Multi-mode real-time obstacle avoidance method and system for wafer transmission robot

Abstract

The invention relates to the technical field of robot obstacle avoidance, in particular to a wafer transfer robot-oriented multi-modal real-time obstacle avoidance method and system, and the method comprises the steps: collecting environment data in real time through a multi-modal data collection device, and obtaining multi-modal environment data; performing space-time alignment on the multi-modal environment data, and establishing a workspace map; collecting operation data of the dynamic obstacle, and inputting the operation data into the constructed trajectory prediction model for trajectory prediction to obtain a trajectory prediction result; performing path planning on the wafer transmission robot to generate an optimal obstacle avoidance path; and verifying the optimal obstacle avoidance path, and outputting the optimal obstacle avoidance path to a motion controller for execution. According to the wafer carrying robot, the problem that the obstacle avoidance capability of the wafer carrying robot is insufficient in a dynamic environment is effectively solved, damage caused by collision in the wafer carrying process is effectively avoided, and the operation accuracy and safety of the wafer carrying robot in a complex production environment are improved.

Description

Technical Field

[0001] This invention relates to the field of robot obstacle avoidance technology, and in particular to a multimodal real-time obstacle avoidance method and system for wafer transfer robots. Background Technology

[0002] Wafer transfer robots are mainly used for operations such as transporting, transferring, and storing wafers. Their accuracy and stability directly affect production efficiency and product quality. With the increasing demands for automation in intelligent manufacturing, obstacle avoidance capability has become one of the key technologies for improving the efficiency and safety of wafer transfer robots. Traditional wafer transfer robots use torque feedback technology to determine whether the robot has collided with an obstacle. When the robot comes into contact with an obstacle, the torque sensor can detect the collision in real time and trigger an emergency stop operation through the feedback mechanism to quickly stop the robot's movement.

[0003] While torque feedback can effectively trigger emergency stops to prevent further robot movement, this method still has significant limitations in wafer handling. Due to the brittleness of wafers, simple emergency stops cannot completely eliminate the effects of collisions, especially when facing suddenly appearing dynamic obstacles. Traditional algorithms often cannot react sufficiently before a collision occurs, and the robot may still experience minor collisions within a short period, leading to irreversible damage to the wafer. Therefore, a multimodal real-time obstacle avoidance method for wafer transport robots is urgently needed. Summary of the Invention

[0004] This invention provides a multimodal real-time obstacle avoidance method and system for wafer transport robots, which can effectively solve the problems in the background art.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A multimodal real-time obstacle avoidance method for wafer transport robots, the method comprising: Multimodal environmental data is obtained by collecting environmental data in real time using a multimodal data acquisition device. The multimodal environment data is spatiotemporally aligned to create a workspace map; Collect the operational data of dynamic obstacles, input the operational data into the constructed trajectory prediction model to predict the trajectory, and obtain the trajectory prediction result; Based on the workspace map and the trajectory prediction results, path planning is performed on the wafer transport robot to generate the optimal obstacle avoidance path. The optimal obstacle avoidance path is verified, and then the optimal obstacle avoidance path is output to the motion controller for execution.

[0006] Furthermore, the deployment of multimodal data acquisition devices includes: The working area of ​​the wafer transfer robot is decomposed, and several candidate installation combinations are generated. Visibility analysis is performed on the candidate installation combinations, and visibility indices are calculated; A comprehensive score is calculated based on the visibility index, and the candidate installation combination with the highest comprehensive score is selected for deployment.

[0007] Furthermore, the visibility metrics include coverage, redundancy, modal confidence, and coverage uniformity.

[0008] Furthermore, spatiotemporal alignment of the multimodal environment data to establish a workspace map includes: A unified time base is applied to multimodal environmental data, and a timestamp is added. Perform extrinsic parameter joint calibration on the multimodal environment data and project it onto the map coordinate system; The spatiotemporally aligned multimodal environmental data is fused with confidence, and the fusion result is written into the map data structure to generate a workspace map.

[0009] Furthermore, the map data structure is divided into a static layer and a dynamic layer. The static layer stores information on fixed obstacles and structural boundaries, while the dynamic layer stores information on dynamic obstacles.

