Polar region ship path planning simulation platform and system based on ASVSim and visual technology

Abstract

The invention belongs to the field of ship path planning simulation, discloses a polar region ship path planning simulation platform and system based on ASVSim and visual technology, and provides a simulation platform which comprises a polar region environment module, a ship model module, a sensor module, a control decision module and a physical simulation module. The polar region environment module is used for modeling a polar region environment by utilizing a three-dimensional rendering engine; the ship model module provides a ship dynamics model; the sensor module simulates various sensors; the control decision module provides an algorithm interaction API; and the physical simulation module integrates data to form a closed loop. The invention further provides a path planning system which comprises a sensing module, a data processing module and a path planning module. According to the invention, the natural environment of polar ship navigation can be simulated vividly, and abundant data sources and reliable test support are provided for verification of a path planning system.

Description

Technical Field

[0001] This invention belongs to the field of ship path planning simulation, specifically relating to a polar ship path planning simulation platform and system based on ASVSim and vision technology. Background Technology

[0002] As the strategic value of the Arctic shipping routes grows increasingly important, planning safe and efficient routes for ships traversing complex ice zones has become a core technology. Developing advanced and reliable route planning and obstacle avoidance systems is crucial for achieving autonomous navigation, and the effectiveness of such systems must be guaranteed through thorough testing and verification. However, the verification process faces significant bottlenecks. Conducting real-world ship trials in the actual polar environment is not only extremely costly, carries enormous environmental risks, and presents extremely complex logistical support, but the rapidly changing sea ice distribution makes it difficult to reproduce experimental conditions, hindering comparative testing and iterative optimization of different systems. This severely restricts the development of intelligent technologies for polar vessels.

[0003] Therefore, using simulation technology for system testing has become an industry consensus and is the most effective way to solve the aforementioned dilemmas. However, existing general-purpose autonomous driving simulation platforms (such as the open-source AirSim) are mainly geared towards land vehicle scenarios and generally lack high-fidelity modeling of specific polar environmental factors, such as the physical properties of sea ice, the collision dynamics between the ship and sea ice, and the unpredictable polar weather. This distortion of the simulation environment leads to a significant gap between system test results and actual applications, failing to provide reliable verification support for the system's effectiveness. Therefore, there is an urgent need in this field for a dedicated simulation platform capable of realistically simulating polar navigation scenarios, and on this basis, to develop a corresponding efficient path planning system to promote the research and application of polar ship path planning technology in a safe and low-cost manner. Summary of the Invention

[0004] To address the technical problems of existing simulation platforms' inaccurate modeling of polar environments, inability to provide reliable system verification, and high costs, risks, and difficulty in reproducing conditions during actual ship trials, this invention provides a polar ship path planning simulation platform and system based on ASVSim and vision technology. This platform can realistically simulate the natural environment of polar ship navigation, providing abundant data sources and reliable testing support for the verification of the path planning system.

[0005] To achieve the above objectives, the present invention provides the following solution: A polar vessel path planning simulation platform based on ASVSim and vision technology, the platform includes: a polar environment module, a vessel model module, a sensor module, a control decision module, and a physical simulation module; The polar environment module is used to perform three-dimensional rendering and modeling of polar environment elements using a three-dimensional rendering engine. The ship model module is used for ship dynamics models that drive ship motion and for modeling the shape of research vessels. The sensor module is used to collect data from the polar environment module in real time; The control decision module is used to receive data from the sensor module and send control commands given by the decision algorithm to the ship model module; The physical simulation module is used to integrate data from the polar environment module, the ship model module, the sensor module, and the control decision module to perform physical simulation and form a closed loop.

