Robot dog path planning method and system based on intelligent inspection

Abstract

The invention discloses a robot dog path planning method and system based on intelligent inspection, and relates to the technical field of robot autonomous navigation. The method comprises the following steps: acquiring temperature and distance data of a target area, a robot dog envelope deformation phase and an impact excitation frequency; calculating a coordinate offset correction cost map through the offset calibration matrix to obtain an environmental geometric boundary; when impact and sampling frequency interfere, a sampling time sequence is adjusted to avoid a displacement peak phase, and a de-aliasing point cloud is output; recognizing a narrow point area based on the point cloud and the boundary, calculating a first phase window when the dynamic envelope width is narrowest, and generating a motion rhythm instruction according to the first phase window to enable the robot dog to pass through the narrow point. The sensing interference of thermal field refraction and vibration aliasing is suppressed, the problem of space-time mismatch under a limit narrow slit is solved, and the routing inspection reliability is improved.

Description

Technical Field

[0001] This invention relates to the field of robot autonomous navigation technology, and in particular to a path planning method and system for a robot dog based on intelligent inspection. Background Technology

[0002] In the inspection processes of quadruped robots in complex industrial environments such as power and chemical industries, the navigation system relies on sensing sensors such as infrared and depth sensors to acquire spatial characteristics of the environment. Existing path planning technologies mostly utilize static geometric models to identify and avoid obstacles. However, when facing extreme conditions such as high temperatures, high vibrations, and extremely narrow gaps, the inspection robot system exhibits the following failure mechanisms: Refraction and drift mechanism at the spatial environment level: The thermal plume generated on the surface of the target in industrial inspection causes non-uniform fluctuations in the refractive index of the surrounding air medium. When the sensing beam traverses this unstable thermal environment, the beam propagation path undergoes a physically nonlinear deflection. Existing map-building algorithms, when processing sensing signals with this deflection phenomenon, generate spatial geometric descriptions that deviate significantly from the actual working conditions. This spatial coordinate drift causes a displacement mismatch between the obstacle avoidance vector generated by the navigation system and the actual obstacle position, making it difficult to maintain stable inspection accuracy of the robot dog in complex thermal environments at the engineering implementation level.

[0003] Signal aliasing mechanism at the perception level: There is a potential interference between the mechanical impact excitation generated during the robot dog's walking and the sensor sampling rhythm. When the impact frequency generated by the foot touching the ground overlaps with the system sampling frequency in the time domain, the environmental feature point cloud will exhibit significant contour blurring and position aliasing. Existing filtering techniques struggle to effectively separate noise features from effective perception features during engineering operation when faced with non-stationary signals caused by mechanical vibration. This feature aliasing at the perception signal level leads to semantic ambiguity in the physical input obtained by the path planning process, increases the randomness of obstacle avoidance logic triggering, and prevents the system from forming a reproducible control loop.

[0004] Spatiotemporal mismatch mechanism at the dynamic passage level: In extreme narrow-slit passage scenarios in industrial settings, a complex dynamic interference relationship exists between the periodic deformation of the robot dog's body width and the constraints of fixed obstacles. Existing control architectures typically abstract the mobile platform as a constant geometry, failing to fully consider the dynamic evolution of the spatial envelope caused by gait rhythms. Under extreme constraints, the spatiotemporal coupling relationship between the body motion phase and the passage space is prone to instability. When faced with this dynamic interference, existing planning algorithms often fail to achieve precise coordination between motion phase and spatial gap at the execution level, leading to mechanical collisions or passage failures when the robot dog passes through key narrow areas.

[0005] The three types of failure mechanisms mentioned above are coupled with each other, making it difficult for existing path planning schemes to form a deterministic obstacle avoidance control closed loop under complex industrial inspection conditions. Summary of the Invention

[0006] This invention provides a path planning method and system for a robot dog based on intelligent inspection, in order to solve the technical problem that in inspection tasks in complex environments, the robot dog's lateral dimension periodic change pattern during gait cycles and the spatial geometric constraints of extreme narrow points are mismatched in time and space. Furthermore, the nonlinear spatial coordinate shift caused by non-uniform thermal field refraction and the temporal aliasing of perception sampling caused by mechanical impact excitation further aggravate the physical quantification error of the planning system on the passable width, ultimately leading to misjudgment of extreme obstacle avoidance points and passage failure.