[0010] Furthermore, based on the workspace map and the trajectory prediction results, path planning is performed on the wafer transport robot to generate the optimal obstacle avoidance path, including: The trajectory prediction results are written into the workspace map at preset time intervals, and an occupancy probability is assigned to the spatial unit of each predicted location to obtain a dynamic workspace map. Several candidate velocity pairs are sampled in the velocity space, and candidate obstacle avoidance paths are generated based on the candidate velocity pairs. The candidate obstacle avoidance path pairs are then filtered through the dynamic workspace map. Define a path cost function, calculate the cost value of the selected candidate obstacle avoidance paths, and select the obstacle avoidance path with the smallest cost value as the optimal obstacle avoidance path.

[0011] Furthermore, filtering the candidate obstacle avoidance path pairs using the dynamic workspace map includes: Set an occupancy probability threshold. If the occupancy probability of any spatial unit on the candidate obstacle avoidance path is greater than or equal to the occupancy probability threshold, then the candidate obstacle avoidance path is discarded. If the occupancy probability of a spatial unit on a candidate obstacle avoidance path is less than the occupancy probability threshold, then the cost of the candidate obstacle avoidance path is calculated.

[0012] Furthermore, this includes: performing dual verification of the optimal obstacle avoidance path, including collision risk verification and vibration stability verification.

[0013] A multimodal real-time obstacle avoidance system for wafer transport robots, the system comprising: The environmental data acquisition module collects environmental data in real time through a multimodal data acquisition device to obtain multimodal environmental data; The spatial map construction module performs spatiotemporal alignment on the multimodal environment data to establish a workspace map; The dynamic trajectory prediction module collects the operational data of dynamic obstacles, inputs the operational data into the constructed trajectory prediction model to perform trajectory prediction, and obtains the trajectory prediction result; The optimal path generation module performs path planning for the wafer transport robot based on the workspace map and the trajectory prediction results, and generates the optimal obstacle avoidance path. The optimal path verification module verifies the optimal obstacle avoidance path and outputs the optimal obstacle avoidance path to the motion controller for execution.

[0014] Furthermore, the optimal path generation module includes: The dynamic map generation unit writes the trajectory prediction results into the workspace map at preset time intervals and assigns an occupancy probability to the spatial cell of each predicted location to obtain a dynamic workspace map. The candidate path filtering unit samples several sets of candidate velocity pairs in the velocity space, generates candidate obstacle avoidance paths based on the candidate velocity pairs, and filters the candidate obstacle avoidance path pairs through the dynamic workspace map. The optimal path selection unit defines a path cost function, calculates the cost value of the filtered candidate obstacle avoidance paths, and selects the obstacle avoidance path with the smallest cost value as the optimal obstacle avoidance path.

[0015] The technical solution of this invention can achieve the following technical effects: It effectively solves the problem of insufficient obstacle avoidance capability of wafer handling robots in dynamic environments. By constructing a workspace map in real time, it achieves comprehensive perception of both static and dynamic obstacles, improving the precise obstacle avoidance capability of wafer handling robots. It adopts dynamic path planning, which greatly improves the efficiency of path planning, and uses a dual verification mechanism to ensure the safety and stability of the obstacle avoidance path. It effectively avoids damage to wafers caused by collisions during handling, and improves the operational accuracy and safety of wafer handling robots in complex production environments.

[0016] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0017] 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 some embodiments recorded in 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 flowchart illustrating a multimodal real-time obstacle avoidance method for wafer transport robots. Figure 2 A flowchart illustrating the setup of a multimodal data acquisition device; Figure 3 A flowchart illustrating the process of creating a workspace map; Figure 4 A flowchart illustrating the process of generating the optimal obstacle avoidance path; Figure 5 This is a schematic diagram of a multimodal real-time obstacle avoidance system for wafer transport robots. Detailed Implementation

[0019] The technical solutions 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.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0021] Example 1 like Figure 1 As shown, a multimodal real-time obstacle avoidance method for wafer transport robots is presented, the method including: S1: Collect environmental data in real time through a multimodal data acquisition device to obtain multimodal environmental data; Specifically, to cover different perception ranges and accuracy and ensure a comprehensive understanding of the surrounding environment, this step uses a multimodal data acquisition device to acquire multi-source perception information of the wafer transfer robot's working environment in real time. This device includes LiDAR, a vision camera, ultrasonic sensors, an inertial measurement unit, and infrared sensors. Each sensor acquires environmental data in different dimensions. For example, LiDAR can provide high-precision distance measurement and point cloud information; the vision camera can acquire information such as the texture, shape, and color of obstacles. Through multimodal data acquisition, complementarity and redundancy can be formed between different perception modalities. This improves the robustness and accuracy of environmental perception and provides more usable features for trajectory prediction and obstacle avoidance planning, significantly enhancing the safe operation capability of the wafer transfer robot in complex dynamic environments.