[0006] Preferably, the polar environment module uses Unreal Engine 5 to perform three-dimensional rendering modeling of floating ice, icebergs and seawater, and simulates the physical collision characteristics of floating ice and icebergs; Specifically: For polar ice floes, instead of creating static meshes, parametric modeling is performed through the UE5 physics engine interface, assigning density and mass attributes to each ice floe instance; a simplified multi-convex hull collider is used to approximate the complex outer contour of the iceberg to balance computational overhead with collision accuracy in path planning; for seawater modeling, fluid simulation based on the Gerstner wave function is used, with a regional vector field superimposed, and each vertex in the wave morphology is... Position over time The changes follow: ; in, The direction of the i-th wave component. For amplitude, For wave number, For location, time, Angular frequency, This is the initial phase.

[0007] Preferably, the ship model module includes an exterior blueprint and a dynamic model; The exterior blueprints were based on the Xue Long polar research vessel, which was used to construct a three-dimensional model of the ship; the dynamic model was provided by the ASVSim plugin.

[0008] Preferably, the sensor module includes an RGB camera, barometer, IMU, GPS, distance sensor, and radar provided by the ASVSim plugin.

[0009] Preferably, the control decision module includes a decision algorithm and a ship control API. The decision algorithm is used to calculate based on the received environmental data and ultimately generate decision commands including the ship's rudder angle and horsepower. The ship control API is used to enable real-time interaction between the Python-based external algorithm environment and the 3D rendering engine environment.

[0010] The present invention also provides a polar vessel path planning system based on ASVSim and vision technology. The system is implemented through the aforementioned platform and includes: a perception module, a data processing module, and a path planning module. The sensing module is used to acquire environmental sensing data and ship status data using the sensor module; The data processing module is used to process the environmental perception data to generate an obstacle map; The path planning module is used to calculate the ship's control strategy based on the obstacle map and the ship's status data.

[0011] Preferably, the process of acquiring environmental perception data and ship status data using sensor modules includes: The system uses a forward-looking RGB camera to acquire local RGB images and radar to acquire depth information; a top-view RGB camera to acquire global RGB images; and IMU and GPS to acquire the ship's positioning and attitude.

[0012] Preferably, the process of processing the environmental perception data to generate an obstacle map includes: The local RGB image is input into an image segmentation model to obtain a segmentation map, thereby obtaining the pixel index of the obstacle in the image coordinate system. The image segmentation model is a zero-shot segmentation model. For pixel index ( , ), query the synchronously acquired depth information, and obtain the depth value. Obtain the intrinsic parameter matrix of the forward-looking RGB camera. and extrinsic transformation matrix Using the aforementioned depth value and the intrinsic parameter matrix The pixel index ( , Transform from the image coordinate system to the camera coordinate system to obtain the camera coordinate points. ; Using the aforementioned extrinsic transformation matrix The camera coordinate points Transform to world coordinate system to obtain 3D points ; and combining multiple of the three-dimensional points To form obstacle point clouds; The obstacle point cloud is orthogonally projected onto a preset horizontal plane to generate a raster-style local obstacle map.

[0013] Preferably, the process of calculating the ship's control strategy based on the obstacle map and the ship's status data includes: The Dynamic Window Method (DWA) is adopted, taking the local obstacle map, the global map generated from the global RGB image, and the ship's attitude, speed, and dynamic constraints as inputs, and outputting the rudder angle and horsepower of the ship. Then, it is transmitted to the UE terminal through the ship control API to complete real-time closed-loop control.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This polar vessel route planning simulation platform is based on UE's powerful rendering capabilities and realistically simulates the polar ice and snow environment, including controllable weather, sea ice and glaciers with physical collisions, and flowing seawater, which can realistically simulate the natural environment of polar vessel navigation.

[0015] 2. This polar vessel path planning simulation platform is based on the ASVSim plugin, integrates a variety of sensors for testing, and provides the external blueprint of the polar research vessel. By calling the ASVSim API, sensor data can be read in real time and the vessel's power unit can be controlled, providing a rich source of data for the verification of the vessel path planning system.