[0007] In view of the above problems, the present invention provides a path planning method for a robot dog based on intelligent inspection, the method comprising the following steps: Step S1: Obtain surface temperature distribution data and surface geometric distance data of the target area, and obtain the envelope deformation phase data, impact excitation frequency data and preset static reference width of the robot dog; Step S2: Calculate the first spatial coordinate offset using a preset offset calibration matrix, and correct the occupancy probability threshold in the local cost map based on the first spatial coordinate offset to obtain the first environmental geometric boundary; Step S3: Determine the interference weight between the impact excitation frequency data and the data acquisition frequency, and when the interference weight exceeds the first preset threshold, adjust the sampling triggering timing of the data acquisition frequency so that the sampling triggering timing avoids the peak phase of the physical displacement generated by the robot dog's movement, and output the de-aliased local point cloud. Step S4: Based on the dealiasing local point cloud, identify narrow point regions in the first environmental geometric boundary where the physical space width is within a first preset width range, and calculate the first phase window corresponding to when the dynamic envelope width of the robot dog reaches the minimum value according to the envelope deformation phase data; Step S5: Generate motion rhythm commands according to the first phase window, so that the longitudinal axis of the robot dog's body passes through the narrow point region within the first phase window.

[0008] Furthermore, step S2 specifically includes: A mapping relationship is established between the first spatial coordinate offset and the temperature difference, and the original distance extracted from the surface geometric distance data; wherein, the first spatial coordinate offset increases monotonically with the increase of the temperature difference and the increase of the original distance, and the first spatial coordinate offset tends to a preset offset limit constant after the temperature difference exceeds a preset temperature threshold.

[0009] Furthermore, the first spatial coordinate offset The calculation formula is as follows: in, This is the first spatial coordinate offset. The offset limiting constant is... The temperature difference in the surface temperature distribution data. The preset reference calibration temperature, The original distance, The preset reference distance constant, and The preset weighting factor is used; the offset calibration matrix contains the preset weighting factor. and .

[0010] Furthermore, the output value of the first spatial coordinate offset is limited to a non-negative range.

[0011] Furthermore, step S3 specifically includes: When the ratio of the impact excitation frequency data to the data acquisition frequency falls into a preset frequency domain aliasing interval, the real-time phase of the robot dog's motion is extracted from the envelope deformation phase data. The sampling trigger timing is adjusted according to the real-time phase so that the deviation between the edge of the trigger pulse of the sampling trigger timing and the phase of the peak phase of the physical displacement is greater than a preset deviation threshold.

[0012] Furthermore, step S4 specifically includes: Based on the envelope deformation phase data, the dynamic envelope width of the robot dog during the motion cycle is determined; A difference function is established between the dynamic envelope width and the geometric width of the narrow point region. Based on the difference function and a preset monotonically increasing judgment function, the passability confidence of the robot dog in the current phase is determined. The passability confidence shows an increasing trend in the critical region where the difference turns from negative to positive.

[0013] Furthermore, the passage confidence is calculated by using the phase passage decision index. The formula is as follows: in, This is the phase passage decision index. The envelope deformation phase data, The geometric width of the narrow point region. The static reference width, The dynamic compression rate is preset based on the physical structure of the robot dog. This is a preset spatial compensation term with the same dimensions as the geometric width. This is the preset gain factor used for dimensional normalization.

[0014] Furthermore, the value of the dynamic compression ratio is limited to a preset range of physical contraction limits of the body, and the first phase window is a continuous phase range where the phase passage decision index is greater than the second preset threshold.

[0015] Furthermore, the upper limit of the first preset width range is 1.2 times the static reference width.

[0016] This invention provides a path planning system for a robot dog based on intelligent inspection, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the aforementioned path planning method for a robot dog based on intelligent inspection.

[0017] The technical solution provided in this application has at least the following technical effects: By quantifying the nonlinear saturation function relationship between surface temperature difference data, original distance and spatial coordinate offset, the grid coordinates in the local cost map that are disturbed by thermal field refraction are compensated for physical displacement. This reduces the image artifacts and contour overlap of the perception boundary under high temperature conditions, ensuring that the robot dog can obtain a geometrically consistent environmental reference benchmark in the high temperature inspection path.