[0022] S2: Perform spatiotemporal alignment on multimodal environment data and establish a workspace map; Based on the above embodiments, multimodal environmental data can reflect information about the space surrounding the wafer transfer robot, including spatial geometry information, dynamic obstacle information, and environmental state information. This step performs time synchronization and spatial calibration on the collected multimodal environmental data, unifying the data from different modal sensors to the same timestamp and coordinate system to achieve spatiotemporal alignment of the data. Based on the aligned data, a workspace map is constructed. The workspace map is used to represent information about static obstacles, dynamic obstacles, and passable areas in the operating environment of the wafer transfer robot, forming a structured workspace map. This can significantly improve the efficiency and real-time performance of obstacle avoidance planning, providing an accurate environmental model for subsequent trajectory prediction and obstacle avoidance planning.

[0023] S3: Collect the operational data of dynamic obstacles, input the operational data into the constructed trajectory prediction model to perform trajectory prediction, and obtain the trajectory prediction result; In this embodiment, dynamic obstacles exist in the operating environment of the wafer transfer robot. If obstacle avoidance planning is only based on the current position, there is a lack of lead time, which poses a risk of collision. In order to make subsequent obstacle avoidance decisions more forward-looking, this step constructs a trajectory prediction model to predict the trajectory of dynamic obstacles. The trajectory prediction model can be flexibly selected according to the real-time requirements of the wafer transfer robot's operating environment, the type of dynamic obstacle, and the computing resource conditions. It includes kinematic models, probabilistic filtering models, deep learning models, etc., which allows obstacle avoidance planning to consider the future state of obstacles rather than just the current state, thereby improving the operational stability of the wafer transfer robot and reducing the risk of wafer damage.

[0024] S4: Based on the workspace map and trajectory prediction results, perform path planning for the wafer transfer robot and generate the optimal obstacle avoidance path; S5: Verify the optimal obstacle avoidance path and output the optimal obstacle avoidance path to the motion controller for execution.

[0025] Specifically, static maps alone cannot handle obstacles that change over time; dynamic predictions alone lack the constraints of fixed obstacles. This step integrates static environmental information with the future state of dynamic obstacles, and generates the optimal obstacle avoidance path under the current conditions through a path planning algorithm. The optimal obstacle avoidance path is then subjected to secondary feasibility and safety verification based on the latest perception data, including continuous-time collision detection and path smoothing, to ensure that there is no collision risk during the execution of the path and that it meets the stability requirements of wafer handling. If the verification is successful, the path parameters are converted into low-level instructions that the motion controller can directly execute; if the verification fails, a hierarchical response strategy is triggered, and the local path is replanned or an emergency stop is initiated based on the risk level.

[0026] The technical solution of this invention effectively solves the problem of insufficient obstacle avoidance capability of wafer handling robots in dynamic environments. By constructing a workspace map in real time, it achieves comprehensive perception of both static and dynamic obstacles, improving the precise obstacle avoidance capability of the wafer handling robot. The adoption of dynamic path planning significantly improves the efficiency of path planning, and the use of a dual verification mechanism ensures the safety and stability of the obstacle avoidance path, effectively avoiding damage to wafers caused by collisions during handling, and improving the operational accuracy and safety of the wafer handling robot in complex production environments.