[0016] 3. Based on the constructed polar ship path planning simulation platform, this invention proposes a path planning system for ship navigation in polar environments, providing an example for researchers using this simulation platform and related path planning. Attached Figure Description

[0017] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a diagram showing the polar environment module of the polar vessel path planning simulation platform according to an embodiment of the present invention.

[0019] Figure 2 This is a diagram showing the contents of the ship model module of the polar ship path planning simulation platform according to an embodiment of the present invention.

[0020] Figure 3 This is a system interaction diagram of the sensor module and control decision module of the polar ship path planning simulation platform according to an embodiment of the present invention.

[0021] Figure 4 This is a system flowchart of the physical simulation module of the polar ship path planning simulation platform according to an embodiment of the present invention.

[0022] Figure 5 This is a flowchart of a path planning implementation system of the polar ship path planning simulation platform according to an embodiment of the present invention.

[0023] Figure 6 This is a polar simulation environment and ship modeling rendering diagram of the polar ship path planning simulation platform according to an embodiment of the present invention.

[0024] Figure 7 This is a visualization of the final effect of a path planning implementation system of a polar ship path planning simulation platform according to an embodiment of the present invention. Detailed Implementation

[0025] 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.

[0026] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0027] This invention provides a polar vessel path planning simulation platform based on ASVSim. This platform aims to provide a high-fidelity, low-cost, and repeatable testing and verification environment for polar vessel path planning systems. The platform's overall architecture consists of five core modules: a polar environment module, a vessel model module, a sensor module, a control decision module, and a physical simulation module.

[0028] The polar environment module, such as Figure 1 As shown, the polar environment module in this embodiment is responsible for constructing and rendering a highly realistic polar navigation scene. Specifically, this platform utilizes the powerful 3D rendering capabilities of Unreal Engine 5 to model key environmental elements unique to the polar regions. These elements include: floating ice and icebergs with physical collision volumes, dynamically flowing seawater, and a variable weather system incorporating effects such as sky, clouds, and snowflakes, providing a near-realistic visual and physical environment for testing the ship path planning system. The effect is as follows... Figure 7As shown. Specifically: For polar ice floes, this application does not simply create static meshes, but performs parametric modeling through the UE5 physics engine interface. Density and mass attributes are assigned to each ice floe instance. A simplified multi-convex hull collider is used to approximate the complex outer contour of the iceberg to balance computational overhead with collision accuracy in path planning. For seawater modeling, fluid simulation based on the Gerstner wave function is used, with a superimposed regional vector field. The wave morphology formula is as follows, for each vertex... Location Over time The changes follow: in The direction of the i-th wave component. For amplitude, For wave number, Angular frequency, This is the initial phase. The variable weather system is based on the variable weather module provided by the ASVSim plugin.

[0029] The ship model module, as shown Figure 2 As shown, the ship model module in this embodiment is used to define and drive the ship as a carrier in the simulation environment. This module mainly includes a static mesh of the exterior blueprint, the mounting of relevant sensors, and the ship's dynamics and kinematics model. This platform references publicly available information from real polar scientific research vessels such as the "Xuelong" to construct a three-dimensional static mesh model of the ship, which is then rendered in UE5. Figure 6 As shown, the ship's motion behavior in the simulation environment is driven by the ship dynamics and kinematics model provided by the ASVSim plugin. This model is responsible for calculating the changes in the ship's attitude, speed, etc., when it receives control commands and is affected by the environment (such as water flow, ice collisions). The model also defines the mounting locations of various sensors.

[0030] The sensor module and control decision module are as follows: Figure 3 As shown, the sensor module and the control decision module together form the data perception and system decision-making closed loop of the simulation platform. The sensor module runs on the UE (User Equipment) and integrates various virtual sensors provided by the ASVSim plugin to simulate the ship's perception capabilities of its surrounding environment. These sensors include RGB cameras for visual information, radar for depth information, GPS and IMU for position and attitude information, and distance sensors, etc. During simulation, these sensors collect data from the virtual environment in real time. The control decision module typically runs in a Python environment outside the UE environment, such as... Figure 3The PyCharm terminal receives data from sensor modules via the sensor API provided by ASVSim. The path planning system calculates based on the received environmental data, ultimately generating decision commands including the ship's rudder angle and horsepower (this is a simulation platform that provides a system framework; users can verify their own path planning algorithms, so specifying the exact algorithm is unnecessary; a preferred embodiment follows). These commands are then sent back to the ship model on the UE terminal via the ship control API provided by ASVSim, thereby controlling the ship's navigation.