[0018] By establishing a phase mapping between the impact excitation frequency data and the timing of the sensor sampling trigger pulse edge, the data acquisition action can avoid the peak and trough intervals of the robot dog's body displacement. Thus, without relying on the performance of the software filter, false feature point clusters caused by mechanical oscillation are filtered out from the sensing source, achieving time-domain gain stability of dynamic environmental point clouds under strong vibration conditions.

[0019] Without altering the robot dog's main mechanical structure, by matching the evolution phase of the robot's lateral envelope width with the geometric constraints of the narrow point region, the system can ensure precise alignment on the time axis between the moment the robot dog's body is at its physical minimum width phase and the displacement process through the narrow point region by adjusting motion control variables such as step frequency, stride length, and foot dwell time. This allows the robot dog to complete non-contact continuous passage without human intervention even in extreme conditions where the obstacle gap is close to or smaller than the robot dog's static reference width.

[0020] The synergistic effect of the above technologies solves the problem of perception error accumulation and execution collision caused by environmental thermal field shift, mechanical vibration interference and spatial phase mismatch in inspection robots. This enables the path planning system to form a deterministic and reproducible narrow-gap passage control closed loop in industrial field environments with complex multi-physics field interference, effectively improving the success rate of robot dogs in high-temperature, high-vibration and extremely narrow spaces. Attached Figure Description

[0021] Figure 1 This is a flowchart of a robot dog path planning method based on intelligent inspection in an embodiment of the present invention; Figure 2 This is a system architecture diagram of a robot dog path planning system based on intelligent inspection, as described in an embodiment of the present invention. Detailed Implementation

[0022] The above technical solutions will now be described in detail with reference to the accompanying drawings and specific embodiments to provide a better understanding of them. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. It should be understood that the present invention is not limited to the exemplary embodiments used only to explain the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention. Furthermore, it should be noted that, for ease of description, only the parts related to the present invention are shown in the drawings, not all of them.

[0023] Example See Figure 1 and Figure 2 This invention provides a path planning method for a robotic dog based on intelligent inspection. This method is executed by a path planning system for a robotic dog based on intelligent inspection, specifically by a processor calling and executing a computer program stored in memory. The method includes the following steps: Step S1: Obtain surface temperature distribution data and surface geometric distance data of the target area, and obtain the envelope deformation phase data, impact excitation frequency data and preset static reference width of the robot dog; Step S2: Calculate the first spatial coordinate offset using a preset offset calibration matrix, and correct the occupancy probability threshold in the local cost map based on the first spatial coordinate offset to obtain the first environmental geometric boundary; Step S3: Determine the interference weight between the impact excitation frequency data and the data acquisition frequency, and when the interference weight exceeds the first preset threshold, adjust the sampling triggering timing of the data acquisition frequency so that the sampling triggering timing avoids the peak phase of the physical displacement generated by the robot dog's movement, and output the de-aliased local point cloud. Step S4: Based on the dealiasing local point cloud, identify narrow point regions in the first environmental geometric boundary where the physical space width is within a first preset width range, and calculate the first phase window corresponding to when the dynamic envelope width of the robot dog reaches the minimum value according to the envelope deformation phase data; Step S5: Generate motion rhythm commands according to the first phase window, so that the longitudinal axis of the robot dog's body passes through the narrow point region within the first phase window.

[0024] The underlying hardware architecture of the intelligent inspection-based robot dog path planning system is built upon a collaborative mechanism between the processor and memory. The memory contains a computer program that executes the path planning algorithm. A physical data channel is established between the processor and memory via a standard bus interconnecting peripheral components within the motherboard. During system power-on initialization, the computer program is loaded from memory into the processor's cache for instruction-level scheduling and execution. The processor then executes the computer program to implement the robot dog's path planning logic.

[0025] The robot dog has an infrared thermal imager, a depth ranging module, a high-precision joint encoder, and a six-axis inertial measurement unit mounted at specific physical locations on its body. The infrared thermal imager and depth ranging module are rigidly fixed to the gimbal area at the front of the robot dog's body, performing a non-contact scanning task that overlaps the field of view of the target area in front. The high-precision joint encoder is integrated into the internal structure of the drive joints of the robot dog's four legs. The six-axis inertial measurement unit is patch-mounted at the center of gravity of the robot dog's body. The infrared thermal imager, depth ranging module, high-precision joint encoder, and six-axis inertial measurement unit all establish a data interconnection channel with the processor through the controller area network bus.

