A multi-source thermal inertia fused mars rover path planning method and system

Abstract

The application discloses a multi-source thermal inertia fusion Mars rover path planning method and system, wherein the method comprises the following steps: generating multi-scale thermal inertia distribution data of a target area; based on this, combining a visible light image carried by the Mars rover, terrain recognition and soil cover thickness inversion are carried out; based on the terrain recognition and soil cover thickness inversion results, a grid cost map is constructed for the target area; based on the grid cost map, global path search is carried out to obtain a minimum cumulative cost path from a starting point to an ending point; in the process of driving the Mars rover along the minimum cumulative cost path, data is acquired in real time by using a vehicle-mounted sensor, a local cost map in the grid cost map is updated, and when it is detected that there is an impassable area in the current execution path, local re-planning is carried out. The method fuses satellite wide-range pre-scouting and vehicle-mounted high-precision detailed investigation, directly embeds thermal inertia mechanical properties into path search, and realizes the conversion from qualitative evaluation to quantitative trajectory.

Description

Technical Field

[0001] This invention relates to the field of autonomous navigation and path planning technology for deep space exploration, and more specifically to a method and system for Mars rover path planning using multi-source thermal inertia fusion. Background Technology

[0002] Risk assessment and prediction are crucial for Mars rover path planning. Due to the complexity of the Martian geological environment, rovers need to plan safe routes in real time during their journey. Existing path planning methods are mainly based on geometric maps, such as elevation maps, slope maps, and rock distribution maps, constructed using visible light or lidar. However, these methods can only avoid surface geometric obstacles and cannot detect the risks posed by terrain with a hard upper crust and a soft lower crust, or by soil structures with varying thicknesses of soft upper and hard lower layers. Several rover accidents involving getting stuck on Mars have been caused by terrain that appears flat but has fragile mechanical properties.

[0003] In the prior art, patent application CN121659815A, entitled "A Non-Contact Passability Prediction Method and System for Mars Rover Based on Thermal Inertia," estimates thermal inertia and maps it to passability indices such as the conic index (CI), then outputs three passability status levels: safe to pass, critical risk, and high risk not to pass, along with corresponding navigation and control suggestions. However, this method only outputs qualitative levels and suggestions, without generating specific driving paths or constructing a cost map and path search algorithm suitable for graph search. When an "avoidance" suggestion is output, how the Mars rover plans its path still requires processing by other systems, lacking a closed loop from perception to a specific trajectory.

[0004] Another patent application, CN121720611A, entitled "A Method and System for Temperature Monitoring of Mars Probes Based on Time-Control Factor Optimization," aims to improve the accuracy of thermal inertia estimation, but it only involves optimizing the time window for temperature data acquisition. Furthermore, existing techniques for analyzing the Martian surface using thermal inertia largely rely on a single data source; or they utilize only orbital satellite thermal infrared data, such as THEMIS data, for large-scale geological mapping with a resolution of hundreds of meters, which cannot meet the meter-level accuracy required for local path planning by the Mars rover; or they only use onboard thermal infrared cameras for local perception, lacking forward-looking prediction of a wider area, thus limiting the field of vision for path planning.

[0005] Therefore, a multi-source thermal inertia path planning method is needed that can integrate large-scale low-resolution satellite data with local high-resolution vehicle data. Summary of the Invention

[0006] In view of the above problems, the present invention proposes a Mars rover path planning method and system based on multi-source thermal inertia fusion, aiming to overcome or at least partially solve the above problems.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] In a first aspect, the present invention provides a Mars rover path planning method based on multi-source thermal inertia fusion, comprising: S10. Acquire multi-source thermal inertia data of the target area and perform fusion processing to generate multi-scale thermal inertia distribution data; S20. Based on the multi-scale thermal inertia distribution data, combined with the visible light images on the Mars rover, perform terrain identification and soil cover thickness inversion. S30. Divide the target area into two-dimensional grids, assign a passability cost to each two-dimensional grid based on the terrain recognition results and the soil cover thickness inversion results, and construct a grid cost map. S40. Perform a global path search based on the grid cost map to obtain the path with the minimum cumulative cost from the starting point to the ending point; S50. During the process of the Mars rover traveling along the minimum cumulative cost path, the onboard sensors acquire data in real time and update the local cost map in the grid cost map. When an impassable area is detected in the current execution path, local replanning is performed to generate a detour path.

[0009] Furthermore, the acquisition and fusion processing of multi-source thermal inertia data of the target region specifically includes: Obtain satellite thermal inertia distribution map based on the thermal infrared imaging system carried by the Mars orbiting satellite, and vehicle thermal inertia distribution map based on the thermal infrared camera on the Mars rover; Unify satellite thermal inertia data and vehicle-mounted thermal inertia data to the Mars geographic coordinate system, and resample all thermal inertia data grid resolution to a uniform resolution; Using the satellite thermal inertia distribution map as the background map, a local fusion window is determined with the rover's current position as the center. Within the local fusion window, for the areas where the satellite thermal inertia distribution map and the rover's thermal inertia distribution map overlap, a weighted average method is used to calculate the fused thermal inertia value. For the areas within the local fusion window that only contain the satellite thermal inertia distribution map, a spatial interpolation method is used to improve the resolution.

[0010] Furthermore, the terrain recognition includes rocky area recognition, thin-shell terrain recognition, and soil-covered structure recognition.

[0011] Furthermore, the thin-shell terrain recognition adopts a comprehensive judgment based on both thermal inertia dynamic features and visible light structural features; The dynamic characteristic criterion of thermal inertia is as follows: based on the multi-scale thermal inertia distribution data and temperature time series, the heating rate of the target region during the heating phase is extracted. and cooling rate during the cooling phase ; Obtain the baseline heating rate of bare soil areas under the same environmental conditions. and reference cooling rate When satisfied and At that time, it was determined to be a candidate for thin-shell terrain; among them This is the threshold coefficient for determining the heating rate; This is the threshold coefficient for determining the cooling rate; The visible light structural feature criteria are: extracting edge warping features, surface crack network features, surface texture features, and color and spectral features from the visible light images on the Mars rover. The thermal inertia dynamic feature criterion and the visible light structural feature criterion are weighted and fused to obtain a thin-shell confidence score; when the thin-shell confidence score is greater than or equal to the confidence threshold, it is determined to be a thin-shell terrain.

