Medical cart docking control method, system, and storage medium

Abstract

The application discloses a medical trolley docking control method and system and a storage medium. The method comprises the following steps: acquiring a target reference point cloud model and a current frame point cloud set; calculating an effectiveness score and a key point position score; constructing an optimization objective function by taking the effectiveness score and the key point position score as weight factors, solving a relative pose, and driving the medical trolley to move; driving a conical locking pin into contact with a V-shaped guide groove, absorbing positioning errors by using a passive translation degree of freedom, and completing docking. The application can effectively eliminate point cloud noise interference and stably complete high-precision stress-free automatic docking.

Description

[0001] Medical trolley docking control method, system and storage medium Technical Field

[0002] This application relates to the field of medical robot and automation equipment technology, specifically to a method, system and storage medium for docking control of medical trolleys. Background Technology

[0003] In medical environments, mobile device docking is often limited by the high reflectivity of metal equipment, leading to distortion points and drift noise in the laser point cloud, causing divergence in pose calculation. In addition, existing mobile chassis have control errors due to mechanical clearances. Under these errors, rigid docking is prone to physical interference and jamming, failing to meet the requirements of extremely high-precision stress-free docking. Summary of the Invention

[0004] This application provides a method, system, and storage medium for docking and controlling medical trolleys, which solves the problem of point cloud distortion and the limits of physical control accuracy.

[0005] This application provides a medical trolley docking control method based on a dual scoring function and a floating mechanism, applicable to a medical trolley with a chassis controller and a mechanical docking assembly including a passive floating base plate and a conical locking pin. The method includes: acquiring a reference point cloud model of the target docking base and acquiring a current frame point cloud set scanned by a line laser profile sensor; extracting the local normal vectors of each point in the current frame point cloud set, and calculating the effectiveness score of the current frame point cloud based on the variance of the angle between the local normal vectors of each point within its neighborhood and the laser reflection intensity echo value; extracting the inner folding vertex of the V-shaped guide groove on the target docking base from the reference point cloud model, and determining the effectiveness score when it is greater than a preset value. In the point cloud of the scoring threshold, the three-dimensional Euclidean distance and curvature similarity between the current point cloud and the vertex of the inner corner are calculated to obtain the key point score; the validity score and the key point score are used as weighting factors to construct a nonlinear optimization objective function to solve the relative pose of the medical trolley relative to the target docking base, and the medical trolley is driven to move according to the relative pose; the conical locking pin of the mechanical docking assembly is driven to extend, so that the conical locking pin contacts the V-shaped guide groove of the target docking base, and the positioning error after the medical trolley stops moving is absorbed by the passive translational degree of freedom of the mechanical docking assembly in the horizontal direction to complete the docking.

[0006] Specifically, the process of calculating the validity score of the current frame point cloud includes: obtaining point p in the current frame point cloud set. i Local normal vector And calculate the local normal vector. The variance of the angle between the vector and the local normal vector of each point in its neighborhood ; Obtain the point p i The corresponding laser reflection intensity echo value I i The laser reflection intensity echo value I is processed by a preset attenuation function. i Obtain the intensity penalty term and the variance of the included angle. The reciprocal of the product of the product and the intensity penalty term is used to obtain the point p. i Validity score S valid,i .

[0007] Specifically, the process of constructing the nonlinear optimization objective function and solving the relative pose is configured as follows: the effectiveness score S is... valid,i With the key point score S key,i Multiplying yields the product for point p. i The overall weight W i Construct an error function with the rotation matrix R and translation vector t of the relative pose as optimization variables. Where pi is a point in the current frame point cloud set, q i To match the inner corner vertex; the Levenberg-Marquardt algorithm is used to iteratively solve the error function E(R,t), and a damping factor is introduced in the Levenberg-Marquardt algorithm. The configuration is dynamically updated based on the residual decrease rate of each iteration until the increment of the error function is less than a preset convergence threshold.