[0026] After the hardware architecture is initialized, the processor initiates a synchronous sensing process based on multi-source heterogeneous features. Through a low-level hardware timer, the processor sends a global synchronization trigger pulse to the infrared thermal imager and the depth ranging module. Upon receiving the global synchronization trigger pulse, the infrared thermal imager captures the infrared radiation energy spectrum of the target area and converts it into digitized surface temperature distribution data using a built-in temperature calibration curve. At the same trigger moment, the depth ranging module emits a probe light wave and receives the corresponding reflected signal, generating surface geometric distance data containing the three-dimensional coordinates of the target area by calculating the photon time of flight. Based on this hardware-level pulse synchronization mechanism, the surface temperature distribution data and the surface geometric distance data are absolutely aligned in the time axis dimension.

[0027] During the physical synchronization period of acquiring external environmental feature data, the processor concurrently reads the low-level register states of the high-precision joint encoder and the six-axis inertial measurement unit. Based on the current physical rotation feedback data of the driven joints, the processor uses the robot's forward kinematics matrix to calculate the dynamic spatial contour of the robot dog's body shell in the current gait, thereby extracting the envelope deformation phase data that exhibits a periodic pattern. The synchronously operating six-axis inertial measurement unit acquires the mechanical oscillation acceleration signal transmitted along the robot dog's body skeleton at the moment the foot touches the ground. For the acquired time-series mechanical oscillation acceleration signal, the processor performs a fast Fourier transform calculation to extract the peak frequency point with the highest energy distribution in the spectrum, and converts the extracted peak frequency point into impact excitation frequency data.

[0028] After acquiring dynamic physical quantities and external environmental data, the processor initiates a read command to the non-volatile memory area via the internal data bus to extract the preset static reference width. The static reference width represents the maximum physical dimension parameter of the robot dog's body cross-section when all four legs are fully extended and it is stationary. The processor stores the static reference width in the cache as a fixed reference boundary for subsequent calculations. The underlying computing platform thus completes the acquisition process of surface temperature distribution data, surface geometric distance data, envelope deformation phase data, impact excitation frequency data, and the static reference width, providing a complete and spatiotemporally aligned raw data array for subsequent coordinate offset calculations.

[0029] To ensure the determinism of the path planning logic, the processor needs to call preset system parameters before performing real-time calculations. Preset weighting factors. and The acquisition process is based on experimental data under a controlled calibration environment. In this environment, multiple sets of different temperature gradients and corresponding standard reference distances are artificially constructed, and the infrared thermal imager and depth ranging module collect raw data of the target area. For the collected surface temperature distribution data and corresponding surface geometric distance data, the processor calls nonlinear least-squares optimization logic (specifically, the Levenberg-Marquardt iterative algorithm or the Gauss-Newton algorithm) to perform multivariate nonlinear fitting calculations, mapping the temperature difference, raw distance, and the actually measured spatial coordinate offsets as a function. Through the above fitting regression process, preset weighting factors are determined. and The specific numerical values ​​for the project. The pre-defined weighting factors determined through fitting. and The system is organized into an offset calibration matrix and stored in memory for the processor to access during the coordinate offset calculation phase.

[0030] The physical contraction and offset limits of the robot body are set based on the test results of the mechanical linkage interference limit of the robot dog. The upper limit of the dynamic compression ratio is derived from the mechanical structural interference displacement test of the robot dog's drive joints at the end of the movement stroke. By measuring the cross-sectional dimensions of the robot dog's body in the maximum physical contraction state, the upper limit of the dynamic compression ratio is calculated to limit the physical boundary of the robot dog's envelope contraction during the movement cycle. Offset limit constant. The settings are based on the sensor ranges of the infrared thermal imager and the depth ranging module, as well as the preset extreme environmental temperature differences. Offset limit constant. The value limits the maximum amount of spatial offset compensation that the system is allowed to generate, thus avoiding calculation results that exceed the scope of physical common sense due to perceived noise.