[0012] Furthermore, the soil cover thickness inversion is performed by calculating the soil cover thickness using a pre-calibrated thermal inertia-soil cover thickness mapping relationship. This mapping relationship is established through ground simulation experiments, and corresponding mapping curves are calibrated and stored as curve libraries for different types of fire soil.

[0013] Furthermore, step S20 also includes a wheel settlement determination step based on the thickness of the overburden layer, specifically including: Obtain the Martian wheel diameter parameters and calculate the estimated settlement of the wheel within the soil cover layer based on the soil cover thickness. When the thickness of the overburden layer is greater than or equal to the critical overburden layer thickness, and the estimated settlement amount is greater than the safe settlement threshold, it is determined to be a high-risk area; when the thickness of the overburden layer is less than the critical overburden layer thickness, and the estimated settlement amount is less than or equal to the safe settlement threshold, it is determined to be a passable area.

[0014] Furthermore, the process of assigning a passability cost to each two-dimensional grid based on the terrain identification results and the soil cover thickness inversion results to construct a grid cost map specifically includes: Set the passability cost of rocks in the rocky region with an equivalent diameter greater than or equal to one-third of the Martian wheel diameter, as well as the passability cost of thin-shell terrain, to the preset maximum cost. For areas with soil-covered structures, the thickness of the soil cover layer is mapped to a passability cost through a continuous cost function. The greater the thickness of the soil cover layer, the higher the passability cost. When the thickness of the soil cover layer exceeds the danger threshold, it is set to the preset maximum cost. The final passability cost of each two-dimensional grid is a weighted sum of terrain cost, slope cost, and safety distance cost; wherein, the slope cost is calculated based on the tilt angle of the surface within the two-dimensional grid, and the safety distance cost is used to maintain a preset safety distance between the planned path and high-risk terrain.

[0015] Furthermore, the global path search employs A Algorithm, the A The algorithm's heuristic function is the Euclidean distance multiplied by a terrain adaptation coefficient; the terrain adaptation coefficient is dynamically adjusted based on the passability cost of the terrain ahead of the path.

[0016] Furthermore, when an impassable area is detected in the current execution path, local replanning is performed to generate a detour path, specifically including: After each update of the local cost map, check whether the path points in the current execution path located in front of the rover fall into a two-dimensional grid with a passability cost of the preset maximum cost value. If so, it is determined that there is an impassable area in the current execution path. The local replanning adopts D The Lite algorithm takes the rover's current position as the starting point and the first safe point after the impassable area on the current execution path as the local endpoint, and re-plans the detour path on the local cost map. If the path length of the local replanning search exceeds a preset multiple of the current execution path length or the search time exceeds a preset duration, and no feasible path is still found, the replanning is deemed to have failed. The rover then returns to the previously recorded safe position and re-executes the global path search.

[0017] Secondly, this invention provides a Mars rover path planning system that integrates multiple sources of thermal inertia. Using the aforementioned method, the system includes: The multi-source thermal inertia acquisition and fusion module is used to acquire multi-source thermal inertia data of the target area and perform fusion processing to generate multi-scale thermal inertia distribution data. The terrain identification and soil cover thickness inversion module is used to perform terrain identification and soil cover thickness inversion based on the multi-scale thermal inertia distribution data and the visible light images on the Mars rover. The grid cost map construction module is used to divide the target area into two-dimensional grids, assign a passability cost to each two-dimensional grid based on the terrain recognition results and the soil cover thickness inversion results, and construct a grid cost map. The global path planning module is used to perform a global path search based on the grid cost map to obtain the path with the minimum cumulative cost from the starting point to the ending point. The local rolling replanning module is used to acquire data in real time using onboard sensors during the rover's journey along the minimum cumulative cost path, update the local cost map in the grid cost map, and perform local replanning to generate a detour path when an impassable area is detected in the current execution path.

[0018] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a Mars rover path planning method and system based on multi-source thermal inertia fusion, which has the following beneficial effects: This invention integrates satellite, vehicle-mounted thermal infrared, and active thermal excitation data to construct a multi-scale thermal inertia distribution, achieving a unified approach to large-scale pre-reconnaissance and local fine-grained perception; it maps the soil cover thickness to a continuous cost function, guiding the path algorithm to automatically weigh risks and distances to obtain a specific path; it realizes the transformation from qualitative passability suggestions to quantitative trajectory planning, while supporting local replanning in dynamic environments.

[0019] This invention uses A With D The Lite algorithm performs global search and local replanning respectively; it identifies thin-shell terrain based on heating / cooling rate and quantifies soil cover accessibility based on thickness continuity cost, thus solving the path planning problem of visually indistinguishable terrain. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of the Mars rover path planning method based on multi-source thermal inertia fusion provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of thermal inertia distribution provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the passage cost provided in an embodiment of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] This invention discloses a multi-source thermal inertia fusion method for Mars rover path planning, which is particularly suitable for path planning of Mars rovers in complex terrains such as thin shells, soil coverings, and rocks. Figure 1 As shown, the method includes the following steps: S10. Acquire multi-source thermal inertia data of the target area and perform fusion processing to generate multi-scale thermal inertia distribution data; S20. Based on multi-scale thermal inertia distribution data and combined with visible light images on the Mars rover, terrain identification and soil cover thickness inversion are performed. S30. Divide the target area into two-dimensional grids, assign a passability cost to each two-dimensional grid based on the terrain recognition results and the soil cover thickness inversion results, and construct a grid cost map. S40. Perform a global path search based on the grid cost map to obtain the path with the minimum cumulative cost from the starting point to the end point; S50. During the process of the Mars rover traveling along the path of minimum cumulative cost, the onboard sensors acquire data in real time and update the local cost map in the grid cost map. When an impassable area is detected in the current execution path, local replanning is performed to generate a detour path.

[0024] Next, each of the above steps will be explained in detail.

[0025] In step S10 above, multi-source thermal inertia data of the target area are acquired and fused to generate multi-scale thermal inertia distribution data; specifically including: (1) Acquisition method of thermal infrared remote sensing by orbiting satellites: Using a thermal infrared imaging system aboard a Mars orbiter, thermal infrared images of the target area at different time phases were acquired over at least one Martian day. The surface brightness temperature of each pixel in each time-phase image was then extracted. Combined with the albedo acquired synchronously by satellite Solar altitude angle and atmospheric parameters The apparent thermal inertia formula is used to calculate the satellite's thermal inertia value:

[0026] in, This represents the satellite's thermal inertia value. This is a comprehensive correction factor for the Martian environment, ranging from 0.6 to 1.2, and is dynamically adjusted based on atmospheric opacity. This represents the maximum surface temperature difference during the satellite observation period. The Martian solar constant; The extinction coefficient is used. The calculation results are used to generate a satellite thermal inertia distribution map with a spatial resolution of 100m to 1km and a coverage area of ​​tens of square kilometers.

