Gantry-mounted dual robotic arms and their path planning methods and systems under radiation environments

Abstract

This invention discloses a path planning method for a gantry-supported dual-arm robotic system in a radiation environment, comprising: discretizing a continuous radiation field into three-dimensional radiation voxels; establishing environmental obstacles, the main robotic arm, slave robotic arm, and gantry structure of the gantry-supported dual-arm system as collision bodies; constructing a state subspace based on the configuration vectors of the main robotic arm and the gantry structure, planning a main path in the state subspace, and obtaining a main path point sequence; temporalizing the main path point sequence to obtain a queryable trajectory in the subspace; constructing an extended space based on the configuration vector and time of the slave robotic arm, planning a slave path in the extended space, and obtaining an extended space trajectory; and synthesizing the queryable trajectory in the subspace and the extended space trajectory on a unified time axis to form a collaborative trajectory for both arms. This invention achieves decoupled collaborative planning for a gantry-supported dual-arm system in a radiation environment, ensuring geometric collision avoidance and radiation safety constraints.

Description

Technical Field

[0001] This invention relates to the field of robot path planning technology, and in particular to a gantry-mounted dual robotic arm and its path planning method and system in a radiation environment. Background Technology

[0002] Nuclear decommissioning is a critical step in ensuring nuclear safety, encompassing high-risk tasks such as the dismantling of radioactive equipment, spent fuel reprocessing, and contaminated area cleanup. Utilizing autonomous specialized robots to replace human labor in these high-risk tasks has become an inevitable trend. Among these, the gantry-mounted dual-arm robotic system combines the wide-range mobility of a gantry crane with the flexibility of dual robotic arms, demonstrating significant engineering advantages in scenarios such as waste bin handling, large pipe cutting, and heavy object dismantling.

[0003] Path planning, a key technology for autonomous operation of gantry-type dual-arm robotic systems, determines the system's trajectory in environments coupled with radiation fields and complex obstacles, impacting operational safety and radiation exposure levels. Existing technologies, within a two-dimensional radiation cost grid modeling framework, include improved particle swarm optimization, GB-RRT*, and DL-RRT* algorithms for low-dose path searching in nuclear facility radiation environments. However, these studies primarily focus on personnel and mobile robots and revolve around two-dimensional ground navigation, making direct transfer to three-dimensional collaborative planning problems in high-dimensional configuration spaces. In the field of high-dimensional robotic arm planning, sampling-based RRT algorithms and their variants dominate. However, existing results largely focus on radiation obstacle avoidance for two-dimensional mobile platforms or geometrical collaboration of dual arms in conventional environments, lacking cumulative radiation dose calculation models that match the three-dimensional motion characteristics of robotic arms. Furthermore, when dealing with high-dimensional configuration space planning, traditional random sampling strategies struggle to simultaneously ensure obstacle avoidance efficiency and minimize radiation dose, resulting in slow algorithm convergence, high exposure risk to critical components, and difficulty meeting the operational requirements of complex nuclear decommissioning scenarios. Summary of the Invention

[0004] To address the issues of insufficient environmental radiation characterization, lack of protection for critical joint components, and low sampling convergence efficiency in the existing technologies, this invention provides a gantry-supported dual robotic arm and its path planning method and system under radiation conditions. This method integrates radiation field constraints and a hybrid sampling strategy, achieving multi-objective collaborative optimization of obstacle avoidance safety, path length, and radiation protection.

[0005] A path planning method for a gantry-mounted dual robotic arm in a radiation environment includes the following steps: Discretize the continuous radiation field into three-dimensional radiation voxels; The environmental obstacles, the main robotic arm, the slave robotic arm, and the gantry structure of the dual robotic arm system are established as the collision body; A state subspace is constructed based on the configuration vectors of the main robotic arm and the gantry structure. A main path is planned within this subspace. Radiation exclusion constraints are verified using the 3D radiating voxels, and geometric collision avoidance constraints are verified using the colliders. A main path point sequence is obtained. This main path point sequence is temporalized to obtain a queryable trajectory in the subspace. An extended space is constructed based on the configuration vectors and time of the secondary robotic arm. Secondary paths are planned within this extended space. Geometric collision avoidance constraints are verified using the queryable trajectory and the colliders. Radiation exclusion constraints are verified using the 3D radiating voxels, and the extended space trajectory is obtained. The queryable trajectory in the subspace and the extended space trajectory are then synthesized on a unified time axis to form a dual-arm cooperative trajectory.