[0031] After acquiring surface temperature distribution data, surface geometric distance data, and preset weighting factors, the processor executes a medium deviation correction process to eliminate refraction errors. The processor first extracts the temperature difference based on the surface temperature distribution data. The original distance d is extracted based on the surface geometric distance data, and then the first spatial coordinate offset is calculated according to the preset weighting factor contained in the offset calibration matrix and the formula for calculating the first spatial coordinate offset. : In the formula for calculating the first spatial coordinate offset and As a preset weighting factor, The preset reference calibration temperature, This is a preset baseline distance constant. First spatial coordinate offset. With temperature difference The increase and the original distance The increase in exhibits a monotonically increasing trend. This is because the formula for calculating the first spatial coordinate offset introduces the hyperbolic tangent function. When temperature difference When the preset temperature threshold is exceeded, the first spatial coordinate offset Gradually approaching the offset limit constant This achieves saturated convergence of the compensation values. Simultaneously, the exponential structure of the last term in the formula for calculating the first spatial coordinate offset ensures that, in the near-field environment, i.e., the original distance... When approaching zero, the first spatial coordinate offset Automatic zeroing. Based on this mathematical structure, the first spatial coordinate offset... The output value is strictly limited to the non-negative range and does not exceed the offset limit constant. Defined physical boundaries.

[0032] The first spatial coordinate offset was calculated. Then, the processor will use the first spatial coordinate offset. As compensation input, the occupancy probability of the local cost map is restored. The processor extracts the grid occupancy probability distribution of the corresponding target area in the local cost map, and for the grid coordinates affected by thermal plume refraction, it uses the first spatial coordinate offset. Inverse translation compensation along the coordinate axes is performed. By correcting the distribution threshold of the occupancy probability in the spatial dimension, the perceptual ghost image, which was originally displaced due to refraction interference, is restored to its true physical spatial position, thereby generating a first environmental geometric boundary in the local cost map after removing optical artifacts. The first environmental geometric boundary provides a spatial reference benchmark with physical truth properties for subsequent identification of narrow point regions.

[0033] After correcting for spatial deviations, the processor performs a dealiasing process in the temporal domain to remove noise interference from mechanical vibrations. The processor extracts the impact excitation frequency data and obtains the current data acquisition frequency from the depth ranging module. The processor calculates the ratio between the impact excitation frequency data and the data acquisition frequency, determining the interference weight by checking if this ratio falls within a preset frequency domain aliasing interval. The frequency domain aliasing interval is defined as a frequency-sensitive region with a specific bandwidth, centered at integer multiples of the data acquisition frequency. When the impact excitation frequency data approaches an integer multiple of the data acquisition frequency, the processor determines that the interference weight exceeds a first preset threshold, indicating a risk of phase overlap between the current mechanical vibration displacement and the sampling time sequence, which could easily lead to physical position artifacts in the generated point cloud data.

[0034] For situations where the interference weight exceeds a first preset threshold, the processor initiates a phase-shifting triggering mechanism based on displacement peak avoidance. The processor extracts the real-time phase of the robot dog's motion from the envelope deformation phase data and determines the physical moment when the robot dog's body reaches its displacement peak within the vibration cycle based on the real-time phase. The processor dynamically delays or advances the edge of the sampling trigger pulse from the depth ranging module, adjusting the sampling trigger timing at the microsecond level to ensure that the triggering time of the sampling trigger pulse avoids the displacement peak phase pointed to by the real-time phase. By locking the data acquisition action within a relatively stable phase range of the robot body, the de-aliased local point cloud output by the depth ranging module eliminates spatial position ambiguity caused by high-frequency vibration, improving the instantaneous accuracy of environmental perception.

[0035] The processor performs dynamic envelope planning for extreme environments based on the de-aliased local point cloud and the first environmental geometric boundary. The processor performs contour extraction algorithms within the first environmental geometric boundary to identify regions where the physical space width between obstacles falls within a first preset width range, defining these as narrow point regions. The upper limit of the first preset width range is set to 1.2 times the static baseline width. For the identified narrow point regions, the processor extracts their geometric width and, combined with envelope deformation phase data, initiates the cross-window prediction logic.