[0027] (2) Acquisition method of the Mars rover's onboard thermal infrared camera: The Mars rover carries an uncooled long-wave infrared camera, operating in the 8–14 μm wavelength range, with a detector resolution of 640 × 512 pixels. It first continuously images a target area within a 1–5 m radius ahead, with sampling intervals ranging from 30 seconds to 2 minutes, and the duration depending on mission requirements, ranging from 30 minutes to 8 hours. Then, the acquired thermal infrared image sequences undergo radiometric calibration, bad pixel correction, non-uniformity correction, and geometric registration. The temperature value of each pixel in each frame is extracted to construct a temperature time series. Combined with the albedo obtained by the onboard environmental perception unit Solar irradiance Calculate the vehicle's thermal inertia value based on atmospheric temperature and pressure parameters. :

[0028] in, This is a correction factor for vehicle-mounted observations, with a value range of 0.7 to 1.1, which is corrected in real time based on the observed zenith angle and atmospheric path length. This is temperature difference data, specifically the difference between the highest and lowest temperatures; The terrain geometry factors are mainly affected by slope and aspect, and are extracted from vehicle-mounted stereo vision or digital elevation maps; the calculation results are used to generate a vehicle-mounted thermal inertia distribution map with a spatial resolution of 0.1m~0.5m.

[0029] (3) Fusion of multi-source thermal inertia data: The satellite thermal inertia data and vehicle-mounted thermal inertia data in the multi-source thermal inertia distribution map are registered and matched using a spatial coordinate system. The specific fusion process is as follows: a) Spatial Coordinate System 1: Satellite thermal inertia data uses the Martian geographic coordinate system (longitude and latitude), while onboard thermal inertia data uses the rover's own coordinate system (with the rover's geometric center as the origin, the forward direction as the positive X-axis, and the left side as the positive Y-axis). First, the rover's precise position in the Martian geographic coordinate system is obtained through its onboard inertial navigation unit. and heading angle Then, each pixel in the vehicle thermal inertia data Convert to the Martian geographic coordinate system using the following formula:

[0030] in: The coordinate transformation of a pixel in the vehicle-mounted thermal infrared image corresponds to the Martian geographical longitude. The coordinate transformation of a pixel in the vehicle-mounted thermal infrared image corresponds to the Martian geographical latitude. The average radius of Mars is used. After the conversion, all data are unified to the Martian geographic coordinate system.

[0031] b) Fusion Strategy and Scope Determination: The satellite thermal inertia distribution map is used as the global background map. Let the current position of the Mars rover be... ,by A square local fusion window with side length L is defined centered on the specified element. The value of L is dynamically determined according to the following rules: When the Mars rover is in cruise mode, the speed is >0.1m / s, and L=20m is taken; When the Mars rover is in scientific exploration or stationary mode, take L = 20~50m; When there is suspicious terrain identified by satellite (such as an area with low thermal inertia anomaly) in the area ahead, take a larger value of 100m; Within the local fusion window, the original satellite data is replaced or corrected using vehicle-mounted high-resolution thermal inertia data.

[0032] c) Weighted average fusion of overlapping areas: For areas where satellite data and vehicle data overlap within the fusion window, a weighted average method is used to calculate the fused thermal inertia value.

[0033] in: This represents the thermal inertia value after fusion. This is the weighting coefficient for satellite thermal inertia. Here are the weighting coefficients for vehicle thermal inertia. The expressions for these two weighting coefficients are:

[0034]

[0035] in: The confidence coefficient for calibrating the vehicle-mounted sensor can be set to 0.95; To calibrate the confidence coefficient for the satellite sensor, a value of 0.85 can be used. For the variance of the vehicle thermal inertia inversion error, take... ; Let the variance of the satellite thermal inertia inversion error be taken as... , It is a unit of thermal inertia; For in-vehicle spatial resolution, a value of 0.1~0.5m is used; For satellite spatial resolution, a range of 10~100m is used. d) Interpolation Enhancement in Non-Overlapping Regions: For regions within the fusion window that only have satellite thermal inertia distribution maps and lack vehicle-mounted data, spatial interpolation methods are used to improve resolution and enhance their usability in local planning. The specific interpolation strategy is as follows: First, calculate the thermal inertia variability function of the satellite data within the fusion window. :

[0036] in: Spatial distance; The distance is Logarithm of data points; For position The satellite's thermal inertia value at that location; This represents the satellite's thermal inertia value at the corresponding point. , This represents the spatial offset in the longitude and latitude directions.

[0037] If the thermal inertia variability function If the growth is slow over short distances, indicating strong spatial autocorrelation, then Kriging interpolation is used. The interpolation formula is:

[0038] in, Indicates at the interpolation point Estimated value of satellite thermal inertia at the location; Indicates at a known point The satellite's thermal inertia value at that location; express The weighting coefficients, and It is obtained by solving the Kriging equations, which satisfies the conditions of unbiased estimation and minimum variance.

[0039] If the thermal inertia variability function If the spatial autocorrelation is weak, then the inverse distance weighted interpolation method is used instead:

[0040] in: interpolation point to a known point The distance; It is a power exponent, with a value range of 1 to 3.

[0041] After interpolation, the spatial resolution of the satellite data area is improved to 2m×2m. At the same time, the boundaries of macroscopic geological units identified by the satellite are preserved in the interpolation results. The specific method is as follows: before interpolation, the satellite data is divided into geological units to identify the boundary lines of different geological units. During the interpolation process, the pixels on both sides of the boundary line are forced not to participate in the interpolation calculation of the other, thereby avoiding the loss of boundary information caused by blurring.

[0042] e) Output Results: The final output is the fused multi-scale thermal inertia distribution map, such as... Figure 2 As shown, it includes: Global background layer: Covers a 100m×100m area around the Mars rover, with a resolution of 2m, and is stored in the global map database; Local high-precision layer: Covers an L×L area around the Mars rover with a resolution of 0.5m, and is stored in a high-speed cache for real-time path planning; Confidence layer: Each grid cell is accompanied by a confidence value (0~1), used for risk measurement in path planning; The fused thermal inertia distribution data is output as a raster layer to the third part for cost map construction.