[0027] To ensure optimal configuration of the multimodal data acquisition device, such as Figure 2 As shown, the multimodal data acquisition device includes: S11: Decompose the working area of ​​the wafer transfer robot and generate several candidate installation combinations; This step spatially decomposes the working area of ​​the wafer transfer robot. Based on the structural characteristics, workflow, and obstacle distribution of the working area, it is divided into multiple functional sub-regions. After decomposition, suitable sensor types and installation methods can be selected for each sub-region. Multiple candidate installation combinations are generated within each functional sub-region. Each candidate installation combination includes the installation location coordinates, installation orientation information, and the correspondence between sensor type and installation location. Within each sub-region, a specific set of sampling points is generated based on the spatial shape, obstacle distribution, and key monitoring point locations. Locations unsuitable for sensor installation are filtered out from the sampling points, retaining suitable and structurally stable points as candidate installation locations. The location coordinates and orientation parameters of each candidate location are recorded. Simultaneously, a multimodal sensor library is established, and performance parameters are recorded for each sensor. Based on the sensing requirements such as the coverage area, accuracy requirements, dynamic obstacle ratio, and lighting conditions of the sub-region, suitable types and quantities are selected from the sensor library. Candidate installation locations are combined with matching sensor types to form candidate installation combinations.

[0028] S12: Perform visibility analysis on candidate installation combinations and calculate visibility indices; S13: Calculate the comprehensive score based on the visibility index, and select the candidate installation combination with the highest comprehensive score for deployment.

[0029] Visibility analysis is performed on candidate installation locations. Based on the field of view parameters, detection distance, resolution, and anti-occlusion capabilities of various sensors, the detection of spatial points within the working area by the sensors at these locations is simulated, and visibility indices are calculated. Based on the visibility indices obtained from the visibility analysis, and combined with constraints such as sensor installation cost, wiring complexity, and bandwidth resources, a comprehensive score is calculated for each candidate installation combination. The scoring method can adopt a weighted summation strategy, and the candidate installation combination with the highest comprehensive score is selected for deployment to ensure the coverage integrity and redundancy of multimodal data and reduce the impact of blind spots on subsequent environmental perception and obstacle avoidance.

[0030] Furthermore, visibility metrics include coverage, redundancy, modal confidence, and coverage uniformity.

[0031] Coverage is the ratio of the effective working space volume covered by the sensor's field of view to the entire working space, reflecting the overall coverage integrity; redundancy is the proportion of the same spatial area covered by two or more sensors, used to improve the system's fault tolerance and robustness; modal confidence is a reliability score calculated based on the detection accuracy and stability of different sensor modes in the area, ensuring data reliability under complex environments such as changes in illumination and reflection interference; coverage uniformity reflects the consistency of coverage density in different spatial areas, avoiding overly dense coverage in some areas and sparse coverage in others, ensuring balanced global sensing performance.

[0032] Based on the above embodiments, such as Figure 3 As shown, spatiotemporal alignment of multimodal environmental data to establish a workspace map includes: S21: Unify the time base for multimodal environment data and add a timestamp; Specifically, this step of time alignment can be achieved by constructing a unified time base using a precise time protocol or an external trigger signal, and by attaching a high-precision timestamp to each modal observation frame. Through unified time base and timestamp management, time domain misalignment caused by asynchronous sampling of different modalities can be eliminated. Since there are inherent sampling time offsets at the hardware and driver levels of sensors of different modalities, fixed time offsets of each modality can be calibrated and compensated through event alignment or cross-correlation methods to eliminate fusion errors caused by time misalignment.

[0033] S22: Perform extrinsic parameter joint calibration on multimodal environment data and project it onto the map coordinate system; Specifically, due to the differences in installation location and orientation of different types of sensors, the data they collect are located in their respective sensor coordinate systems. Direct use would lead to spatial alignment errors. Therefore, this step uses external parameter calibration methods such as calibration plate method, laser-camera joint calibration method, and multi-view geometry method to determine the coordinate system relationship of each sensor and obtain the rigid transformation matrix from each sensor coordinate system to the map coordinate system. After completing the external parameter calibration, each modal data is projected onto the map coordinate system through the corresponding rigid transformation matrix, ensuring that the detection results of different sensors for the same object coincide in spatial position, which is convenient for subsequent confidence fusion and accurate obstacle localization.

[0034] S23: Perform confidence fusion on the spatiotemporally aligned multimodal environmental data, write the fusion result into the map data structure, and generate a workspace map.

[0035] After completing spatiotemporal alignment and projection, the multi-source observations at the same spatial location are weighted according to real-time data quality and modal characteristics. The confidence level can be obtained based on indicators such as echo intensity, depth confidence, image clarity and historical consistency. The observation results at the same spatial location are weighted and synthesized according to the detection reliability of each sensor mode in different regions. The fused observation results are written into a three-dimensional voxel map data structure to generate a workspace map containing static obstacles, dynamic obstacles and passable areas.