[0031] The physical simulation module is based on the ASVSim plugin and is used to integrate data from the external environment, ship model, sensors, etc., to perform physical simulation and output simulation results.

[0032] It should be explained that ASVSim is a simulation plugin for ships and other surface vehicles, based on the Unreal Engine UE5. ASVSim's advantage lies in its integration of specialized ship dynamics and kinematic models, as well as various commonly used navigation sensor models (such as RGB cameras, radar, GPS, and IMU), on top of Unreal Engine's powerful rendering capabilities. Furthermore, ASVSim provides a complete set of application programming interfaces (APIs), supporting real-time reading and control of ship status and sensor data within the simulation world from external development environments such as Python. This provides an efficient and convenient interactive channel for realizing closed-loop verification of complex "perception-decision-control" systems.

[0033] Specifically, refer to Figure 4 The physical simulation module performs unified simulation calculations and generates simulation results based on information from the polar environment module, ship model module, sensor module, and control decision module. The physical simulation module can calculate the ship's motion state and generate its navigation trajectory based on environmental parameters from the polar environment module (such as water pressure and sea ice collisions) and dynamic parameters from the ship model module. Furthermore, the physical simulation module can drive the sensor module to collect data based on the ship's real-time motion state and the environmental model, generating the ship's perception data. Further, the physical simulation module receives decision data (such as rudder angle and horsepower commands) from the control decision module and uses it as input for the next simulation moment to calculate the ship's control results and generate updated ship motion state data, thus forming a complete simulation closed loop.

[0034] It should be understood that the simulation platform constructed above provides a basic environment for the development and verification of various polar vessel path planning systems. To illustrate the practical application of the aforementioned polar vessel path planning simulation platform and provide a feasible technical solution, this invention further discloses a path planning implementation system based on this platform, as a preferred embodiment of the invention.

[0035] It should be explained that this is a vision-based ship path planning system, including: a perception module, a data processing module, and a path planning module. The flowchart of the specific system implementation is as follows: Figure 5 As shown in the flowchart, the implementation details will be explained in detail below.

[0036] The perception module uses an RGB camera, radar, IMU, and GPS sensors. Data from the UE-side sensors is read into the PyCharm data processing module via the ASVSim sensor API. This includes obtaining RGB images from an RGB camera positioned 100 meters above the ship in the environment; obtaining RGB images and camera pose from an onboard RGB camera in the ship's direction of travel; obtaining depth information from radar; and obtaining ship status (pose, speed, etc.) from the IMU and GPS.

[0037] The data processing module processes the input perception data and converts it into a global map and a local obstacle map, which are then output to the DWA system for path planning.

[0038] Specifically, such as Figure 5 As shown, the global map is obtained from RGB images output by an aerial RGB camera. Obstacles such as glaciers and ice floes are predominantly white, while seawater is predominantly bluish-black. For simplicity, obstacles can be directly distinguished by color. The pixel value range is 0-255. A threshold of 100 is set; values ​​above 100 are considered obstacles, and values ​​below 100 are considered feasible areas, thus obtaining a global obstacle map.

[0039] It should be explained that the global map only performs global planning and does not require a detailed obstacle map. If the obstacles in the test environment are complex and difficult to distinguish simply by color, the global RGB image can be input into the segmentation model to obtain the segmentation map of the obstacles, and then a raster map can be constructed to obtain the global obstacle map.