[0006] Furthermore, the discretization of the continuous radiation field into three-dimensional radiation voxels includes: The three-dimensional mesh side length is determined based on the minimum joint motor equivalent diameter and safety factor in the system, wherein the mesh side length is less than or equal to the ratio of the minimum joint motor equivalent diameter to the safety factor; Based on the location of the radiation source, the inverse square law of distance attenuation, the dielectric exponential attenuation term, and the direction factor of the most favorable receiving surface, the radiation intensity at the center of the voxel is calculated as the numerical attribute of the radiation voxel.

[0007] Furthermore, the establishment of environmental obstacles, the main robotic arm, the slave robotic arm, and the gantry structure as collision bodies in the gantry-supported dual robotic arm system includes: Each link of the main robotic arm and the slave robotic arm is approximated as a capsule, the gantry structure is represented by multiple cuboids, and the environmental obstacles are represented by spheres aligned with axes and enclosing boxes. Radiation voxels whose dose rates exceed the radiation exclusion threshold are included in the exclusion set and used as environmental barriers in edge validation. A safety margin that adapts to the radiation level is set, and the safety margin increases as the radiation dose rate increases.

[0008] Furthermore, the geometric collision avoidance constraints are verified based on the collider, including the use of hierarchical dynamic collision detection: Wide-phase screening: A static axis-aligned bounding box hierarchical tree is established for environmental obstacles. The bounding boxes of each collider are calculated for discrete configurations, and potential intersecting candidate obstacles are queried in the hierarchical tree. For dynamic collisions between master and slave robotic arms, the overlapping area point set of the workspace of the two arms is pre-calculated. An axis-aligned bounding box is constructed for the overlapping area point set as a trigger area. When the sampling point on the axis of the connecting capsule enters the trigger area, the narrow-phase collision fine detection between the arms is triggered. Narrow phase precision detection: Determines whether colliding bodies intersect based on the geometric distance between them; Edge continuity verification: The planned edges are interpolated using an adaptive discrete step count, and the wide-phase screening and narrow-phase fine-tuning are performed for each discrete configuration.

[0009] Furthermore, the radiation forbidden constraint is verified based on the three-dimensional radiation voxel, including: Wide phase no-entry screening: Voxel index mapping is performed on the center point of the joint motor, and neighborhood expansion is performed by combining the equivalent radius of the motor and the adaptive safety margin. The conservative dose rate in the neighborhood voxel set is queried. If the dose rate threshold is exceeded, it is determined to be infeasible. Narrow-phase cumulative dose calculation: For configurations that pass the wide-phase no-entry screening, the spatial intersection of the articulated motor cylinder and voxels is verified based on the separated axis theorem. The radiation intensity of the intersecting voxels is weighted, the weighted equivalent dose rate of the articulated motor is calculated, and the cumulative radiation dose is obtained by integrating along the path.

[0010] Furthermore, the temporalization of the main path point list and the planning of secondary paths in the extended space include: Based on the maximum speed of the gantry and the maximum angular velocity of the master and slave robotic arm joints, calculate the shortest feasible execution time for the edge formed by adjacent path points of the main path, and define a uniform time step to achieve consistency between geometric discrete points and time discrete points. In the extended space, the main arm is regarded as a dynamic obstacle that changes over time. The state of the slave arm is queried through the configuration of the main arm over time to perform dynamic inter-arm collision detection. When the slave arm cannot cross the area occupied by the main arm within a specific time period, the slave arm is allowed to take a waiting action, that is, advance time while maintaining its posture.

[0011] Furthermore, path planning in the state subspace and the extended space employs a hybrid sampling strategy based on radiation field information, including: Perform uniform sampling within the feasible boundary with a first probability; The target region is biased and sampled with a second probability; Low-radiation area guided sampling is performed with a third probability: uniform random sampling points are generated and filled into the candidate pool after passing the radiation exclusion judgment. The radiation guidance index of the candidate samples is calculated and sorted in ascending order. One sample is randomly selected from the set of the top K samples as the guided sampling output. The radiation guidance index is the sum of the joint motor dose rates.