[0036] Furthermore, the map data structure is divided into a static layer and a dynamic layer. The static layer stores information about fixed obstacles and structural boundaries, while the dynamic layer stores information about dynamic obstacles.

[0037] As a preferred embodiment, the workspace map adopts a layered structure, including a static layer and a dynamic layer. The static layer is used to store environmental elements in the wafer transfer robot's operating environment that do not change over time, such as the robot body, fences, tracks, and other fixed obstacles. The dynamic layer is used to store dynamic obstacle information in the environment that changes over time, including its current position, speed, acceleration, and historical trajectory data. The layered map structure ensures high-precision modeling of the static environment while improving the real-time updating capability of dynamic environment information, enabling the wafer transfer robot to perform efficient path planning and obstacle avoidance in complex and mixed operating environments, reducing computational overhead and enhancing the robustness and adaptability of the system.

[0038] Based on the above embodiments, such as Figure 4 As shown, based on the workspace map and trajectory prediction results, path planning is performed on the wafer transport robot to generate the optimal obstacle avoidance path, including: S41: Write the trajectory prediction results into the workspace map at preset time intervals, and assign an occupancy probability to the spatial unit of each predicted location to obtain a dynamic workspace map. In this embodiment, the future positions of dynamic obstacles corresponding to the trajectory prediction results are mapped to a three-dimensional workspace map that is fused with static obstacle information at a set time interval. Within the future prediction time domain, an occupancy probability value is assigned to the spatial unit corresponding to the future prediction position. The occupancy probability reflects the possibility that the unit will be occupied by an obstacle at that future moment, forming a map data structure that reflects dynamic occupancy information that changes over time. This ensures that the map data not only contains the static distribution of fixed obstacles, but also reflects the movement trend of dynamic obstacles in real time, thereby providing the latest environmental status for path planning.

[0039] S42: Sample several candidate velocity pairs in the velocity space, generate candidate obstacle avoidance paths based on the candidate velocity pairs, and filter the candidate obstacle avoidance path pairs through a dynamic workspace map; S43: Define the path cost function, calculate the cost of the candidate obstacle avoidance paths after screening, and select the obstacle avoidance path with the minimum cost as the optimal obstacle avoidance path.

[0040] Within the velocity space, a deterministic sampling method is used to sample several candidate velocity pairs. Based on each candidate velocity pair, the kinematic model of the wafer transport robot is used to perform trajectory integration in the planning time domain, generating corresponding candidate obstacle avoidance paths. Using a dynamic workspace map, the occupancy probability of each spatial unit on the path is queried, filtering out a set of candidate paths that meet geometric feasibility requirements, thus eliminating high-risk paths in advance. After filtering, the cost value of the remaining candidate obstacle avoidance paths is calculated. The path cost can be weighted and summed according to preset weights. The path cost function includes at least the minimum obstacle distance, path deviation, motion smoothness, and dynamic obstacle risk weights. To select the optimal obstacle avoidance path, a min-heap priority queue with cost as the key value can be used to manage candidate trajectories, selecting the path with the lowest cost value as the optimal obstacle avoidance path, ensuring that the wafer transport robot can safely, smoothly, and efficiently complete the transport task even with dynamic obstacles.

[0041] Furthermore, the selection of candidate obstacle avoidance path pairs through a dynamic workspace map includes: S421: Set an occupancy probability threshold. If the occupancy probability of any spatial unit on the candidate obstacle avoidance path is greater than or equal to the occupancy probability threshold, the candidate obstacle avoidance path will be discarded. S422: If the occupancy probability of a spatial unit on a candidate obstacle avoidance path is less than the occupancy probability threshold, then calculate the cost of the candidate obstacle avoidance path.

[0042] As a preferred embodiment, when screening candidate obstacle avoidance paths using a dynamic workspace map, an occupancy probability threshold is first set. This threshold is used to determine whether there is an unacceptable collision risk in the spatial units traversed by the path. For each candidate obstacle avoidance path, all spatial units it passes through are traversed along the path. If the occupancy probability of any spatial unit is greater than or equal to the occupancy probability threshold, it is determined that the path has a high-risk dynamic obstacle within the prediction time window, and the path is directly discarded and no longer participates in subsequent cost calculations. For candidate obstacle avoidance paths that are not discarded, their cost is calculated. This embodiment, through the occupancy probability threshold screening mechanism, can eliminate obviously infeasible paths during the planning stage, reduce invalid calculations, and improve planning efficiency, thereby achieving a balance between safety and operational efficiency.

