Method, device and computer equipment for authorized release of medical robot dynamic control logic

By receiving and inspecting medical robot task objects through a unified entry point, identifying the source and change type of dynamic control logic, conducting sandbox verification and risk scoring, and matching authorization levels, the problem of binding authorization and dynamic release in existing medical robot systems is solved, realizing full-link automated authorization management and security assurance.

CN123069334APending Publication Date: 2026-11-10CHANGZHOU QINGZHI INTELLIGENT TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611024693.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-07-10
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing medical robot systems lack a binding mechanism for authorization and dynamic release, which makes it impossible to match tasks with the precise authorization of medical personnel at the corresponding level, and to achieve deep coupling between technology verification and clinical authorization decision-making and full-chain automated authorization control.

Method used

The system receives medical robot task objects through a unified entry point, performs a general boundary check, identifies the source type and change granularity type of dynamic control logic, generates a unique identifier and content hash value, conducts sandbox verification and authorization management, obtains multi-dimensional assessment data for risk scoring, matches authorization levels, continuously verifies the authorization status during execution, and records key events to form an immutable chain of evidence.

Benefits of technology

It enables differentiated control over parameter-level, node-level, topology-level, behavior tree-level, and version-level changes, ensuring that authorization matches risk, preventing unauthorized modifications, providing differentiated security responses, supporting fully auditable and traceable authorization management, and meeting the compliance requirements of medical device regulatory agencies.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123069334A_ABST
    Figure CN123069334A_ABST
Patent Text Reader

Abstract

The application relates to an authorization issuing method, device and computer equipment for dynamic control logic of a medical robot. The method comprises the following steps: receiving a task object and performing total boundary checking to lock an execution range; identifying a source type and a change granularity type of the dynamic control logic, judging whether to trigger re-sandbox verification and authorization; performing sandbox verification in an isolated simulation environment; obtaining multi-dimensional evaluation data for weighted risk scoring after verification, mapping to a risk level, matching an authorization level to generate an authorization application; pushing to an authorized terminal and performing corresponding operations; performing version access control on the authorized control logic, generating a release conclusion and performing corresponding release operations; continuously checking the authorization state during execution; and recording key events in the whole process in an unalterable format to form an evidence chain. Therefore, the whole-process authorization control of the medical robot control logic from generation to execution is realized, and the safety and traceability of clinical deployment are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the fields of medical robotics and artificial intelligence, and in particular to a licensed method, apparatus, computer device, computer-readable storage medium, and computer program product for dynamic control logic of a medical robot. Background Technology

[0002] With the deep integration of artificial intelligence, adaptive control, digital twins and medical robotics, medical devices such as rehabilitation robots, surgical robots and assisted nursing robots are evolving from "fixed program executors" to "dynamic control logic execution platforms".

[0003] In traditional technologies, technology validation and clinical authorization decisions are considered two separate processes, with different tasks requiring authorization from doctors, therapists, nurses, or administrators. However, existing robotic systems lack authorization and dynamic release binding mechanisms, resulting in tasks failing to match precise authorization from medical personnel at the appropriate levels.

[0004] Therefore, there is an urgent need for a solution that can deeply couple technology validation with clinical licensing decisions to achieve automated licensing management across the entire process. Summary of the Invention

[0005] Based on this, it is necessary to provide a method, device, computer equipment, computer-readable storage medium, and computer program product for licensing and issuing dynamic control logic of medical robots that can deeply couple technology verification with clinical licensing decisions and realize automated licensing management across the entire chain, addressing the aforementioned technical problems.

[0006] In a first aspect, this application provides a method for distributing dynamic control logic for a medical robot, the method comprising:

[0007] The system receives medical robot task objects through a unified entry point, performs a total boundary check on the control parameters in the task objects, locks the control parameters that pass the check as the execution range, and generates a task identifier.

[0008] Identify the source type and change granularity type of the dynamic control logic to be released, generate a unique identifier, version information and content hash value of the dynamic control logic, and determine whether to trigger re-sandbox verification and re-authorization based on the change granularity type;

[0009] The dynamic control logic is subjected to sandbox verification in an isolated simulation environment to obtain the verification results;

[0010] If the verification result is passed, assessment data from multiple dimensions is obtained, a risk score is calculated by weighting the assessment data from multiple dimensions, the risk score is mapped to the corresponding risk level, and the corresponding authorization level is matched according to the risk level to generate an authorization application.