[0012] Furthermore, during the path planning process, a multi-objective cost function is used for parent selection and reconnection in the RRT* algorithm: The cost of each edge is defined as the weighted sum of the cumulative radiation dose of all joint motors on that edge and the path length in the mission space. During the parent selection phase, the total cost of the path through the candidate parent node in the neighborhood is compared, and the node with the lowest cost is selected as the parent node. During the reconnection phase, it is determined whether the total cost can be reduced by going through the current node to the neighboring nodes. If so, reconnection is performed.

[0013] A path planning system for a gantry-mounted dual robotic arm in a radiation environment, used to implement the aforementioned path planning method, includes: The radiation modeling module is used to discretize a continuous radiation field into three-dimensional radiation voxels; The collision modeling module is used to create collision bodies from environmental obstacles, the main robotic arm, the slave robotic arm, and the gantry structure of the gantry-supported dual robotic arm system. The main arm planning module is used to construct a state subspace based on the configuration vectors of the main robotic arm and the gantry structure, plan the main path in the state subspace, verify the radiation prohibition constraint based on the three-dimensional radiation voxels, verify the geometric collision avoidance constraint based on the collider, and obtain the main path point sequence; the main path point sequence is temporalized to obtain a queryable trajectory in the subspace. The slave arm planning module is used to construct an extended space based on the configuration vector and time of the slave arm, plan the slave path in the extended space, verify the geometric collision avoidance constraints based on the queryable trajectory and the collider, verify the radiation prohibition constraints based on the three-dimensional radiation voxels, and obtain the extended space trajectory. The trajectory synthesis module is used to synthesize the subspace queryable trajectory and the extended space trajectory on a unified time axis to form a dual-arm collaborative trajectory.

[0014] Furthermore, the system also includes: The hybrid sampling module is used to generate sampling points by executing a hybrid sampling strategy based on radiation field information, including uniform sampling, target bias sampling, and low-radiation region guided sampling with a set probability. The cost update module is used to select the parent and reconnect the RRT* algorithm using a multi-objective cost function. The edge cost is defined as the weighted sum of the cumulative radiation dose of the joint motor and the path length in the task space, and the neighborhood cost is compared and reconnected for update.

[0015] Beneficial effects: This invention constructs a three-dimensional voxel radiation field and joint motor cumulative dose model adapted to the motion characteristics of robotic arms, achieving accurate quantification of radiation risk of key components and overcoming the limitation of traditional two-dimensional grid modeling in matching the three-dimensional motion characteristics of robotic arms. A hierarchical dynamic collision detection method based on overlapping workspace prediction is designed. By combining wide-phase screening and narrow-phase fine detection, invalid inter-arm detection calculations are significantly eliminated, greatly reducing the computational overhead in the high-dimensional dual-arm collaboration process. By introducing a radiation-guided hybrid sampling strategy and a multi-objective cost function, the algorithm converges to low-radiation regions while reducing sampling blindness, achieving multi-objective collaborative optimization of obstacle avoidance safety, path length, and radiation protection, significantly reducing the cumulative radiation dose of joints and greatly reducing computation time. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application 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 some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic flowchart of the main method steps of the present invention; Figure 2 This is a schematic diagram of the method flow of the present invention; Figure 3 This is a schematic diagram of the collision detection method of the present invention. Detailed Implementation

[0018] To make the above-mentioned objectives, features, and advantages of this application more apparent and understandable, the specific embodiments of this application are described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of this application. Therefore, this application is not limited to the specific embodiments disclosed below.

[0019] In the description of this application, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0020] In this application, unless otherwise expressly specified and limited, the terms "installation," "connection," "joining," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components, unless otherwise expressly limited. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0021] This invention provides a path planning method and system for a gantry-mounted dual robotic arm in a radiation environment. Its core lies in proposing an improved path planning algorithm—RadC-RRT*—that integrates radiation field constraints and a hybrid sampling strategy. This addresses the bottlenecks faced by existing methods in dealing with the coupling problem between high-dimensional configuration space and three-dimensional non-uniform radiation field, such as insufficient environmental radiation characterization, lack of protection for key joint components, and low sampling convergence efficiency.