[0008] Specifically, the mechanical docking assembly also includes a cross-roller guide and a return spring; the passive floating base plate is mounted on the front end of the medical trolley via the cross-roller guide, providing a physical translation tolerance range of ±8 mm; the passive floating base plate is held in a central reference position by the elastic constraint of the return spring under no external force compression.

[0009] Specifically, the process of absorbing positioning errors by utilizing the passive translational degree of freedom of the mechanical docking assembly in the horizontal direction to complete the docking is configured as follows: when the calculated residual of the relative pose is less than the preset distance tolerance, the chassis controller is controlled to stop the movement of the medical trolley; the tapered locking pin is driven to extend along the set longitudinal axis and insert into the V-shaped guide groove; during the insertion process, the outer conical surface of the tapered locking pin and the inner inclined surface of the V-shaped guide groove physically interfere and squeeze, and the horizontal component force generated by this interference and squeezing overcomes the tension of the reset spring, driving the passive floating base plate to slide laterally along the cross-roller guide rail, compensating for the mechanical residual error generated when the medical trolley stops moving.

[0010] This application embodiment also provides a high-precision docking system for a medical trolley, including: a line laser profile sensor for acquiring a reference point cloud model of a target docking base and acquiring a current frame point cloud set; a mechanical docking assembly, mounted on the medical trolley, including a passive floating base plate and a tapered locking pin; and a chassis controller, communicatively connected to the line laser profile sensor and the mechanical docking assembly, configured to execute the following control logic: extracting the local normal vectors of each point in the current frame point cloud set, calculating the validity score of the current frame point cloud based on the variance of the angle between the local normal vectors of each point within the neighborhood of the point and the laser reflection intensity echo value; and extracting the target docking base from the reference point cloud model. For the inner corner vertex of the V-shaped guide groove, in the point cloud where the effectiveness score is greater than a preset score threshold, the three-dimensional Euclidean distance and curvature similarity between the current point cloud and the inner corner vertex are calculated to obtain the key point score. The effectiveness score and the key point score are used as weighting factors to construct a nonlinear optimization objective function to solve the relative pose of the medical trolley relative to the target docking base, and the medical trolley is driven to move according to the relative pose. The conical locking pin of the mechanical docking assembly is driven to extend, so that the conical locking pin contacts the V-shaped guide groove of the target docking base. The passive translational degree of freedom of the mechanical docking assembly in the horizontal direction is used to absorb the positioning error after the medical trolley stops moving and complete the docking.

[0011] Specifically, the chassis controller is configured to calculate the effectiveness score by obtaining the midpoint p in the current frame point cloud set. i Local normal vector And calculate the local normal vector. The variance of the angle between the vector and the local normal vector of each point in its neighborhood ; Obtain the point p i The corresponding laser reflection intensity echo value I i The laser reflection intensity echo value I is processed by a preset attenuation function. i Obtain the intensity penalty term and the variance of the included angle. The reciprocal of the product is multiplied by the intensity penalty term to obtain the effectiveness score.

[0012] Specifically, the mechanical docking assembly also includes a cross-roller guide and a return spring; the passive floating base plate is mounted through the cross-roller guide, providing a lateral physical translation tolerance range; the passive floating base plate is constrained by the return spring and kept in a central reference position under the condition of no external force compression.

[0013] Specifically, the chassis controller uses an iterative algorithm for nonlinear optimization when solving the relative pose, and the product of the effectiveness score and the key point score is configured as a dynamic confidence weight factor for each input data point in the error optimization function.

[0014] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described medical trolley docking control method.

[0015] In summary, this application effectively eliminates point cloud noise interference caused by high-density metal reflection through a dual scoring function system, ensuring the convergence and stability of pose calculation in the complex environment of the operating room. By introducing a passive mechanical floating docking mechanism with horizontal degrees of freedom at the hardware level, the inherent mechanical clearance error of the mobile chassis is directly absorbed. Based on this synergistic coupling mechanism of algorithm-guided positioning and mechanical floating correction, this application stably achieves high-precision stress-free physical docking under strong interference environments. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the overall structure of the high-precision docking system for medical trolleys provided in an embodiment of the present invention.