[0043] To ensure the safety and stability of the wafer transfer robot when executing the optimal obstacle avoidance path, the optimal obstacle avoidance path is subject to dual verification of collision risk and vibration stability.

[0044] Specifically, the generated obstacle avoidance path undergoes dual verification. Collision risk verification calculates the minimum distance between each point on the path and surrounding obstacles at each time step and compares it with a set safe distance threshold. If the minimum distance between each point on the path and an obstacle is greater than or equal to the safe distance threshold, the path is considered safe. If it is less than the threshold, the path is considered infeasible and must be replanned. Collision risk verification ensures that collisions do not occur during path planning by checking the relationship between the minimum distance between each point on the path and the obstacle and the safe distance, thus guaranteeing the safe operation of the robot. Vibration stability verification verifies the curvature and rate of curvature change of the path. A large rate of curvature change can lead to sharp turns in the path, potentially causing significant vibration and affecting the stability of the wafer. It is necessary to ensure that the rate of curvature change of the path is within a certain range. If the rate of curvature change is within the maximum rate of curvature change threshold, the path is considered safe. Vibration stability verification avoids sharp turns and vibrations in the path by limiting the rate of curvature change of the path, ensuring that the wafer is not subjected to excessive vibration during transportation, thereby improving transmission accuracy and safety.

[0045] Example 2 like Figure 5 As shown, a multimodal real-time obstacle avoidance system for wafer transport robots includes: The environmental data acquisition module collects environmental data in real time through a multimodal data acquisition device to obtain multimodal environmental data; The spatial map building module performs spatiotemporal alignment of multimodal environmental data to create a workspace map; The dynamic trajectory prediction module collects the operational data of dynamic obstacles, inputs the operational data into the constructed trajectory prediction model to predict the trajectory, and obtains the trajectory prediction result. The optimal path generation module performs path planning for the wafer transport robot based on the workspace map and trajectory prediction results, generating the optimal obstacle avoidance path. The optimal path verification module verifies the optimal obstacle avoidance path and outputs the optimal obstacle avoidance path to the motion controller for execution.

[0046] The adjustment system described above in this invention can effectively realize a multimodal real-time obstacle avoidance method for wafer transfer robots, and the technical effects it can achieve are as described in the above embodiments, and will not be repeated here.

[0047] Furthermore, the optimal path generation module includes: The dynamic map generation unit writes the trajectory prediction results into the workspace map at preset time intervals and assigns an occupancy probability to the spatial cell of each predicted location to obtain a dynamic workspace map. The candidate path filtering unit samples several sets of candidate velocity pairs in the velocity space, generates candidate obstacle avoidance paths based on the candidate velocity pairs, and filters the candidate obstacle avoidance path pairs through a dynamic workspace map. The optimal path selection unit defines a path cost function, calculates the cost value of the candidate obstacle avoidance paths after screening, and selects the obstacle avoidance path with the minimum cost value as the optimal obstacle avoidance path.

[0048] Similarly, the above-mentioned optimization schemes for the system can also achieve the optimization effects corresponding to the methods in Embodiment 1, which will not be repeated here.

[0049] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and accompanying drawings are merely exemplary illustrations of the application as defined herein, and are to be considered as covering any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Thus, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A multimodal real-time obstacle avoidance method for wafer transport robots, characterized in that, The method includes: Multimodal environmental data is obtained by collecting environmental data in real time using a multimodal data acquisition device. The multimodal environment data is spatiotemporally aligned to create a workspace map; Collect the operational data of dynamic obstacles, input the operational data into the constructed trajectory prediction model to predict the trajectory, and obtain the trajectory prediction result; Based on the workspace map and the trajectory prediction results, path planning is performed on the wafer transport robot to generate the optimal obstacle avoidance path. The optimal obstacle avoidance path is verified, and then the optimal obstacle avoidance path is output to the motion controller for execution.