[0022] Example 1: Reference Figures 1-3 As shown, this embodiment provides a path planning method for a gantry-mounted dual robotic arm in a radiation environment, mainly including the following steps: Step S1: Discretize the continuous radiation field into three-dimensional radiation voxels: To address the spatial variability of radiation distribution in a nuclear environment, a radiation field model based on a three-dimensional voxel mesh is constructed. To ensure that the spatial resolution of radiation dose calculations covers key components of the robotic arm, especially the radiation gradient changes experienced by the joint motors during movement, a three-dimensional mesh generation method based on joint size characteristics is proposed. Each joint motor is simplified as a cylindrical model, and according to spatial sampling theory, the mesh side length... Must meet: ; in, The minimum joint motor equivalent diameter in the system. A safety factor is used to suppress dose estimation bias caused by undersampling and improve assessment conservatism. A three-dimensional cubic mesh containing radiation information is called a radiation voxel, and the set of voxels is denoted as . Voxels The center coordinates are Its geometric domain is represented as: ; The numerical property of a radiation voxel is defined as the radiation intensity at the voxel's center, used for subsequent trajectory integration to obtain the cumulative dose. Let the radiation source be a point source, located at... The initial radiation intensity of the source is Voxel under point source irradiation The radiation intensity was calculated using voxel center sampling as follows: ; This reflects the inverse squared distance decay law. This is an exponential decay term, considering the absorption and scattering of radiation by the medium. No time decay, Let be the direction factor. To avoid ambiguity in the definition of the direction factor due to the non-uniqueness of the voxel normal, the incident direction is defined as: ; And let the set of normals of the six faces of the voxel be The direction factor is obtained by taking the most favorable receiving surface: ; Therefore, the voxel field of radiation can be written as During system motion, any key point can be quickly queried by matching its corresponding voxel value with the voxel index mapping; and the dose rate can be further numerically integrated along the planned path to obtain the cumulative dose, providing a foundation for subsequent collision determination under radiation constraints and the construction of the RadC-RRT* cost function.

[0023] Step S2: Establish the environmental obstacles, the main robotic arm and slave robotic arm of the gantry-supported dual robotic arm system, and the gantry structure as the collision body: To meet the requirements of high-frequency feasibility assessment, a hierarchical dynamic collision detection framework is constructed, progressing from wide-phase screening to narrow-phase fine detection to edge continuity verification. The system configuration vector is defined as: ; in, This represents the three translational degrees of freedom of the gantry in the world coordinate system. These are the joint angle vectors of the master and slave robotic arms, respectively. Since the master and slave robotic arms are fixedly mounted on the same gantry base, their relative mounting poses are constant. Therefore, collisions between the arms can be determined in the gantry base coordinate system, while collisions between the entire machine and the environment must also consider the impact on the world-frame pose.

[0024] The collision body is divided into three parts: the robotic arm link collision body, the gantry structure collision body, and the set of environmental obstacles. Each link of the master and slave robotic arms is approximated as a capsule, characterized by the endpoints of the link axes and the radius. The radius includes the body envelope and safety margin, thus obtaining the robotic arm link collision body. The rigid body structures, such as the gantry columns, beams, trolleys, and sliding platforms, are represented by a set of cuboids, forming a collection of colliding bodies. Environmental obstacles are represented by a set of spheres and AABBs, denoted as . Under radiation confinement, for voxels The dose rate Let the radiation exclusion threshold be... ,like: ; The voxel is then included in the forbidden set and participates in edge verification as an environmental barrier in the wide phase, i.e. This allows for hard constraint avoidance in extremely high radiation regions. Simultaneously, to improve motion conservation in radiation gradient regions, a safety margin that adapts to the radiation level is set. ; in, Based on the safety margin, For reference dose rate, where is the weighting coefficient. This mechanism allows the system to automatically increase the collision conservative boundary near high-radiation regions. In summary, the set of collision pairs that the system needs to consider is: ; The pose transformation of Lungmen in the world system can be written as a pure homogeneous translation transformation: ； Therefore, under any configuration, the base coordinates of the robotic arm and the gantry structure can be obtained from the world coordinates through translation.