[0017] Figure 2 This is a partial three-dimensional structural cross-sectional view of the mechanical docking assembly provided in an embodiment of the present invention.

[0018] Figure 3 This is a block diagram of the logic structure of the control system provided in an embodiment of the present invention.

[0019] Figure 4 This is a flowchart illustrating the medical trolley docking control method provided in an embodiment of the present invention.

[0020] Figure 5 This is a schematic diagram illustrating the flow of the validity score calculation steps provided in the embodiments of the present invention.

[0021] Explanation of reference numerals in the attached figures

[0022] In the diagram: 101-Medical trolley, 102-Target docking base, 103-Line laser profile sensor, 104-Chassis controller, 200-Mechanical docking assembly, 201-Fixing frame, 202-Passive floating base plate, 203-Conical locking pin, 204-V-shaped guide groove, 205-Cross-shaped roller guide, 206-Reset spring, 207-Push rod motor, 301-Data acquisition unit, 302-Effect evaluation unit, 303-Feature matching unit, 304-Pose calculation unit, 305-Actuation drive unit. Detailed Implementation

[0023] To make the technical solution, structural features, and beneficial effects of this application clearer and more explicit, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0024] like Figure 1 As shown, this application embodiment provides a high-precision docking system for a medical trolley. This high-precision docking system is mainly deployed on a medical trolley 101 in the medical field. The system includes a chassis controller 104, a line laser profile sensor 103, and a front-end mechanical docking assembly 200. The line laser profile sensor 103 is a short-baseline line laser profile sensor. The line laser profile sensor 103 is used to acquire the reference point cloud model of the target docking base 102 and to acquire the current frame point cloud set during actual operation. The line laser profile sensor 103 is installed at the lower front end of the medical trolley 101, with a ground clearance of 150mm to 200mm, and its scanning field of view is horizontally oriented towards the target docking base 102. By setting the ground clearance to 150mm to 200mm, direct interference from most high-suspended reflective instruments in the ward can be effectively avoided, and the system is precisely aligned with the effective characteristic height range of the target docking base 102. The chassis controller 104 is communicatively connected to the line laser profile sensor 103 and the mechanical docking assembly 200. In addition, as a fundamental support for the system, the medical trolley 101 is equipped with a chassis servo drive motor and its related motion control bus circuitry to execute movement commands issued by the chassis controller 104. Simultaneously, the system is also equipped with a power conversion module to maintain stable operation of the line laser profile sensor 103 and the chassis controller 104. It should be noted that the specific model of the line laser profile sensor 103 does not constitute the sole limitation of this application; those skilled in the art can also achieve the functions of this application using a high-frequency structured light sensor with similar performance.

[0025] like Figure 2As shown, the mechanical docking assembly 200 includes a fixing frame 201, a passive floating base plate 202, and two sets of asymmetrically arranged conical locking pins 203. The fixing frame 201 is rigidly connected to the chassis structure of the medical trolley 101. The passive floating base plate 202 is connected to the fixing frame 201 via four sets of cross-roller guides 205 and a return spring 206. By employing this specific mechanical constraint structure of four sets of cross-roller guides 205, the passive floating base plate 202 is allowed to have a passive translational freedom of ±8mm in the horizontal XY plane. Under no external force compression, the passive floating base plate 202 is held in a central reference position by the elastic constraint of the return spring 206. On the surface of the target docking base 102, a V-shaped guide groove 204 is correspondingly provided to cooperate with the conical locking pins 203. The inlet chamfer angle of the V-shaped guide groove 204 is set between 15 degrees and 20 degrees. Before docking, the high-precision docking system for the medical trolley relies on the aforementioned pure kinematic control to move the medical trolley 101 towards the target docking base 102. Limited by the friction fluctuations of the existing chassis tires and the mechanical backlash of the servo motor gears, the theoretical stopping accuracy of pure kinematic control is typically around ±10mm. In this embodiment, the chassis controller 104 receives point cloud data from the laser sensor, executes the dual scoring function algorithm (described in detail later), and outputs drive commands to control the movement of the medical trolley 101 chassis until the tapered locking pin 203 is fully inserted into the V-shaped guide groove 204.