[0040] Specifically, such as Figure 5 As shown, the local obstacle map construction process aims to accurately convert the two-dimensional image information acquired from the sensor into a physical obstacle distribution in the world coordinate system, and then project it to obtain an obstacle map. The construction process is divided into image segmentation and obstacle pixel indexing, camera parameter acquisition, coordinate system transformation and 3D point cloud generation, and projection to generate a local obstacle map. Figure 4 Each part.

[0041] The image segmentation and obstacle pixel indexing process begins by inputting the shipborne forward-looking RGB image acquired by the perception module into a pre-trained image segmentation model. In this embodiment, a model with strong zero-shot segmentation capabilities, such as the Segment Anything Model (SAM), is preferably used to adapt to the variable icy and snowy environment of the polar regions.

[0042] Furthermore, the segmentation model processes the input RGB image and outputs a segmentation map. In this segmentation map, each pixel is assigned a category label (e.g., ice floes, icebergs, open water, sky, etc.), treating ice floes and icebergs as obstacles and open water as non-obstacles. Thus, a set of indices for all pixels belonging to the obstacle category in the image coordinate system (u, v) can be obtained.

[0043] The camera parameter acquisition includes the camera intrinsic parameter matrix. In this embodiment, the external parameter matrix T is obtained by dynamic calculation based on the preset attributes of the camera in the simulation environment and the real-time camera pose.

[0044] Furthermore, the camera intrinsic parameter matrix The projection characteristics of the camera itself are described. In this embodiment, the camera intrinsic parameters are calculated based on the image sensor size and the camera's horizontal field of view (FOV). Specifically, given the image width (img_width), image height (img_height), and the camera's horizontal field of view (fov) obtained from the simulation environment, the elements of the intrinsic parameter matrix K are calculated as follows: Principal coordinates ( The center of the image is usually assumed to be: focal length and They are equal, and the formula for calculation is: Thus, a complete camera intrinsic parameter matrix is ​​constructed. : Furthermore, the camera extrinsic transformation matrix This describes the position and pose of the camera coordinate system relative to the world coordinate system, used to implement point transformations between the two coordinate systems. In this embodiment, this matrix is ​​constructed based on the real-time pose of the camera (camera_pose) in the simulation environment. (Extrinsic parameter matrix) The calculations for each element are as follows: First, extract the camera's position vector in the world coordinate system from camera_pose. and attitude quaternions The camera's coordinates along the X, Y, and Z axes in the world coordinate system are respectively... The attitude quaternion scalar (real part) is The vector (imaginary part) is : Then, the attitude quaternion Convert to a 3×3 rotation matrix This matrix represents the rotation transformation from the camera coordinate system to the world coordinate system.

[0045] Finally, combining the rotation matrix And position vector (i.e., translation vector) Construct a 4×4 homogeneous transformation matrix This matrix is ​​the complete extrinsic transformation matrix. Its form is as follows, where... This represents the projection of the j-th axis of the camera coordinate system onto the i-th axis of the world coordinate system. The coordinate system transformation and 3D point cloud generation are the core of realizing the transformation from 2D image to 3D space. For each obstacle pixel index obtained in the aforementioned steps ( , Perform the following coordinate transformation process: First, query the pixel coordinates from the depth image acquired synchronously with the RGB image. , The corresponding depth value This depth value This represents the distance of the pixel along the Z-axis of the camera coordinate system.

[0046] Secondly, the pixel coordinates ( , ) and its depth value Transform from image coordinate system to camera coordinate system. This is achieved using the camera intrinsic parameter matrix. The parameters defined in the code are used to calculate the three-dimensional coordinates of the point in the camera coordinate system. The conversion formula is: in, Internal parameter matrix The inverse matrix.

[0047] Then, the points in the camera coordinate system Transform to world coordinate system to obtain 3D points This process involves left-multiplying the camera extrinsic transformation matrix. To achieve this, matrix multiplication in homogeneous coordinates requires first setting the camera coordinates. Augmented to homogeneous coordinate form The conversion formula is as follows: By repeatedly performing the above coordinate transformation process on all identified obstacle pixels, a set of three-dimensional points that accurately describes the shape and position of the obstacle surface in the world coordinate system can be obtained, which is the final obstacle point cloud.