[0025] Step S3: Construct a state subspace based on the configuration vectors of the main robotic arm and the gantry structure; plan the main path in the state subspace; verify the radiation prohibition constraint based on the three-dimensional radiation voxels; verify the geometric collision avoidance constraint based on the colliders; and obtain the main path point sequence. Phase I state is defined as: ； Nearest neighbor search is performed using a weighted ∞ norm metric. ; From RadC-RRT* Expand to For the edges, a discrete approximation of continuous collision detection is used, and the edge interpolation value is: ; Where N is the off-step number. To balance efficiency and security, an adaptive off-step number is adopted: ; in, The threshold for the maximum single-step translation of the gantry. This is the threshold for the maximum single-step change in joint angle. In subsequent dual-arm collaborative RRT*, to ensure that the master and slave robotic arms and the gantry execute edges synchronously on the same time axis, the edge execution duration can be defined: ; in, The maximum speed of the Dragon Gate The maximum angular velocity of the master and slave robotic arm joints. The time step can be directly unified as: ; No. The discrete time points are: ; Two types of fast constraint checks are performed for each interpolation point: geometric collision avoidance constraint verification and radiation prohibition constraint verification.

[0026] Geometric collision avoidance constraint verification employs hierarchical dynamic collision detection: 1) Wide-phase screening: For static environments, static AABB and BVH trees are established for environmental obstacles. For each discrete configuration, the AABB of each capsule and gantry box is calculated, and the set of potential intersecting candidate obstacles is queried in the BVH. In discrete configurations... Let the k-th gantry crane AABB be defined as... The j-th environmental AABB obstacle and the environmental sphere obstacle are Master-slave robotic arm The root connecting rod capsule body is To address the potential dynamic collision risks in dual-arm collaboration, static BVH trees are insufficient for wide-phase filtering. Therefore, a hierarchical collision detection method based on overlapping workspace regions is proposed. Monte Carlo methods are used to generate point clouds of the workspace for each single arm, and Boolean operations are used to obtain the point set of the overlapping region. ; To improve online decision-making efficiency, an axis-aligned bounding box is constructed for the point set as the trigger region: ; For any link capsule body's axial segment, uniformly arranged along its axis... One sampling point: ; If there exists any sampling point that satisfies: ; If the link fails to pass the test, it will be triggered to participate in the subsequent inter-arm narrow-phase collision fine inspection; otherwise, the fine inspection of the link and the opposite link will be skipped.

[0027] 2) Narrow-phase precision detection: Determines whether colliders intersect based on the geometric distance between them. and An intersection is determined if and only if all three coordinate axes overlap in their intervals; for and Let the nearest point of the center of the sphere on AABB be the point of intersection. If the distance is less than the radius of the sphere, then they intersect. For the capsule and the sphere, the capsule and AABB, and the capsule and the capsule, calculate the nearest point of the sphere's center to the line segment, the minimum distance of the line segment to AABB, and the minimum distance of the line segment to the line segment, respectively. If they are less than the radius of the capsule, then they intersect.

[0028] Verification of radiation exclusion constraints includes: 1) Wide Phase No-Entry Screening: For any configuration, let the set of center points of the joint motors of the master and slave robotic arms be... For any motor center point Voxel index mapping is used to determine which voxel it falls into: ; in Let be the reference origin for the voxel mesh. To cover the solid dimensions and discretization errors of the motor, neighborhood expansion is introduced. Let the equivalent radius of the motor be... Safety margin is Then the neighborhood radius can be taken as: ; And define a conservative dose rate: ; in For Centered on, with radius A set of voxel neighborhood indices. Given a dose rate threshold. The radiation wide-phase forbidden criterion is defined as: ; 2) Narrow-phase cumulative dose calculation: After a discrete configuration on the edge passes the radiation exclusion verification, a voxel-weighted dose rate estimate is further applied to each articulated motor. For the cylindrical articulated motor, for The voxels in the model are used to verify whether the cylinder and voxels truly intersect using the SAT separation axis theorem. For both the cylinder and voxels, seven separation axes need to be checked: the three principal axes of the voxels and the axis of the cylinder. Cross product of the cylinder axis and the principal axis of the voxel: ; For any split axis If all projections on the axes satisfy the intersection condition, then the cylinder and the voxel are considered to intersect. The complete intersection criterion of the separating axis theorem can be expressed as: ; After obtaining the voxel intersecting the motor, the weighted equivalent dose rate of the motor under the given configuration is defined as: ; in, Let represent the number of all voxels intersecting the cylinder, then the th... The cumulative dose of the motors on this side is: ; After edge verification, the edge enters cost evaluation and participates in the RRT* rewriting and reconnection process. Phase I outputs the main path node list: ; Step S4: Time-distill the main path point list to obtain the queryable trajectory in the subspace: The output path of stage I is time-diminished. The edges formed by adjacent path points are... According to the above formula, its shortest feasible execution time is: ; And order Define a uniform time step: ; This achieves consistency between geometrically discrete points and time-discrete points. For any Define the time interpolation of the main path: ; Thus, during the arm planning phase, time can be used to... The only confirmed synchronous configuration of the main arm and the gantry.