[0026] like Figure 3 As shown, the chassis controller 104 can be logically divided into a data acquisition unit 301, an effectiveness evaluation unit 302, a feature matching unit 303, a pose calculation unit 304, and an execution drive unit 305. These units interact with each other via an internal high-speed data bus to coordinate instructions and calculation matrices. Based on the above hardware architecture and control logic units, the chassis controller 104 is configured to execute the core docking control method.

[0027] Combination Figure 4 The control flow shown in this application embodiment provides a medical trolley docking control method based on a dual scoring function and a floating mechanism. The method is executed by the aforementioned chassis controller 104 and specifically includes the following processing steps.

[0028] In step S401, the chassis controller 104 acquires the reference point cloud model of the target docking base 102 and the current frame point cloud set scanned by the line laser profile sensor 103. The acquired data will be cached in the high-speed random access memory of the chassis controller 104 for subsequent low-latency calculations.

[0029] In step S402, the local normal vectors of each point in the current frame point cloud set are extracted. Based on the variance of the angle between the local normal vectors of each point within its neighborhood and the laser reflection intensity echo value, the validity score of the current frame point cloud is calculated. The medical operating room environment contains a large number of surface-polished metal devices, which generate significant specular reflection and multipath effects. By evaluating the validity score, the chassis controller 104 can filter out reliable physical surface point clouds in real time.

[0030] Specifically, the process of calculating the validity score of the current frame point cloud includes the following sub-steps. In step S501, the chassis controller 104 acquires a specific point p in the current frame point cloud set. i Local normal vector And calculate the local normal vector. The variance of the angle between the vector and the local normal vector of each point in its neighborhood Angle variance This reflects the degree of normal fluctuation on the point cloud surface in a local region. Typically, a continuous, smooth solid surface has a very small variance in the normal angle. In step S502, the system synchronously acquires point p. i The corresponding laser reflection intensity echo value I i In step S503, the laser reflection intensity echo value I is processed by a preset attenuation function. i Obtain the intensity penalty term and the included angle variance. Multiplying the reciprocal of the product of the strength penalty term and the product of ... and the product of the i Validity score S valid,i .

[0031] In this embodiment, the calculation logic of the validity score is deeply integrated with the saturation characteristics of laser echo. The chassis controller 104 sets that when the variance of the angle between the normal vectors of all points in the neighborhood of a point is greater than a preset variance threshold, and the laser reflection intensity echo value I of that area is... i When the value is greater than 240, the system determines that the laser reflection intensity echo saturation range threshold has reached the trigger state. At this time, the chassis controller 104 determines that the area is a false multipath point cloud caused by high metal reflectivity. For such false point clouds, the chassis controller 104 will directly assign a validity score close to 0 through an attenuation function. Conversely, for physical areas with continuous and smooth normal vector variance and normal echo intensity, the system assigns a higher validity score S. valid,i By employing a calculation method that combines curvature variance and echo intensity threshold, the logical characteristics of the random distribution of normals in multipath drift point clouds can be utilized to achieve efficient and delay-free removal of complex mirror noise.

[0032] In step S403, the chassis controller 104 extracts key feature points of the V-shaped guide groove 204 on the target docking base 102 from the reference point cloud model. Within the current point cloud set where the validity score is greater than a preset score threshold, it calculates the three-dimensional Euclidean distance and curvature similarity between the current point cloud and the key feature points to obtain a key point score. In this embodiment, the system specifically selects the inner folded corner vertex of the V-shaped guide groove 204 as a high-confidence key feature point. Specifically, the system extracts local geometric topological features from the valid point cloud using a fast point feature histogram algorithm, completes matching, and outputs the key point score S for each point. key,i This mechanism, which relies on strong geometric features to output key point scores, provides a reliable benchmark for the rapid convergence of the subsequent objective function. It should be noted that the key feature point extraction in this application is not limited to inner corner vertices; those skilled in the art can also use the groove wall edge lines as topological features to achieve this matching process.