[0048] The projection-generated local obstacle map converts a 3D obstacle point cloud into a raster map that can be directly used for 2D path planning. Specifically, it transforms each 3D point in the point cloud into a raster map. Orthogonally project onto a preset horizontal plane (usually sea level with Z=0), ignoring its height component. A set of two-dimensional coordinate points is obtained. Finally, these 2D projection points are filled into a 2D raster map with a preset size and resolution. All rasters occupied by projection points are marked as "obstacles," while other rasters are marked as "passable." This completes the construction of a local obstacle map for one frame, which can be updated in real time and used as input to the DWA algorithm.

[0049] The path planning module is responsible for calculating the optimal navigation control commands in real time, including the ship's rudder angle and horsepower, based on the obstacle map generated by the data processing module and the ship's own state obtained by the perception module. Then, it uses ASVSim's ship control API to perform closed-loop control on the ship in the simulation platform.

[0050] Specifically, in this embodiment, the path planning module employs the Dynamic Window Method (DWA algorithm) as the local path planning and obstacle avoidance algorithm. The algorithm's inputs are the global map, the local obstacle map, and the ship's state, and its outputs are the linear velocity and angular velocity corresponding to the optimal trajectory. Subsequently, the linear velocity and angular velocity are converted into the corresponding rudder angle and power for the ship to execute.

[0051] The global map is used to provide macroscopic target guidance for the DWA algorithm, ensuring that the ship can navigate toward the final goal while avoiding local obstacles.

[0052] The local obstacle map is the main basis for the DWA algorithm to perform collision detection and trajectory evaluation.

[0053] The ship's state includes, but is not limited to, its position, attitude, linear velocity, and angular velocity. Furthermore, inherent dynamic constraints such as the ship's physical dimensions and maximum turning radius are also used as input to ensure that the generated trajectory is actually executable by the ship.

[0054] It's important to explain that the core of the DWA algorithm lies in trajectory sampling and evaluation within the velocity space. First, based on the input ship's current speed and dynamic constraints, the algorithm generates a dynamic window consisting of multiple pairs of linear and angular velocity combinations. Then, for each pair of velocity commands within the window, the algorithm performs trajectory deduction within a short time step, simulating multiple candidate navigation trajectories. Next, all candidate trajectories are scored using a pre-defined evaluation function. This function typically considers three aspects: target orientation (the consistency between the trajectory's endpoint and the global path direction); obstacle avoidance safety (the distance between the trajectory and the nearest obstacle in the local obstacle map); and navigation speed (the magnitude of the linear velocity corresponding to the trajectory), encouraging efficient navigation while ensuring safety. Finally, the algorithm selects the trajectory with the highest overall score and uses the corresponding linear and angular velocity commands as the optimal result for this decision.

[0055] The optimal speed command is translated into specific ship control strategies, which in this embodiment are the ship's rudder angle and horsepower values. These control strategies are sent in real-time from the Python development environment to the ship model in the UE simulation environment via the ship control API provided by the ASVSim plugin. Upon receiving the commands, the ship model executes the corresponding actions, thus completing a full closed-loop iteration from environmental perception and intelligent decision-making to platform control, driving the ship to navigate safely and autonomously in the complex polar ice environment. The final effect is visualized as follows: Figure 7 As shown.

[0056] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A polar vessel path planning simulation platform based on ASVSim and vision technology, characterized in that, The platform includes: a polar environment module, a ship model module, a sensor module, a control decision module, and a physical simulation module; The polar environment module is used to perform three-dimensional rendering and modeling of polar environment elements using a three-dimensional rendering engine. The ship model module is used for ship dynamics models that drive ship motion and for modeling the shape of research vessels. The sensor module is used to collect data from the polar environment module in real time; The control decision module is used to receive data from the sensor module and send control commands given by the decision algorithm to the ship model module; The physical simulation module is used to integrate data from the polar environment module, the ship model module, the sensor module, and the control decision module to perform physical simulation and form a closed loop.