[0043] In step S20 above, terrain identification and soil cover thickness inversion are performed based on multi-scale thermal inertia distribution data and combined with visible light images from the Mars rover. The visible light images from the Mars rover are used to extract the geometric shape, texture, edges, and structural features of the surface, complementing the thermal inertia features to jointly support terrain identification. This step specifically includes: (1) Stone area identification: When the thermal inertia value after fusion The threshold for determining thermal inertia in the area of ​​the stone is greater than the threshold value. Furthermore, the variance of local spatial heterogeneity after fusion The variance of the thermal inertia of the region greater than the threshold for judging the stone block area At that time, it was initially determined to be a candidate area for stones. The value range is 180~220. Simultaneously analyze the visible light images on the Mars rover and extract the following features: a) edge gradient amplitude is higher than a preset threshold, forming a closed polygonal outline; b) surface texture roughness is significantly higher than the surrounding bare soil; c) there is obvious shadow or highlight reflection; d) aspect ratio and roundness conform to the morphological characteristics of gravel or rock. When at least two of the thermal inertia criterion and visual criterion are satisfied, it is finally determined to be a rocky region. Further, the equivalent diameter is calculated based on the area of ​​the connected domain of the rock. Where A represents the area of ​​the connected region of the stones; when , The Martian wheel track is marked; areas marked as high-risk and impassable rocky zones.

[0044] (2) Thin-shell terrain identification: Thin-shell terrain is the key identification target of this invention. It is comprehensively judged by dual criteria of dynamic characteristics of thermal inertia and visible light structural characteristics. When either criterion triggers a high-risk level, it is marked as thin-shell.

[0045] a) Determination of dynamic characteristics of thermal inertia: Extract the heating rate of the target region during the heating phase. :

[0046] in, This refers to the temperature change during the heating phase, i.e., the temperature difference during the heating process, expressed in °C. The duration of the heating phase, measured in seconds (s).

[0047] Extract the cooling rate of the target area during the cooling phase. :

[0048] in, The temperature change is the amount of cooling, i.e., the temperature drop, expressed in °C. The duration of cooling is expressed in seconds (s).

[0049] Obtain the baseline heating rate of bare soil areas under the same environmental conditions. and reference cooling rate When satisfied and At that time, it was determined to be a candidate for thin-shell terrain; among them The threshold coefficient for determining the heating rate is 0.2 to 0.5. The threshold coefficient for determining the cooling rate is 0.2 to 0.5. The thin-shell surface has a small heat capacity and heats up quickly, while the soft soil underneath provides thermal insulation, resulting in slow cooling. This exhibits a "fast rise and slow fall" thermal response curve characteristic.

[0050] b) Determination of visible light structural features: The following typical visual features of the thin-shell terrain were extracted from the visible light images on the Mars rover: Edge warping features: At the boundaries of thin-shell strata (such as the contact zone with surrounding bare soil or rock), visible light images can detect an upward curling or broken upturned shape at the crustal edge. The Canny edge detection operator is used to extract the edges, and the gray-level gradient change rate on both sides of the edge line is calculated. When the gradient directions on both sides of the edge show a significant divergence (gradient angle > 90°) and the edge line has a continuous arc-shaped upturned shape, it is determined to be an edge warping feature.

[0051] Surface fracture network: Thin-crust strata develop polygonal drying or shrinkage fractures under arid or thermal stress. Morphological closure operations are used to extract fracture lines, and the number of fracture intersections and fracture widths (typically 1-5 mm, appearing as dark, thin lines in visible light images) per unit area are statistically analyzed. A fracture feature is defined as a fracture density > 0.5 fractures / m² and a fracture network exhibiting a closed polygonal grid structure.

[0052] Color and spectral characteristics: Due to the presence of cementing materials (such as sulfates, chlorides, or thermal alteration products), thin-crust strata often exhibit a lighter or brighter hue than the surrounding sand and soil in the visible light band, and their spectral reflectance curves may show absorption characteristics in specific wavelength ranges (such as 0.8–1.0 μm). Normalized difference indices are calculated using multispectral or color images to enhance the contrast between the thin crust and bare soil.

[0053] c) Comprehensive judgment: The thermal inertia dynamic characteristic criterion and the visible light structural characteristic criterion are weighted and fused to calculate the thin-shell confidence score. :

[0054] in: The score for determining thermal inertia is 1.0 if the "rapid rise and slow fall" condition is met, and otherwise a continuous value of 0 to 0.5 is taken. The score for visual judgment is a comprehensive score of features such as edge warping, crack networks, and texture anomalies, which is a weighted sum of each sub-feature. , These are the corresponding weighting coefficients, taken as 0.6 and 0.4 respectively.

[0055] When the confidence score of the thin shell is greater than or equal to the confidence threshold ( When the visual features are not obvious (e.g., the thin shell is covered by a thin layer of sand), the thermal inertia features can still trigger identification; conversely, if the thermal inertia data is affected by the weather (e.g., cloudy weather interferes with temperature differences), the visual features can issue an early warning independently.

[0056] (3) Identification of soil cover structure and thickness inversion: a) Determination of thermal inertia: When the thermal inertia value after fusion satisfy At that time, among them, This represents the threshold for determining the thermal inertia of soft, sandy soil. This indicates the threshold for determining the thermal inertia of the stone region, and , .

[0057] b) Inversion of overburden thickness: Calling the pre-calibrated thermal inertia-overburden thickness mapping relationship Calculate the thickness of the overburden layer (Unit: cm). This mapping relationship was established through ground simulation experiments, and the mapping function was obtained by nonlinear regression fitting, in the form of:

[0058] Where h represents the fitted soil cover thickness; h eff Indicates the effective cover layer thickness; Multiple mapping curves are established for different types of fire soil (different particle sizes, densities, and degrees of cementation). In practical applications, the matching mapping relationship is automatically selected based on the thermal inertia value, temperature response characteristics, and the particle size distribution estimated in the visible light image.

[0059] Mapping relationship This can be achieved in the following ways: In ground-based laboratories and Mars simulation sites, simulated Martian soil samples with varying thicknesses of upper cover layers were prepared, covering typical Martian soil types (including aeolian sand, volcanic ash, etc.). For each type of Martian soil, under controlled thermal conditions (simulating Martian atmospheric pressure, diurnal temperature variation, and solar irradiance), the relationship between thickness and thermal inertia was established using the following method: The volcanic soil sample was laid in a test trench of fixed size, with the thickness gradually increasing from 0.5 cm to 20 cm, in increments of 0.5 cm to 1.0 cm. Using a thermal imager with the same wavelength (8–14 μm) as the thermal infrared camera on the Mars rover, the apparent thermal inertia at different thicknesses was measured according to the day-night time-division method in step S10. Simultaneously, the cone index CI at each thickness was measured using a cone penetrator as a verification indicator; each thickness was measured 5 times, and the average value was taken to eliminate random errors.