[0033] In step S404, the chassis controller 104 uses the effectiveness score and key point score as dynamic confidence weighting factors to construct a nonlinear optimization objective function, solves the relative pose of the medical trolley 101 relative to the target docking base 102, and drives the medical trolley 101 to move according to the relative pose.

[0034] Specifically, the chassis controller 104 will target point p i Validity score S valid,i Key Point Score key,i Multiply to obtain the product for point p i The overall weight W i The system then constructs an error function with the rotation matrix R and translation vector t of the relative pose as optimization variables. Specifically, the system performs the calculation of the total cost objective function, and the above error optimization logic is implemented through the following mathematical formula:

[0035]

[0036] The equivalent form of this formula can also be expressed as: In the above formula, E(R,t) represents the total error optimization objective function; R represents the relative pose rotation matrix to be solved; t represents the relative pose translation vector to be solved; N represents the total number of valid point clouds participating in the optimization; W v (i) The effectiveness score S corresponding to the above point cloud valid,i W k (i) Corresponding to the above key point score S key,i The product of the two corresponds to the overall weight W. i ;p source,i Or rather, p i This represents the current frame source point cloud of the input data; ptarget,i Or rather, the matching q i This refers to the target inner corner vertex that has been matched in the model.

[0037] After constructing the objective function, the chassis controller 104 uses the Levonberg-Marquardt algorithm to iteratively solve the error function E(R,t). During this nonlinear iterative optimization process, highly reflective metallic noise points, due to their extremely low effectiveness scores, have a comprehensive weight approaching zero and are safely excluded from the contribution region of the gradient descent calculation. In this solution logic, the damping factor introduced in the Levonberg-Marquardt algorithm... The chassis controller 104 is configured to dynamically and adaptively update the residual based on the rate of decrease in each iteration step, ensuring rapid positive definiteness of the Hessian matrix until the iteration increment of the error function E(R,t) decreases to less than a preset convergence threshold. By employing a Levenberg-Marquardt iterative solution algorithm with a double-weighted penalty term, the computational characteristics of stripping gradient contributions from low-scoring data can be utilized, ensuring that the accurate pose data solved by the system does not undergo mathematical nonlinear shifts. The chassis controller 104 then guides the movement of the medical trolley 101 chassis via a chassis servo drive motor based on the high-precision pose data output by the algorithm.

[0038] In step S405, when the calculated residual of the relative pose is less than the preset distance tolerance, the chassis controller 104 is controlled to stop the movement of the medical cart 101. At this time, due to the grip friction of the rubber tires and the mechanical clearance of the motor, the final physical stopping pose guided by the algorithm layer usually still has a physical residual error. According to the system calibration, this physical residual error falls within the range of 3mm to 5mm. At this time, the system drives the conical locking pin 203 of the mechanical docking assembly 200 to extend. Specifically, the chassis controller 104 drives the push rod motor 207 to extend the conical locking pin 203, so that the conical locking pin 203 contacts the V-shaped guide groove 204 of the target docking base 102. By utilizing the passive translational degree of freedom of the mechanical docking assembly 200 in the horizontal direction, the positioning error after the medical cart 101 stops moving is absorbed to complete the final rigid docking.