2. The multimodal real-time obstacle avoidance method for wafer transport robots according to claim 1, characterized in that, The deployment of multimodal data acquisition devices includes: The working area of ​​the wafer transfer robot is decomposed, and several candidate installation combinations are generated. Visibility analysis is performed on the candidate installation combinations, and visibility indices are calculated; A comprehensive score is calculated based on the visibility index, and the candidate installation combination with the highest comprehensive score is selected for deployment.

3. The multimodal real-time obstacle avoidance method for wafer transport robots according to claim 2, characterized in that, The visibility metrics include coverage, redundancy, modal confidence, and coverage uniformity.

4. The multimodal real-time obstacle avoidance method for wafer transport robots according to claim 1, characterized in that, Spatiotemporal alignment of the multimodal environment data to establish a workspace map includes: A unified time base is applied to multimodal environmental data, and a timestamp is added. Perform extrinsic parameter joint calibration on the multimodal environment data and project it onto the map coordinate system; The spatiotemporally aligned multimodal environmental data is fused with confidence, and the fusion result is written into the map data structure to generate a workspace map.

5. The multimodal real-time obstacle avoidance method for wafer transport robots according to claim 4, characterized in that, include: The map data structure is divided into a static layer and a dynamic layer. The static layer stores information on fixed obstacles and structural boundaries, while the dynamic layer stores information on dynamic obstacles.

6. The multimodal real-time obstacle avoidance method for wafer transport robots according to claim 1, characterized in that, Based on the workspace map and the trajectory prediction results, path planning is performed on the wafer transport robot to generate the optimal obstacle avoidance path, including: The trajectory prediction results are written into the workspace map at preset time intervals, and an occupancy probability is assigned to the spatial unit of each predicted location to obtain a dynamic workspace map. Several candidate velocity pairs are sampled in the velocity space, and candidate obstacle avoidance paths are generated based on the candidate velocity pairs. The candidate obstacle avoidance path pairs are then filtered through the dynamic workspace map. Define a path cost function, calculate the cost value of the selected candidate obstacle avoidance paths, and select the obstacle avoidance path with the smallest cost value as the optimal obstacle avoidance path.

7. The multimodal real-time obstacle avoidance method for wafer transport robots according to claim 6, characterized in that, Filtering the candidate obstacle avoidance path pairs using the dynamic workspace map includes: Set an occupancy probability threshold. If the occupancy probability of any spatial unit on the candidate obstacle avoidance path is greater than or equal to the occupancy probability threshold, then the candidate obstacle avoidance path is discarded. If the occupancy probability of a spatial unit on a candidate obstacle avoidance path is less than the occupancy probability threshold, then the cost of the candidate obstacle avoidance path is calculated.

8. The multimodal real-time obstacle avoidance method for wafer transport robots according to claim 1, characterized in that, include: The optimal obstacle avoidance path is subjected to dual verification of collision risk and vibration stability.

9. A multimodal real-time obstacle avoidance system for wafer transport robots, characterized in that, The system includes: The environmental data acquisition module collects environmental data in real time through a multimodal data acquisition device to obtain multimodal environmental data; The spatial map construction module performs spatiotemporal alignment on the multimodal environment data to establish a workspace map; The dynamic trajectory prediction module collects the operational data of dynamic obstacles, inputs the operational data into the constructed trajectory prediction model to perform trajectory prediction, and obtains the trajectory prediction result; The optimal path generation module performs path planning for the wafer transport robot based on the workspace map and the trajectory prediction results, and generates the optimal obstacle avoidance path. The optimal path verification module verifies the optimal obstacle avoidance path and outputs the optimal obstacle avoidance path to the motion controller for execution.

10. The multimodal real-time obstacle avoidance system for wafer transport robots according to claim 9, characterized in that, The optimal path generation module includes: The dynamic map generation unit writes the trajectory prediction results into the workspace map at preset time intervals and assigns an occupancy probability to the spatial cell of each predicted location to obtain a dynamic workspace map. The candidate path filtering unit samples several sets of candidate velocity pairs in the velocity space, generates candidate obstacle avoidance paths based on the candidate velocity pairs, and filters the candidate obstacle avoidance path pairs through the dynamic workspace map. The optimal path selection unit defines a path cost function, calculates the cost value of the filtered candidate obstacle avoidance paths, and selects the obstacle avoidance path with the smallest cost value as the optimal obstacle avoidance path.

Images