[0029] Step S5: Construct an extended space based on the configuration vector and time of the robotic arm, plan a path in the extended space, verify geometric collision avoidance constraints based on the queryable trajectory and the collider, verify radiation prohibition constraints based on the three-dimensional radiation voxels, and obtain the extended space trajectory: The arm planning is carried out within the extended space: ; in The total duration of the main path. This model treats the main arm as a dynamic obstacle that changes over time, passable from any state of the slave arm. Query the main arm configuration. Using synchronous interpolation along the arm edges, for each discrete point, query the main arm configuration, perform hierarchical inter-arm collision detection, slave arm-environment collision detection, and radial wide-phase exclusion filtering. If any discrete point does not satisfy the constraints, the edge is directly rejected. Under dynamic constraints, the slave arm may be unable to traverse the intersection area occupied by the main arm at certain time periods. To enhance cooperative feasibility, the slave arm is allowed to take a waiting action: ; This involves advancing the time while maintaining the same posture, allowing the main arm to pass through the conflict zone before the secondary arm continues its movement. Waiting actions can be incorporated as a special control input into the sampling extension and cost assessment.

[0030] When planning paths in the state subspace and the extended space, a hybrid sampling strategy based on radiation field information is adopted, dividing the samples into three categories and mixing them with probabilities: ; With probability Perform uniform sampling within the feasible boundary; with probability Bias sampling is performed on the target region; with probability Guided sampling in low-radiation areas: Generate uniform random sampling points and fill the candidate pool M after passing the radiation exclusion check. For candidate samples that pass the exclusion check, define the radiation guidance index as the sum of the joint motor dose rates: ; Let the candidate pool size be M and the Top-K parameters be K. The algorithm steps are as follows: Initialize the candidate set; Repeatedly generate uniform random sampling points and perform radiation exclusion judgment until the candidate pool M is filled; according to Sort the samples in ascending order and select the top K samples to form a set. To ensure diversity, one sample is randomly selected as the guide sampling output.

[0031] Simultaneously, a multi-objective cost function is employed for parent selection and reconnection in the RRT* algorithm: the cost of each edge is defined as the sum of the cumulative radiation doses of all joint motors on that edge, while the task space path length is added as a constraint to suppress path redundancy. Therefore, the cumulative dose of the motors on the edge is: ; The edge radiation cost is defined as: ; In RRT* optimization, the total path cost is the sum of edge costs: ; in The length of the task space path. As the weight. To highlight the radiation protection objective, take To minimize the primary driving dose, during the parent selection phase, candidate parent nodes in the neighborhood are compared: ; During the reconnection phase, the system checks whether the total cost can be reduced by going through the current node to neighboring nodes. If so, it reconnects.

[0032] Step S6: Combine the subspace queryable trajectory and the extended space trajectory on a unified time axis to form a dual-arm cooperative trajectory.

[0033] Example 2: This embodiment provides a path planning system for a gantry-mounted dual robotic arm in a radiation environment, used to implement the path planning method described in Embodiment 1, including: The radiation modeling module is used to discretize a continuous radiation field into three-dimensional radiation voxels; The collision modeling module is used to create collision bodies from environmental obstacles, the main robotic arm, the slave robotic arm, and the gantry structure of the gantry-supported dual robotic arm system. The main arm planning module is used to construct a state subspace based on the configuration vectors of the main robotic arm and the gantry structure, plan the main path in the state subspace, verify the radiation prohibition constraint based on the three-dimensional radiation voxels, verify the geometric collision avoidance constraint based on the collider, and obtain the main path point sequence; the main path point sequence is temporalized to obtain a queryable trajectory in the subspace. The slave arm planning module is used to construct an extended space based on the configuration vector and time of the slave arm, plan the slave path in the extended space, verify the geometric collision avoidance constraints based on the queryable trajectory and the collider, verify the radiation prohibition constraints based on the three-dimensional radiation voxels, and obtain the extended space trajectory. The trajectory synthesis module is used to synthesize the subspace queryable trajectory and the extended space trajectory on a unified time axis to form a dual-arm collaborative trajectory.