[0036] The processor first establishes a calculation model for real-time passage capability. Based on the envelope deformation phase data, the processor calculates the dynamic envelope width of the robot dog's fuselage relative to the travel axis under the current motion rhythm. The processor constructs a difference function, defined as the physical difference between the geometric width of the narrow point region and the dynamic envelope width. To quantify passage reliability, the processor introduces a preset monotonically increasing decision function. When the result of the difference function changes from negative to positive, that is, when the geometric width of the narrow point region begins to exceed the dynamic envelope width, the monotonically increasing decision function generates an extremely high gradient response in the critical region, causing the output passage confidence to exhibit a non-linear growth trend, thereby accurately capturing the transient moment when the fuselage's physical envelope contracts to a point sufficient to pass through the narrow point region.

[0037] To generate the first phase window available for scheduling by the executor, the processor calculates the phase passage decision index according to the formula for calculating the phase passage decision index. : In the formula for calculating the phase-based access decision index For envelope deformation phase data, The geometric width of the narrow point region. The static reference width, The dynamic compression rate is preset based on the physical structure of the robot dog. This is a pre-defined spatial compensation term with the same dimensions as the geometric width, and its physical units are consistent with the geometric width. It is used in the formula for calculating the phase access decision index. This is a preset gain factor used for dimensional normalization. Its dimension is the reciprocal of length, and its unit is reciprocal of geometric width, used to cancel out the physical properties of the length dimension in the formula. The formula for calculating the phase passage decision index is... The structure simulates the envelope width evolution of a quadruped robot dog under different gait phases. The processor obtains the phase travel decision index, which dynamically fluctuates with the envelope deformation phase data, by performing logical operations to calculate the phase travel decision index formula. Subsequently, the processor extracts the phase passage decision index. The continuous phase interval exceeding the second preset threshold is defined as the first phase window. The first phase window represents the phase range within which the physical envelope of the robot dog satisfies the safe passage conditions in each gait cycle.

[0038] After acquiring the first phase window, the processor converts the phase range of the first phase window into a corresponding timestamp sequence via the internal control bus and maps it to the stepping control axis of the drive mechanism. The actuator dynamically adjusts the robot dog's gait frequency parameters or foot dwell time based on the received motion rhythm commands. During this process, the processor compares the time overlap between the current gait phase and the first phase window in real time. If the current gait phase prediction deviates from the first phase window, the processor immediately corrects the motion rhythm commands through a closed-loop feedback loop, accelerating or decelerating the current gait cycle to ensure that the robot dog's body is at the physical phase moment of minimum dynamic envelope width, achieving precise time overlap with the physical displacement process through the narrow area. Through this spatiotemporal coordinated scheduling, the robot dog achieves non-contact, smooth passage through extremely narrow spaces.

[0039] To address complex and ever-changing industrial inspection scenarios, this system supports equivalent replacements of various sensing modules. In inspection conditions with severe photoelectric interference such as strong dust, water mist, or reflections from glass curtain walls, the physical means of acquiring surface geometric distance data are not limited to depth ranging modules; ultrasonic dot matrix sensors or millimeter-wave radar modules can also be used to perform detection actions. Ultrasonic dot matrix sensors obtain obstacle distances through the time difference of sound wave return, while millimeter-wave radar modules utilize the Doppler effect of electromagnetic waves to calculate spatial positions. These equivalent sensing modules can also output surface geometric distance data containing the three-dimensional coordinates of the target area, which is then used as the raw distance input into the coordinate offset calculation logic. The physical quantities acquired based on different sensing mechanisms maintain consistency in data structure, without affecting the subsequent calculation path of the first spatial coordinate offset and the environmental boundary reconstruction effect.

[0040] When calculating the first spatial coordinate offset, the mathematical model used to achieve numerical saturation convergence also involves various equivalent logical transformations. Besides the hyperbolic tangent function used in the formula for calculating the first spatial coordinate offset, the processor can also call the arctangent function, error function, or a specific piecewise linear saturation function to perform nonlinear mapping calculations. Both the arctangent function and the error function possess the physical evolution characteristics of monotonically increasing and having horizontal asymptotes. When temperature difference data fluctuates significantly, these monotonically bounded functions can produce the same technical effect as the hyperbolic tangent function, that is, limiting the first spatial coordinate offset within a preset physical boundary, preventing the compensation value from failing due to mathematical divergence, thus preventing the path planning instruction from failing. Regardless of the specific function model used, as long as it conforms to the evolution law of tending towards the physical extreme value with increasing temperature difference, it can effectively suppress refraction errors and accurately restore the physical position.