[0060] For each type of volcanic soil, a nonlinear regression was used to fit the relationship between thickness h and apparent thermal inertia. The mapping relationship between the two is shown in the experiment. The results indicate that the thickness of the cover layer and the thermal inertia exhibit a monotonically increasing but asymptotically saturated relationship: when the cover layer thickness is small, the underlying high thermal inertia bedrock or hard shell significantly contributes to the apparent thermal inertia; as the thickness increases, the thermal inertia gradually approaches that of the pure cover material. Therefore, a logarithmic-power hybrid model is chosen.

[0061] in , , , The regression coefficients obtained by least squares fitting are calibrated separately for each type of fire soil. The goodness of fit R0 2 The requirement is not less than 0.95, and the standard deviation of the residuals is less than 0.5 cm.

[0062] The mapping functions and applicable ranges corresponding to different soil types are stored as a curve library. Each record includes: soil type number and name, effective range of thermal inertia, mapping function parameters, and typical temperature response feature templates. During the actual Mars rover operation, the most suitable mapping curve is automatically selected based on the fused thermal inertia value obtained in step S10 and the temperature response dynamic features and visible light image features extracted simultaneously in step S20.

[0063] c) Wheel settlement determination based on soil cover thickness: The core danger of the soil cover structure is that when the Mars rover travels on the soft soil cover, the soil cover may be sheared and broken, causing the rover to sink until it touches the underlying hard bedrock, or continue to sink to an unacceptable depth.

[0064] This invention establishes the following subsidence determination model: Let the diameter of the Martian wheel be D (in meters), the wheel width be B (in meters), the vertical load be W (in N), and the thickness of the overburden layer be h (in centimeters). The underlying bedrock is considered an incompressible rigid stratum. The overburden layer is soft soil, and its bearing capacity is characterized by the conic index CI (in MPa). The mapping relationship between CI and thermal inertia I is established in advance through ground tests.

[0065] When a vehicle travels on a soft soil cover, the primary cause of subsidence is the compression and shearing of the soil cover. When the soil cover thickness *h* is less than the depth of influence of the stress applied by the vehicle, the underlying bedrock provides support, limiting further subsidence. This invention defines "effective soil cover thickness". And calculate the estimated settlement of the wheel in the pure overburden layer. :

[0066] in: The attenuation coefficient is related to the cover material, ranging from 0.5 to 1.5, and is calibrated experimentally. The physical meaning of this formula is: when the cover layer is thick, the settlement approaches the maximum settlement value of pure soft soil; when the cover layer is thin, the underlying bedrock limits the settlement through stress diffusion effect.

[0067] Estimated subsidence volume With respect to the safe subsidence threshold allowed by the Mars rover (Take 1 / 2 or 3~5cm of the ground clearance) for comparison, and solve for the critical overburden thickness that leads to dangerous settlement. .

[0068] make = Solving for:

[0069] Among them, S max To the maximum estimated subsidence; When the thickness of the cover layer h is greater than or equal to the critical thickness of the cover layer And the estimated subsidence volume Greater than the safe subsidence threshold When the soil cover thickness h is less than the critical soil cover thickness, it is identified as a high-risk area; And the estimated subsidence volume Less than or equal to the safe settlement threshold At that time, it is determined to be a passable area.

[0070] For ease of engineering application, the critical overburden thickness is defined as follows: Simplified to fixed thresholds directly measured through ground simulation experiments, such as those measured for typical Mars rover parameters. =10cm. Also, add the intermediate threshold: This is used to distinguish medium-risk areas.

[0071] The final subsidence risk assessment results are shown in Table 1.

[0072] Table 1: Results of Settlement Risk Assessment

[0073] For each soil-covered structure grid, output the following information: soil cover thickness h (unit: cm); estimated settlement. (Unit: cm); Subsidence risk level (low / medium / high); Corresponding passability cost: Low risk cost 20~40, medium risk cost 41~70, high risk cost 100.

[0074] (4) Other terrain types For areas with exposed hard bedrock (integrated thermal inertia) 180, and the visible light image shows blocky, continuous rock strata or large rock outcrops with no loose surface cover), is identified as high-bearing-capacity terrain, which has the lowest cost. For bare soil areas (integrating thermal inertia) (and there were no obvious abnormalities in visual features), so it was determined to be a normal passage area.

[0075] In step S30 above, the target area is divided into two-dimensional grids. Based on the terrain recognition results and the soil cover thickness inversion results, a passability cost is assigned to each two-dimensional grid to construct a grid cost map. Specifically, the planning area is divided into two-dimensional grids, with the side length of each grid determined based on the data source: 0.1m~0.5m is used for grids covering areas with vehicle-mounted data, and 2m~5m is used for grids covering areas with only satellite data. For each grid... Imposing a cost of passage The specific rules are as follows: (1) Lowest cost area: Cost=0~10, including areas with exposed hard bedrock (thermal inertia I>180) and areas with dense bare soil (thermal inertia 100≤I≤180, slope<10°). (2) Low-cost area: Cost=20~40, including sparse small stone area and thin soil cover area (thickness 2cm<h<5cm) (3) Medium cost area: Cost=41~70, including soft sandy areas (thermal inertia 50≤I≤80, Cost=60) and medium-thickness soil cover areas (5cm≤h≤8cm). (4) High-cost areas: Cost = 71~99, including areas with thick soil cover (8cm≤h≤12cm) and areas with dense rocks. (5) Impassable areas: Set to the preset maximum cost, i.e. =100, including thin-shell terrain, large boulders (diameter greater than or equal to 1 / 3 of the diameter of a wheel), regional cliff pits with a soil cover thickness h>12cm, etc.

[0076] For the soil-covered structure region, the cost function takes a continuously differentiable form:

[0077] in: =20, =100, =12cm, It is a non-linear exponent (with a value range of 0.8 to 1.5). The dangerous threshold for the thickness of the overburden layer, when At that time, it is set to the preset maximum value, that is =100. This continuous mapping ensures that small changes in thickness cause a smooth transition in cost, avoiding discontinuous jumps in path search.