[0039] In practice, the chassis controller 104 sends a command to drive the asymmetrically arranged tapered locking pins 203 to extend forward along a set longitudinal axis. During the insertion process, the outer conical surface of the tapered locking pins 203 comes into physical contact and interferes with the inner 15- to 20-degree chamfered slope of the V-shaped guide groove 204. The horizontal component force generated by this interference compression directly forces the passive floating base plate 202 to overcome the reverse tension of the return spring 206, thereby driving the passive floating base plate 202 to produce a slight lateral slip along the cross roller guide 205. This lateral slip achieved through passive mechanical degrees of freedom accurately compensates for and absorbs physical residual errors of 3 mm to 5 mm, ultimately allowing the tapered locking pins 203 to fall smoothly and completely into the bottom of the V-shaped guide groove 204. By employing a mechanical docking assembly 200 with cross-roller guides 205 in conjunction with a chamfered V-shaped guide groove 204, the mechanical adaptive characteristic of converting inclined plane extrusion force into lateral displacement can be utilized to achieve stress-free closure at the final execution end, meeting the medical-grade 2mm system docking accuracy requirements. Without the anti-interference guidance of the algorithm layer, pure kinematic errors would cause the conical locking pin to fall outside the ±8mm tolerance range, resulting in a hard impact; without the physical compensation slippage of the mechanical layer, algorithms alone cannot eliminate the execution dead zone of the underlying hardware. Therefore, the above steps and physical structure constitute an inseparable closed-loop synergy.

[0040] This application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program. When the computer program is loaded and executed by the processing chip built into the chassis controller 104, the medical trolley docking control method based on a dual scoring function, as described in steps S401 to S405 of the above embodiments, can be implemented. The computer-readable storage medium can be a read-only memory, random access memory, flash memory chip, or any other form of portable digital storage medium. Those skilled in the art should understand that any equivalent compilation or refactoring of the computer program, as long as it enables the automatic docking of the drive mechanical assembly with the algorithm, falls within the protection scope of this application.

Claims

1. A medical trolley docking control method based on dual scoring functions and a floating mechanism, applied to a medical trolley having a chassis controller and a mechanical docking assembly including a passive floating base plate and a conical locking pin, characterized in that, This includes acquiring a reference point cloud model of the target docking base and acquiring a current frame point cloud set scanned by a line laser profile sensor; extracting the local normal vectors of each point in the current frame point cloud set, and calculating the validity score of the current frame point cloud based on the variance of the angle between the local normal vectors of each point within its neighborhood and the laser reflection intensity echo value; extracting the inner corner vertex of the V-shaped guide groove on the target docking base from the reference point cloud model, and calculating the three-dimensional Euclidean distance and curvature similarity between the current point cloud and the inner corner vertex in the point cloud with a validity score greater than a preset score threshold. The system obtains key point scores; uses the effectiveness scores and key point scores as weighting factors to construct a nonlinear optimization objective function, solves for the relative pose of the medical cart relative to the target docking base, and drives the medical cart to move according to the relative pose; drives the conical locking pin of the mechanical docking assembly to extend, so that the conical locking pin contacts the V-shaped guide groove of the target docking base, and uses the passive translational degree of freedom of the mechanical docking assembly in the horizontal direction to absorb the positioning error after the medical cart stops moving to complete the docking.

2. The medical trolley docking control method as described in claim 1, characterized in that, The process of calculating the validity score of the current frame point cloud includes obtaining point p in the current frame point cloud set. i Local normal vector And calculate the local normal vector. The variance of the angle between the vector and the local normal vector of each point in its neighborhood ; Obtain the point p i The corresponding laser reflection intensity echo value I i The laser reflection intensity echo value I is processed by a preset attenuation function. i Obtain the intensity penalty term and the variance of the included angle. The reciprocal of the product of the product and the intensity penalty term is used to obtain the point p. i Validity score S valid,i .

3. The medical trolley docking control method as described in claim 2, characterized in that, The process of constructing the nonlinear optimization objective function and solving the relative pose is configured to include the effectiveness score S. valid,i With the key point score S key,i Multiplying yields the product for point p. i The overall weight W i Construct an error function with the rotation matrix R and translation vector t of the relative pose as optimization variables. , where p i For the points in the current frame point cloud set, q i For the matched inner corner vertex; The error function E(R,t) is solved iteratively using the Levenberg-Marquardt algorithm, and a damping factor is introduced in the Levenberg-Marquardt algorithm. The configuration is dynamically updated based on the residual decrease rate of each iteration until the increment of the error function is less than a preset convergence threshold.

