedback circulation of resonance, and realizing millisecond-level real-time inhibition of the coupling resonance.
9. The method for planning and optimizing the gait of the complex terrain inspection robot path according to claim 1, wherein the specific process of the full-link uncertainty closed-loop correction and the chain failure pre-judgment in the step S5 is as follows: The method comprises the steps of collecting state data and error information of each module in real time, constructing a full-link uncertainty nonlinear conduction model, quantifying the conduction rule and amplification effect of each link uncertainty, dividing the linkage failure risk into three grades, namely low, medium and high, maintaining the current operation parameters at the low risk grade, triggering the reverse closed loop correction of a cross-module at the medium risk grade, suspending the current task at the high risk grade, switching the safety gait and triggering the global path re-planning; The cross-module reverse closed loop correction comprises the steps of reversely correcting local path parameters when gait tracking errors exceed standards, reversely improving regional perception resolution and correcting trafficability constraint when perception uncertainty exceeds standards, and reversely optimizing gait parameters when fuselage stability fluctuation exceeds standards; And dynamically updating the threshold parameters, the weight coefficients and the uncertainty conduction model of each module based on the historical operation data and the error correction result.
10. The method for planning and optimizing gait of the complex terrain inspection robot path according to claim 1, wherein the specific process of collaborative control of the tower foundation inspection operation in step S6 is as follows: After the robot reaches a tower foundation operation stop point, according to the operation precision requirement of ground resistance detection, the linked gait optimization and dynamics control module adjusts the posture of the robot body, the foot end support distribution and the joint impedance parameters, so that the robot body enters an operation high-stability mode; the method comprises the steps of collecting tower footing images through a binocular vision camera, identifying the positions of tower footing defects and grounding electrodes, solving the optimal operation pose of a mechanical arm, and dynamically fine-adjusting the position of the robot body to realize no dead angle coverage of a detection area; planning a flexible operation track of the mechanical arm detection probe, synchronizing an operation action instruction to a dynamics control module in advance to counteract operation disturbance, and completing automatic detection of the grounding resistor; and detecting the wireless signal intensity of the operation area in real time, adaptively selecting a communication mode to transmit back detection data and early warning information, and controlling the robot to recover a normal walking mode after the operation is finished, and going to the next inspection target point according to a planned path.

Description

Path planning and gait optimizing method for complex terrain inspection robot Technical Field The invention relates to the technical field of robots, in particular to a path planning and gait optimization method of a complex terrain inspection robot. Background With the rapid development of industrial intelligence and robot technology, the inspection robot gradually replaces manpower, is widely applied to various scenes such as power grids, underground mines, oil and gas pipelines, chemical industry parks, mountain photovoltaics, water conservancy embankments, geological disaster rescue, edge protection patrol and the like, the scenes generally have unstructured complex features such as rugged topography, large gradient fluctuation, disordered obstacle distribution, limited space and the like, part of scenes are accompanied with severe environments such as high temperature, high humidity, high dust, inflammable and explosive, poisonous and harmful environments and the like, the traditional manual inspection mode has various pain points such as low operation efficiency, high safety risk, incomplete inspection coverage, poor data traceability, incapacity of operation in extreme environments and the like, and the autonomous inspection robot with all-terrain traffic capacity has become an industry requirement. The path planning and gait optimization are two core technologies for realizing autonomous operation of the complex terrain inspection robot, the former is responsible for generating a global passable path meeting the requirements of inspection tasks and a local obstacle avoidance track, the latter is responsible for matching terrain features and path constraints, the motion gait of the robot is dynamically adjusted, the passing stability, the operation efficiency and the cruising ability of the robot under complex terrain are ensured, and the cooperative adaptation ability of the two directly determines the operation boundary and the reliability of the inspection robot in an unstructured environment. For the above and related art, there are often the following drawbacks: 1. The conventional path planning and gait optimization technology of the complicated mountain foot-type inspection robot lacks systematic cognition on the multi-scale heterogeneity of mountain terrain and the bidirectional time-varying characteristics of foot-to-ground interaction, a technical architecture of unidirectional serial open loop is generally adopted, and bidirectional coordination of each link of path planning, gait optimization and execution control cannot be realized, so that full-link linkage failure caused by cross-scale decision mismatch, full-period target conflict of walking and operation and uncertainty nonlinear amplification of macro planning and micro execution is caused; 2. in the prior art, deep research on a multi-body dynamics cross-domain coupling mechanism of a foot-type floating-base robot is lacking, a control strategy of walking and operation decoupling is generally adopted, sudden changes of system dynamics characteristics caused by terrain time variation cannot be dealt with, meanwhile, the embedded edge computing force constraint and the real-time solving requirement of a complex dynamics model cannot be considered, the problems that coupling resonance is out of control easily and operation precision cannot be guaranteed in the contact operation process of the robot are caused, and synchronous optimization of resonance suppression and operation precision cannot be realized in the prior art. Disclosure of Invention The invention aims to solve the technical problem that the prior art has the defect that the mutation of the dynamic characteristics of the system caused by the time variation of the terrain cannot be dealt with, and therefore, a path planning and gait optimization method of a complex terrain inspection robot is provided. In order to achieve the purpose, the application adopts the following technical scheme that the path planning and gait optimizing method of the complex terrain inspection robot comprises the following steps: s1, acquiring multi-scale topographic data of a robot operation area, wherein the multi-scale topographic data comprises a macroscopic grid map and a microscopic topographic elevation model; s2, constructing a multi-constraint collaborative planning objective function integrating microscopic footdrop reachability constraint, full-period terrain evolution constraint, robot kinematics constraint and inspection task constraint based on multi-scale terrain data, and generating a global path containing a footdrop sequence; s3, based on a microscopic terrain elevation model and a foot drop point sequence, establishing a foot-to-ground interaction time-varying characteristic model, quantifying the transformation effect of single-step treading on the terrain and the accumulated deformation rule of multi-step treading, and respectively optimizing gait parameters for a walking sta