[0078] (6) Multi-cost fusion, the final cost is the weighted sum of terrain cost, slope cost and safety distance cost:

[0079] in: For grid The final passability cost is used by the path planning algorithm to calculate the cumulative cost; The cost of the terrain foundation is determined by the preceding steps (terrain type, topsoil thickness, thin shell, rocks, etc.). For example, bare soil is 10, a topsoil thickness of 6cm is approximately 56, and a thin shell or large rocks is 100. The slope cost is calculated based on the tilt angle of the surface within the grid. The steeper the slope, the higher the cost, reflecting the difficulty and risk of the Mars rover climbing or tilting. This safety distance is used to maintain a safe distance between the path and high-risk areas (such as thin shells or deep pits). The closer to the high-risk area, the higher the cost, guiding the path to actively avoid it. , , These are the weighting coefficients, The weight of terrain cost is assigned as the highest weight of 0.6 because it directly determines core passability risks such as subsidence and slippage. The weight of slope cost reflects the impact of slope on passability, second only to terrain, with a weight ratio of 0.3. The weight of the safe distance cost is used as an auxiliary factor to improve path safety, with a weight ratio of 0.1. The toll cost map is as follows: Figure 3 As shown.

[0080] In step S40 above, a global path search is performed based on the grid cost map to obtain the path with the minimum cumulative cost from the starting point to the ending point. The aforementioned raster cost map incorporates terrain attributes such as thin-shell terrain (impassable, cost 100), rocky areas (high-risk impassable based on diameter), and soil-covered areas (mapped using a continuous cost function). The goal of global path planning is to find a path from the starting point S to the ending point G that minimizes the cumulative passability cost, while automatically avoiding high-cost terrain (such as thin-shell terrain, soil-covered terrain, and rocky terrain). The specific method is as follows: (1) Input definition: The starting point S is the current grid coordinates of the Mars rover (provided by inertial navigation and wheel odometry, mapped to the cost map coordinate system), the ending point G is the grid coordinates of the scientific target point or the point specified by the mission command, the cost map M is W×H in size, and each grid m stores the final passability cost calculated in step S30. (Range 0~100, where 100 indicates impassable). For thin-shell terrain grids Marked as an obstacle; for soil-covered areas, the cost varies continuously; for rocky areas; based on the ratio of the equivalent diameter to the wheel diameter, those with a diameter greater than or equal to 1 / 3 of the wheel diameter are marked as impassable, while those with a diameter less than this threshold are given a high cost (71~99).

[0081] (2) A Algorithm initialization: The open list (openList) initially contains the starting point S, and its movement cost is... ,in g The cumulative cost to reach node S, and , Let S be the Euclidean distance from S to G. The estimated total cost for any grid n on the cost map is calculated using... The closed list is initially empty, and records the parent node index for each grid cell.

[0082] (3) Node expansion and cost calculation: Extract the node n with the smallest f value from openList as the current node. If n = G, the path search is complete. Otherwise, move n into closedList. For n's neighboring node m (8-neighborhood, i.e., up, down, left, right, and four diagonal directions), if m is in closedList, skip it; if... If the terrain is covered with thin shells, large rocks, or impassable soil, skip that node (treat it as an obstacle); for soil-covered areas and medium-risk terrain... Calculate the actual cost of traveling from S to m via n:

[0083] in, This represents the geometric distance between adjacent grid cells (1.0m for the straight side and 0.0m for the diagonal). m); This is the motion cost coefficient (ranging from 1.0 to 1.5). Terrain cost influencing factor; The final passability cost of the grid (0~99); This represents the cumulative cost to reach node n.

[0084] Let g(m) denote the current known optimal cumulative cost from the starting point S to node m. Initially, the cumulative cost g(s) of the starting point S is 0, and the cumulative costs of all nodes except the starting point are initialized to infinity. As the algorithm iterates, if m is not in the openList, or if the actual cost g(s) calculated through any adjacent node n is... temp If (m) is less than the currently stored g(m), then use g temp (m) Update g(m), that is This ensures that g(m) always records the current optimal cumulative cost from the starting point to node m. Calculate the heuristic function:

[0085] in, For heuristic function values, , Let x and y be the x and y coordinates of node m; , Let G be the x and y coordinates of the target endpoint G; It is a Euclidean distance.

[0086] calculate Set the parent node to n. If m is not in openList, add it.

[0087] (4) Selection of heuristic function: Euclidean distance multiplied by terrain adaptation coefficient is used to avoid excessive path detours through flat but high-cost areas.

[0088] in: This is a terrain adjustment factor that dynamically increases if high-cost areas appear consecutively ahead of the current path. Coefficients are used to encourage algorithms to deviate from the high-cost band as early as possible; This represents the Euclidean distance from node m to the target point; (5) Detour strategies for thin-shell terrain and special terrain: Since thin-shell terrain is marked as impassable in the cost map ( A The algorithm cannot traverse thin shells. However, for large areas of thin shells, there may be no feasible path. In this case, the algorithm will search near the boundary of the thin shell, utilizing a safe distance cost (in step S30). The guide path maintains a safe distance of at least one grid cell from the thin shell to prevent accidental entry due to positioning errors. Large rocks (impassable) are also directly prohibited from entry by the cost map.

[0089] (6) Path smoothing post-processing: A The algorithm outputs the original path point sequence as follows:

[0090] Perform the following processing: remove redundant points; cubic spline interpolation (interpolate the remaining key points to generate a smooth curvature path, and check whether it falls into an impassable grid during the interpolation process. If it does, tighten the smoothing parameters or revert to the original polyline); maintain a safe distance (the nearest distance between each path point on the smoothed path and the thin shell or large rock is ≥1 grid, i.e. 0.5m).

[0091] (7) Output global path: Output the smoothed global path point sequence ,in This is the current location of the Mars rover. The target point is defined as follows. This sequence contains the geographic coordinates (Mars geographic coordinate system) of each path point and the corresponding accessibility cost (used for local replanning judgment in step S50). The global path is also stored in a path queue for use in subsequent step S60 path tracking control.