[0034] The hybrid sampling module is used to generate sampling points by executing a hybrid sampling strategy based on radiation field information, including uniform sampling, target bias sampling, and low-radiation region guided sampling with a set probability. The cost update module is used to select the parent and reconnect the RRT* algorithm using a multi-objective cost function. The edge cost is defined as the weighted sum of the cumulative radiation dose of the joint motor and the path length in the task space, and the neighborhood cost is compared and reconnected for update.

[0035] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0036] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A path planning method for a gantry-mounted dual robotic arm in a radiation environment, characterized in that, Includes the following steps: Discretize the continuous radiation field into three-dimensional radiation voxels; The environmental obstacles, the main robotic arm, the slave robotic arm, and the gantry structure of the dual robotic arm system are established as the collision body; A state subspace is constructed based on the configuration vectors of the main robotic arm and the gantry structure. A main path is planned in the state subspace. Radiation prohibition constraints are verified based on the three-dimensional radiation voxels. Geometric collision avoidance constraints are verified based on the colliders. A main path point sequence is obtained. The main path point sequence is temporalized to obtain a queryable trajectory in the subspace. An extended space is constructed based on the configuration vector and time of the robotic arm. A path is planned in the extended space. Geometric collision avoidance constraints are verified based on the queryable trajectory and the collider. Radiation prohibition constraints are verified based on the three-dimensional radiation voxels. The extended space trajectory is obtained. The queryable trajectory in the subspace and the extended space trajectory are synthesized into a dual-arm cooperative trajectory on a unified time axis.

2. The path planning method for a gantry-mounted dual robotic arm in a radiation environment according to claim 1, characterized in that, Discretizing the continuous radiation field into three-dimensional radiation voxels includes: The three-dimensional mesh side length is determined based on the minimum joint motor equivalent diameter and safety factor in the system, wherein the mesh side length is less than or equal to the ratio of the minimum joint motor equivalent diameter to the safety factor; Based on the location of the radiation source, the inverse square law of distance attenuation, the dielectric exponential attenuation term, and the direction factor of the most favorable receiving surface, the radiation intensity at the center of the voxel is calculated as the numerical attribute of the radiation voxel.

3. The path planning method for a gantry-mounted dual robotic arm in a radiation environment according to claim 1, characterized in that, The process of establishing environmental obstacles, the main robotic arm, the slave robotic arm, and the gantry structure of the dual robotic arm system as a collision body includes: Each link of the main robotic arm and the slave robotic arm is approximated as a capsule, the gantry structure is represented by multiple cuboids, and the environmental obstacles are represented by spheres aligned with axes and enclosing boxes. Radiation voxels whose dose rates exceed the radiation exclusion threshold are included in the exclusion set and used as environmental barriers in edge validation. A safety margin that adapts to the radiation level is set, and the safety margin increases as the radiation dose rate increases.

4. The path planning method for a gantry-mounted dual robotic arm in a radiation environment according to claim 1, characterized in that, Verifying geometric collision avoidance constraints based on the collider includes employing hierarchical dynamic collision detection: Wide-phase screening: A static axis-aligned bounding box hierarchical tree is established for environmental obstacles. The bounding boxes of each collider are calculated for discrete configurations, and potential intersecting candidate obstacles are queried in the hierarchical tree. For dynamic collisions between master and slave robotic arms, the overlapping area point set of the workspace of the two arms is pre-calculated. An axis-aligned bounding box is constructed for the overlapping area point set as a trigger area. When the sampling point on the axis of the connecting capsule enters the trigger area, the narrow-phase collision fine detection between the arms is triggered. Narrow phase precision detection: Determines whether colliding bodies intersect based on the geometric distance between them; Edge continuity verification: The planned edges are interpolated using an adaptive discrete step count, and the wide-phase screening and narrow-phase fine-tuning are performed for each discrete configuration.