[0041] Thus, through the physical support of the hardware implementation architecture, the synchronous perception of multi-source features, the offline calibration of system parameters, the coordinate correction of medium deviation, the dealiasing processing of the perception time domain, and the window prediction of dynamic envelope, the robot dog path planning system based on intelligent inspection has built a complete technical closed loop from environmental perception to precise execution.

[0042] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A path planning method for a robot dog based on intelligent inspection, characterized in that, The method includes the following steps: Step S1: Obtain surface temperature distribution data and surface geometric distance data of the target area, and obtain the envelope deformation phase data, impact excitation frequency data and preset static reference width of the robot dog; Step S2: Calculate the first spatial coordinate offset using a preset offset calibration matrix, and correct the occupancy probability threshold in the local cost map based on the first spatial coordinate offset to obtain the first environmental geometric boundary; Step S3: Determine the interference weight between the impact excitation frequency data and the data acquisition frequency, and when the interference weight exceeds the first preset threshold, adjust the sampling triggering timing of the data acquisition frequency so that the sampling triggering timing avoids the peak phase of the physical displacement generated by the robot dog's movement, and output the de-aliased local point cloud. Step S4: Based on the dealiasing local point cloud, identify narrow point regions in the first environmental geometric boundary where the physical space width is within a first preset width range, and calculate the first phase window corresponding to when the dynamic envelope width of the robot dog reaches the minimum value according to the envelope deformation phase data; Step S5: Generate motion rhythm commands according to the first phase window, so that the longitudinal axis of the robot dog's body passes through the narrow point region within the first phase window.

2. The method according to claim 1, characterized in that, Step S2 specifically includes: A mapping relationship is established between the first spatial coordinate offset and the temperature difference, and the original distance extracted from the surface geometric distance data; wherein, the first spatial coordinate offset increases monotonically with the increase of the temperature difference and the increase of the original distance, and the first spatial coordinate offset tends to a preset offset limit constant after the temperature difference exceeds a preset temperature threshold.

3. The method according to claim 2, characterized in that, First spatial coordinate offset The calculation formula is as follows: in, This is the first spatial coordinate offset. The offset limiting constant is... The temperature difference in the surface temperature distribution data. The preset reference calibration temperature, The original distance, The preset reference distance constant, and The preset weighting factor is used; the offset calibration matrix contains the preset weighting factor. and .

4. The method according to claim 1, characterized in that, The output value of the first spatial coordinate offset is limited to a non-negative range.

5. The method according to claim 1, characterized in that, Step S3 specifically includes: When the ratio of the impact excitation frequency data to the data acquisition frequency falls into a preset frequency domain aliasing interval, the real-time phase of the robot dog's motion is extracted from the envelope deformation phase data. The sampling trigger timing is adjusted according to the real-time phase so that the deviation between the edge of the trigger pulse of the sampling trigger timing and the phase of the peak phase of the physical displacement is greater than a preset deviation threshold.

6. The method according to claim 1, characterized in that, Step S4 specifically includes: Based on the envelope deformation phase data, the dynamic envelope width of the robot dog during the motion cycle is determined; A difference function is established between the dynamic envelope width and the geometric width of the narrow point region. Based on the difference function and a preset monotonically increasing judgment function, the passability confidence of the robot dog in the current phase is determined. The passability confidence shows an increasing trend in the critical region where the difference turns from negative to positive.

7. The method according to claim 6, characterized in that, The passage confidence score is calculated by using the phase passage decision index. The formula is as follows: in, This is the phase passage decision index. The envelope deformation phase data, The geometric width of the narrow point region. The static reference width, The dynamic compression rate is preset based on the physical structure of the robot dog. This is a preset spatial compensation term with the same dimensions as the geometric width. This is the preset gain factor used for dimensional normalization.

8. The method according to claim 7, characterized in that, The value of the dynamic compression ratio is limited to a preset range of physical contraction limits of the body, and the first phase window is a continuous phase range in which the phase passage decision index is greater than the second preset threshold.

9. The method according to claim 1, characterized in that, The upper limit of the first preset width range is 1.2 times the static reference width.

10. A path planning system for a robot dog based on intelligent inspection, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement a path planning method for a robot dog based on intelligent inspection as described in any one of claims 1 to 9.

Images