[0092] In step S50 above, as the Mars rover travels along the path of minimum cumulative cost, onboard sensors acquire data in real time to update the local cost map in the grid cost map. When an impassable area is detected in the current execution path, local replanning is performed to generate a detour path. Specifically: (1) Local environmental perception and map update: The Mars rover travels at a speed of 0.1 m / s to 0.3 m / s. Every 10 minutes or every 30 minutes, it stops to collect thermal inertia data (using the collection method in step S10 above) to obtain the thermal inertia distribution in the range of 5 m × 5 m to 10 m × 10 m in front. Based on this, the local cost map is updated and merged with the global cost map. (2) Path congestion detection: After each local cost map update, the path points in the current path point sequence located within 5m~10m in front of the rover are projected onto the local cost map. If there is a two-dimensional grid with a passability cost of the preset highest cost value (i.e., thin shell, large rock, or impassable soil cover area), then the path is determined to be congested.

[0093] (3) D The Lite local replanning algorithm: Starting from the rover's current position Slocal, and ending at the first safe point Glocal after the blockage point on the global path, it replans the detour path on the local cost map. Lite employs an incremental search, starting from Glocal and searching backwards, achieving planning through two searches. The search is complete when the rhs and g values ​​in Slocal are equal, and a local path Plocal is generated by backtracking based on the optimal parent node. Here, g represents the currently known shortest cumulative cost from the local endpoint to the current node; rhs represents the minimum estimated cost of the current node calculated by adding the transition cost to the g value of the current node's predecessor node.

[0094] (4) Handling replanning failures: If the path length searched by local replanning exceeds a preset multiple (e.g., 2 times) of the current path length or the search time exceeds a preset duration (e.g., 5 seconds) and no feasible path is found, the replanning is deemed to have failed. At this time, the rover will revert to the "safe position" of the previous record and then re-execute global path planning.

[0095] (5) Replanning execution: The newly generated local path Plocal is concatenated with the untraveled path after Glocal in the global path to form a new complete path, and the target path of the path tracking controller is updated.

[0096] In another embodiment, it also includes: S60. Convert the planned path point sequence into Mars rover motion control commands, and combine the cost map from step S30 with the terrain recognition results from step S20 to achieve adaptive control that adapts to thermal inertia terrain. The specific method is as follows: (1) Path tracking algorithm: A pure tracking algorithm is adopted, and the steering curvature is calculated based on the rover's current pose, the current target point, and the look-ahead distance. The look-ahead distance is determined based on the travel speed. v Adaptive adjustment:

[0097] in, The value range is 1.0 to 2.0. The minimum look-ahead distance is 1m to 2m. When there is high-risk terrain ahead (such as medium-risk areas with a soil cover thickness of >8cm or dense rock areas), the system automatically increases the look-ahead distance (multiplied by a coefficient of 1.2 to 1.5) so that the rover can turn around in advance and avoid entering the danger zone and then making emergency adjustments.

[0098] (2) Control command generation: The default value of the linear velocity command is 0.2 m / s, which is dynamically adjusted according to the cost map attributes of the waypoint. The angular velocity command is calculated based on the path curvature and the current heading deviation, with a maximum angular velocity not exceeding 30° / s. The system dynamically adjusts the linear velocity and steering smoothing parameters based on the thermal inertia inversion results of the grid where the waypoint is located and the terrain type. See Table 2 for details: Table 2: Dynamic Adjustment Parameters for Linear Velocity and Steering Smoothing

[0099] When driving in soil-covered areas, to reduce the risk of shear failure of the soil layer, the angular velocity command is low-pass filtered, and the rate of change of angular velocity is strictly limited to 3° / s²~6° / s² (decreasing with increasing soil layer thickness). When the path curvature is large, the system prioritizes reducing linear velocity rather than increasing angular velocity to reduce the lateral bulldozing effect of the wheels.

[0100] (3) Driving status monitoring and adaptive adjustment: Real-time monitoring of wheel slip rate ,like If the slippage persists for more than 3 seconds, it is considered slippage. The linear velocity is automatically reduced by 20%, and a local perception update is triggered, focusing on checking whether the thermal inertia of the current area is underestimated (e.g., the actual soil cover thickness is greater than the inverted value). The vehicle's pitch angle is monitored in real time, and the deviation between the estimated settlement depth and the actual settlement depth is calculated based on the soil cover thickness inversion results from step S20. If the actual settlement depth exceeds 5cm, the vehicle stops immediately, the soil cover thickness threshold for that grid is lowered (e.g., from 12cm to 8cm), the current area is marked as high-risk, and local replanning in step S50 is triggered. At the same time, the corrected soil cover-cost mapping parameters are stored in the database for subsequent path planning.

[0101] Based on the same inventive concept, embodiments of the present invention also provide a Mars rover path planning system based on multi-source thermal inertia fusion, comprising: The multi-source thermal inertia acquisition and fusion module is used to acquire multi-source thermal inertia data of the target area and perform fusion processing to generate multi-scale thermal inertia distribution data. The terrain recognition and soil cover thickness inversion module is used to perform terrain recognition and soil cover thickness inversion based on multi-scale thermal inertia distribution data and visible light images on the Mars rover. The grid cost map construction module is used to divide the target area into two-dimensional grids, and assign a passability cost to each two-dimensional grid based on the terrain recognition results and the soil cover thickness inversion results to construct a grid cost map. The global path planning module is used to perform global path search based on the grid cost map to obtain the path with the minimum cumulative cost from the starting point to the end point; The local rolling replanning module is used to update the local cost map in the grid cost map by using onboard sensors to acquire data in real time while the Mars rover is traveling along the minimum cumulative cost path. When an impassable area is detected in the current execution path, local replanning is performed to generate a detour path.

[0102] Since the principle behind the problem solved by this system is similar to the aforementioned multi-source thermal inertia fusion method for Mars rover path planning, the implementation of this system can refer to the implementation of the aforementioned method, and the repetitive parts will not be repeated.

[0103] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0104] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A Mars rover path planning method based on multi-source thermal inertia fusion, characterized in that, include: S10. Acquire multi-source thermal inertia data of the target area and perform fusion processing to generate multi-scale thermal inertia distribution data; S20. Based on the multi-scale thermal inertia distribution data, combined with the visible light images on the Mars rover, perform terrain identification and soil cover thickness inversion. S30. Divide the target area into two-dimensional grids, assign a passability cost to each two-dimensional grid based on the terrain recognition results and the soil cover thickness inversion results, and construct a grid cost map. S40. Perform a global path search based on the grid cost map to obtain the path with the minimum cumulative cost from the starting point to the ending point; S50. During the process of the Mars rover traveling along the minimum cumulative cost path, the onboard sensors acquire data in real time and update the local cost map in the grid cost map. When an impassable area is detected in the current execution path, local replanning is performed to generate a detour path.