5. The path planning method for a gantry-mounted dual robotic arm in a radiation environment according to claim 1, characterized in that, Verification of radiation exclusion constraints based on the aforementioned three-dimensional radiation voxel includes: Wide phase no-entry screening: Voxel index mapping is performed on the center point of the joint motor, and neighborhood expansion is performed by combining the equivalent radius of the motor and the adaptive safety margin. The conservative dose rate in the neighborhood voxel set is queried. If the dose rate threshold is exceeded, it is determined to be infeasible. Narrow-phase cumulative dose calculation: For configurations that pass the wide-phase no-entry screening, the spatial intersection of the articulated motor cylinder and voxels is verified based on the separated axis theorem. The radiation intensity of the intersecting voxels is weighted, the weighted equivalent dose rate of the articulated motor is calculated, and the cumulative radiation dose is obtained by integrating along the path.

6. The path planning method for a gantry-mounted dual robotic arm in a radiation environment according to claim 1, characterized in that, The temporalization of the main path point list and the planning of secondary paths in the extended space include: Based on the maximum speed of the gantry and the maximum angular velocity of the master and slave robotic arm joints, calculate the shortest feasible execution time for the edge formed by adjacent path points of the main path, and define a uniform time step to achieve consistency between geometric discrete points and time discrete points. In the extended space, the main arm is regarded as a dynamic obstacle that changes over time. The state of the slave arm is queried through the configuration of the main arm over time to perform dynamic inter-arm collision detection. When the slave arm cannot cross the area occupied by the main arm within a specific time period, the slave arm is allowed to take a waiting action, that is, advance time while maintaining its posture.

7. The path planning method for a gantry-mounted dual robotic arm in a radiation environment according to claim 1, characterized in that, Path planning is performed in the state subspace and the extended space, employing a hybrid sampling strategy based on radiation field information, including: Perform uniform sampling within the feasible boundary with a first probability; The target region is biased and sampled with a second probability; Low-radiation area guided sampling is performed with a third probability: uniform random sampling points are generated and filled into the candidate pool after passing the radiation exclusion judgment. The radiation guidance index of the candidate samples is calculated and sorted in ascending order. One sample is randomly selected from the set of the top K samples as the guided sampling output. The radiation guidance index is the sum of the joint motor dose rates.

8. A path planning method for a gantry-mounted dual robotic arm in a radiation environment according to claim 5 or 7, characterized in that, During path planning, a multi-objective cost function is used for parent selection and reconnection in the RRT* algorithm: The cost of each edge is defined as the weighted sum of the cumulative radiation dose of all joint motors on that edge and the path length in the mission space. During the parent selection phase, the total cost of the path through the candidate parent node in the neighborhood is compared, and the node with the lowest cost is selected as the parent node. During the reconnection phase, it is determined whether the total cost can be reduced by going through the current node to the neighboring nodes. If so, reconnection is performed.

9. A path planning system for a gantry-mounted dual robotic arm in a radiation environment, characterized in that, To implement the path planning method as described in any one of claims 1-8, comprising: The radiation modeling module is used to discretize a continuous radiation field into three-dimensional radiation voxels; The collision modeling module is used to create collision bodies from environmental obstacles, the main robotic arm, the slave robotic arm, and the gantry structure of the gantry-supported dual robotic arm system. The main arm planning module is used to construct a state subspace based on the configuration vectors of the main robotic arm and the gantry structure, plan the main path in the state subspace, verify the radiation prohibition constraint based on the three-dimensional radiation voxels, verify the geometric collision avoidance constraint based on the collider, and obtain the main path point sequence; the main path point sequence is temporalized to obtain a queryable trajectory in the subspace. The slave arm planning module is used to construct an extended space based on the configuration vector and time of the slave arm, plan the slave path in the extended space, verify the geometric collision avoidance constraints based on the queryable trajectory and the collider, verify the radiation prohibition constraints based on the three-dimensional radiation voxels, and obtain the extended space trajectory. The trajectory synthesis module is used to synthesize the subspace queryable trajectory and the extended space trajectory on a unified time axis to form a dual-arm collaborative trajectory.

10. A path planning system for a gantry-mounted dual robotic arm in a radiation environment according to claim 9, characterized in that, The system also includes: The hybrid sampling module is used to generate sampling points by executing a hybrid sampling strategy based on radiation field information, including uniform sampling, target bias sampling, and low-radiation region guided sampling with a set probability. The cost update module is used to select the parent and reconnect the RRT* algorithm using a multi-objective cost function. The edge cost is defined as the weighted sum of the cumulative radiation dose of the joint motor and the path length in the task space, and the neighborhood cost is compared and reconnected for update.

Images