2. The platform according to claim 1, characterized in that, The polar environment module uses Unreal Engine 5 to perform 3D rendering and modeling of floating ice, icebergs and seawater, and simulates the physical collision characteristics of floating ice and icebergs. Specifically: For polar ice floes, instead of creating static meshes, parametric modeling is performed through the UE5 physics engine interface, assigning density and mass attributes to each ice floe instance; a simplified multi-convex hull collider is used to approximate the complex outer contour of the iceberg to balance computational overhead with collision accuracy in path planning; for seawater modeling, fluid simulation based on the Gerstner wave function is used, with a regional vector field superimposed, and each vertex in the wave morphology is... Position over time The changes follow: ； in, The direction of the i-th wave component. For amplitude, For wave number, For location, time, Angular frequency, This is the initial phase.

3. The platform according to claim 1, characterized in that, The ship model module includes an exterior blueprint and a dynamic model; The exterior blueprints were based on the Xue Long polar research vessel, which was used to construct a three-dimensional model of the ship; the dynamic model was provided by the ASVSim plugin.

4. The platform according to claim 1, characterized in that, The sensor module includes an RGB camera, barometer, IMU, GPS, distance sensor, and radar provided by the ASVSim plugin.

5. The platform according to claim 1, characterized in that, The control decision module includes a decision algorithm and a ship control API. The decision algorithm is used to calculate based on the received environmental data and ultimately generate decision commands including the ship's rudder angle and horsepower. The ship control API is used to enable real-time interaction between the Python-based external algorithm environment and the 3D rendering engine environment.

6. A polar vessel path planning system based on ASVSim and vision technology, wherein the system is implemented using the platform described in any one of claims 1-5, characterized in that, The system includes: a perception module, a data processing module, and a path planning module; The sensing module is used to acquire environmental sensing data and ship status data using the sensor module; The data processing module is used to process the environmental perception data to generate an obstacle map; The path planning module is used to calculate the ship's control strategy based on the obstacle map and the ship's status data.

7. The system according to claim 6, characterized in that, The process of acquiring environmental perception data and ship status data using sensor modules includes: The system uses a forward-looking RGB camera to acquire local RGB images and radar to acquire depth information; a top-view RGB camera to acquire global RGB images; and IMU and GPS to acquire the ship's positioning and attitude.

8. The system according to claim 7, characterized in that, The process of processing the environmental perception data to generate an obstacle map includes: The local RGB image is input into an image segmentation model to obtain a segmentation map, so as to obtain the pixel index of the obstacle in the image coordinate system. The image segmentation model is a zero-shot segmentation model. For pixel index ( , ), query the synchronously acquired depth information, and obtain the depth value. Obtain the intrinsic parameter matrix of the forward-looking RGB camera. and extrinsic transformation matrix Using the depth value and the intrinsic parameter matrix The pixel index ( , Transform from the image coordinate system to the camera coordinate system to obtain the camera coordinate points. ; Using the aforementioned extrinsic transformation matrix The camera coordinate points Transform to world coordinate system to obtain 3D points ; and combining multiple of the three-dimensional points To form obstacle point clouds; The obstacle point cloud is orthogonally projected onto a preset horizontal plane to generate a raster-style local obstacle map.

9. The system according to claim 8, characterized in that, The process of calculating the ship's control strategy based on the obstacle map and the ship's status data includes: The Dynamic Window Method (DWA) is adopted, taking the local obstacle map, the global map generated from the global RGB image, and the ship's attitude, speed, and dynamic constraints as inputs, and outputting the rudder angle and horsepower of the ship. Then, it is transmitted to the UE terminal through the ship control API to complete real-time closed-loop control.

Images