4. The medical trolley docking control method as described in claim 1, characterized in that, The mechanical docking assembly also includes a cross-roller guide and a return spring; the passive floating base plate is mounted on the front end of the medical trolley via the cross-roller guide, providing a physical translation tolerance range of ±8 mm; the passive floating base plate is held in the central reference position by the elastic constraint of the return spring under the condition of no external force compression.

5. The medical trolley docking control method as described in claim 4, characterized in that, The process of using the passive translational degree of freedom of the mechanical docking assembly in the horizontal direction to absorb positioning errors and complete docking is configured as follows: when the calculated residual of the relative pose is less than the preset distance tolerance, the chassis controller is controlled to stop the movement of the medical trolley; the conical locking pin is driven to extend along the set longitudinal axis and insert into the V-shaped guide groove; during the insertion process, the outer conical surface of the conical locking pin and the inner inclined surface of the V-shaped guide groove physically interfere and squeeze, and the horizontal component force generated by the interference and squeezing overcomes the tension of the reset spring, driving the passive floating base plate to slide laterally along the cross-roller guide rail, compensating for the mechanical residual error generated when the medical trolley stops moving.

6. A medical trolley docking system, characterized in that, The system includes a line laser profile sensor for acquiring a reference point cloud model of the target docking base and for acquiring the current frame point cloud set; a mechanical docking assembly mounted on a medical trolley, comprising a passive floating base plate and a conical locking pin; and a chassis controller, communicatively connected to the line laser profile sensor and the mechanical docking assembly, configured to execute control logic to extract the local normal vectors of each point in the current frame point cloud set, and to calculate the validity score of the current frame point cloud based on the variance of the angle between the local normal vectors of each point within its neighborhood and the laser reflection intensity echo value; and to extract the inner folding vertex of the V-shaped guide groove on the target docking base from the reference point cloud model, and to determine the validity score when the validity score is greater than a preset value. In the point cloud with thresholds, the three-dimensional Euclidean distance and curvature similarity between the current point cloud and the vertex of the inner corner are calculated to obtain the key point score; the validity score and the key point score are used as weighting factors to construct a nonlinear optimization objective function to solve the relative pose of the medical trolley relative to the target docking base, and the medical trolley is driven to move according to the relative pose; the conical locking pin of the mechanical docking assembly is driven to extend, so that the conical locking pin contacts the V-shaped guide groove of the target docking base, and the positioning error after the medical trolley stops moving is absorbed by the passive translational degree of freedom of the mechanical docking assembly in the horizontal direction to complete the docking.

7. The medical trolley docking system as described in claim 6, characterized in that, The chassis controller is configured to calculate the validity score and obtain the point p in the current frame point cloud set in the following manner. i Local normal vector And calculate the local normal vector. The variance of the angle between the vector and the local normal vector of each point in its neighborhood ; Obtain the point p i The corresponding laser reflection intensity echo value I i The laser reflection intensity echo value I is processed by a preset attenuation function. i Obtain the intensity penalty term and the variance of the included angle. The reciprocal of the product is multiplied by the intensity penalty term to obtain the effectiveness score.

8. The medical trolley docking system as described in claim 6, characterized in that, The mechanical docking assembly also includes a cross-roller guide and a return spring; the passive floating base plate is mounted via the cross-roller guide, providing a physical translation tolerance range of ±8 mm; the passive floating base plate is constrained by the return spring and kept in the central reference position under the condition of no external force compression.

9. The medical trolley docking system as described in claim 6, characterized in that, The chassis controller employs an iterative algorithm for nonlinear optimization when solving for relative pose, and incorporates the product of the effectiveness score and the key point score as a dynamic confidence weight factor for each input data point into the error optimization function.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the medical trolley docking control method as described in claim 1.

Images