2. The Mars rover path planning method based on multi-source thermal inertia fusion as described in claim 1, characterized in that, The acquisition and fusion processing of multi-source thermal inertia data of the target region specifically includes: Obtain satellite thermal inertia distribution map based on the thermal infrared imaging system carried by the Mars orbiting satellite, and vehicle thermal inertia distribution map based on the thermal infrared camera on the Mars rover; Unify satellite thermal inertia data and vehicle-mounted thermal inertia data to the Mars geographic coordinate system, and resample all thermal inertia data grid resolution to a uniform resolution; Using the satellite thermal inertia distribution map as the background map, a local fusion window is determined with the rover's current position as the center. Within the local fusion window, for the areas where the satellite thermal inertia distribution map and the rover's thermal inertia distribution map overlap, a weighted average method is used to calculate the fused thermal inertia value. For the areas within the local fusion window that only contain the satellite thermal inertia distribution map, a spatial interpolation method is used to improve the resolution.

3. The Mars rover path planning method based on multi-source thermal inertia fusion as described in claim 1, characterized in that, The terrain identification includes rocky area identification, thin-shell terrain identification, and soil-covered structure identification.

4. The Mars rover path planning method based on multi-source thermal inertia fusion as described in claim 3, characterized in that, The thin-shell terrain identification uses a combination of dynamic thermal inertia features and visible light structural features for comprehensive judgment. The dynamic characteristic criterion of thermal inertia is as follows: based on the multi-scale thermal inertia distribution data and temperature time series, the heating rate of the target region during the heating phase is extracted. and cooling rate during the cooling phase ; Obtain the baseline heating rate of bare soil areas under the same environmental conditions. and reference cooling rate When satisfied and At that time, it was determined to be a candidate for thin-shell terrain; among them This is the threshold coefficient for determining the heating rate; This is the threshold coefficient for determining the cooling rate; The visible light structural feature criteria are: extracting edge warping features, surface crack network features, surface texture features, and color and spectral features from the visible light images on the Mars rover. The thermal inertia dynamic feature criterion and the visible light structural feature criterion are weighted and fused to obtain a thin-shell confidence score; when the thin-shell confidence score is greater than or equal to the confidence threshold, it is determined to be a thin-shell terrain.

5. The Mars rover path planning method based on multi-source thermal inertia fusion as described in claim 1, characterized in that, The soil cover thickness inversion is performed by calculating the soil cover thickness using a pre-calibrated thermal inertia-soil cover thickness mapping relationship. This mapping relationship is established through ground simulation experiments, and corresponding mapping curves are calibrated and stored as curve libraries for different types of fire soil.

6. The Mars rover path planning method based on multi-source thermal inertia fusion as described in claim 1, characterized in that, S20 also includes a wheel settlement discrimination step based on the thickness of the overburden layer, specifically including: Obtain the Martian wheel diameter parameters and calculate the estimated settlement of the wheel within the soil cover layer based on the soil cover thickness. When the thickness of the overburden layer is greater than or equal to the critical overburden layer thickness, and the estimated settlement amount is greater than the safe settlement threshold, it is determined to be a high-risk area; when the thickness of the overburden layer is less than the critical overburden layer thickness, and the estimated settlement amount is less than or equal to the safe settlement threshold, it is determined to be a passable area.

7. The Mars rover path planning method based on multi-source thermal inertia fusion as described in claim 1, characterized in that, The method of assigning a passability cost to each two-dimensional grid based on terrain identification results and overburden thickness inversion results to construct a grid cost map specifically includes: Set the passability cost of rocks in the rocky region with an equivalent diameter greater than or equal to one-third of the Martian wheel diameter, as well as the passability cost of thin-shell terrain, to the preset maximum cost. For areas with soil-covered structures, the thickness of the soil cover layer is mapped to a passability cost through a continuous cost function. The greater the thickness of the soil cover layer, the higher the passability cost. When the thickness of the soil cover layer exceeds the danger threshold, it is set to the preset maximum cost. The final passability cost of each two-dimensional grid is a weighted sum of terrain cost, slope cost, and safety distance cost; wherein, the slope cost is calculated based on the tilt angle of the surface within the two-dimensional grid, and the safety distance cost is used to maintain a preset safety distance between the planned path and high-risk terrain.

8. The Mars rover path planning method based on multi-source thermal inertia fusion as described in claim 1, characterized in that, The global path search uses A Algorithm, the A The algorithm's heuristic function is the Euclidean distance multiplied by a terrain adaptation coefficient; the terrain adaptation coefficient is dynamically adjusted based on the passability cost of the terrain ahead of the path.

9. The Mars rover path planning method based on multi-source thermal inertia fusion as described in claim 1, characterized in that, When an impassable area is detected in the current execution path, local replanning is performed to generate an alternative path, specifically including: After each update of the local cost map, check whether the path points in the current execution path located in front of the rover fall into a two-dimensional grid with a passability cost of the preset maximum cost value. If so, it is determined that there is an impassable area in the current execution path. The local replanning adopts D The Lite algorithm takes the rover's current position as the starting point and the first safe point after the impassable area on the current execution path as the local endpoint, and re-plans the detour path on the local cost map. If the path length of the local replanning search exceeds a preset multiple of the current execution path length or the search time exceeds a preset duration, and no feasible path is still found, the replanning is deemed to have failed. The rover then returns to the previously recorded safe position and re-executes the global path search.

10. A Mars rover path planning system with multi-source thermal inertia fusion, characterized in that, The system comprising the method of any one of claims 1-9, wherein the method is: The multi-source thermal inertia acquisition and fusion module is used to acquire multi-source thermal inertia data of the target area and perform fusion processing to generate multi-scale thermal inertia distribution data. The terrain identification and soil cover thickness inversion module is used to perform terrain identification and soil cover thickness inversion based on the multi-scale thermal inertia distribution data and the visible light images on the Mars rover. The grid cost map construction module is used to divide the target area into two-dimensional grids, assign a passability cost to each two-dimensional grid based on the terrain recognition results and the soil cover thickness inversion results, and construct a grid cost map. The global path planning module is used to perform a global path search based on the grid cost map to obtain the path with the minimum cumulative cost from the starting point to the ending point. The local rolling replanning module is used to acquire data in real time using onboard sensors during the rover's journey along the minimum cumulative cost path, update the local cost map in the grid cost map, and perform local replanning to generate a detour path when an impassable area is detected in the current execution path.

Images