[0011] The authorization request is pushed to the matching authorization user's terminal, the authorization operation instruction returned by the authorization user is received, and the corresponding operation is performed according to the authorization operation instruction;

[0012] The system performs version access control audits on the authorized dynamic control logic, verifies the validity of the authorization, and generates a release conclusion that allows, restricts, or prohibits release based on the audit results.

[0013] The corresponding release operation is executed according to the release conclusion, and the authorization status is continuously verified during the execution of the dynamic control logic. When the authorization fails, the corresponding security response action is triggered according to the task risk level and the current execution status.

[0014] The key events and operation records of the dynamic control logic from task access to execution completion are recorded and stored in an immutable format to form an evidence chain, which is used to support audit traceability and version rollback management.

[0015] In one embodiment, the source type includes any one or more of the following: real-time generation by external artificial intelligence (AI), rule engine generation, combination of medical task templates, parameter adjustment, doctor modification authorization, developer resubmission, and user script configuration.

[0016] The types of change granularity include parameter-level changes, node-level changes, topology-level changes, behavior tree-level changes, security boundary-level changes, and version-level changes.

[0017] The step of determining whether to trigger re-sandbox verification and re-authorization based on the change granularity type includes:

[0018] If the change granularity type is a parameter-level change, and the parameter-level change is within a preset security domain, then re-sandbox verification and re-authorization are exempted.

[0019] If the change granularity type is a parameter-level change, and the parameter-level change is not within the preset security domain, then re-sandbox verification and re-authorization are triggered.

[0020] If the change granularity type is any of node-level change, topology-level change, behavior tree-level change, security boundary-level change, or version-level change, re-sandbox verification and re-authorization will be triggered.

[0021] In one embodiment, the assessment data of the multiple dimensions includes six dimensions: patient harm probability, task invasiveness, failure consequences, patient status, device contact level, and the change granularity type.

[0022] The risk levels include four levels: R1, R2, R3, and R4, arranged from low risk to high risk.

[0023] The process of matching the corresponding authorization level based on the risk level includes:

[0024] R1 level matching system automatically authorizes or conducts random checks afterward;

[0025] R2 level matching requires authorization from the attending physician, rehabilitation therapist, or nursing station.

[0026] R3 level matching pre-approval authorization from attending physician or senior physician;

[0027] R4 level matching is prohibited from publishing by default.

[0028] In one embodiment, the authorization operation instruction includes one or more of the following: a confirmation instruction, a restriction confirmation instruction, a modification authorization instruction, and a denial instruction;

[0029] The step of executing the corresponding operation according to the authorized operation instruction includes:

[0030] Upon receiving the confirmation instruction, perform version access control verification;

[0031] Upon receiving the restriction confirmation instruction, record the additional constraints imposed on the authorized person carried in the restriction confirmation instruction, and perform version access control audit;

[0032] Upon receiving the modification authorization instruction, the system checks whether the modified content exceeds the preset authorization boundary. If it does, the modification is rejected and a resubmission is prompted. If it does not exceed the boundary, the modified content is recorded and version access control review is performed.

[0033] Upon receiving the rejection instruction, the reason for rejection is recorded and a feedback loop is initiated. The reason for rejection is written into the evidence chain and fed back to the developer, doctor, or task generator. It is allowed to modify the control logic based on the reason for rejection. The modified control logic needs to be re-verified through the sandbox. Only after the sandbox verification is passed can the authorization request be generated again.

[0034] In one embodiment, the preset authorization boundary includes: modifiable items and non-modifiable items;

[0035] The modifiable items include one or more of the following: execution time, execution location, low-risk training duration, number of actions, voice prompt content, reminder frequency, target ward, low-risk navigation target, and authorization period;

[0036] The unmodifiable items include one or more of the following: safety hard limits, maximum speed, maximum torque, maximum contact force, patient safety distance, prohibited action library, R4 risk level tasks, control logic nodes that have not passed sandbox verification, high-risk topology, emergency stop logic, doctor takeover logic, and registration verification boundaries.

[0037] In one embodiment, the verification of authorization validity includes at least one of the following: verifying whether the current authorization has timed out, verifying whether the current authorization has expired, verifying whether the current authorization has been revoked, and verifying whether the scope of the current authorization matches the current task.

[0038] During the authorization waiting period, monitor the response time of the authorizing party. If an authorization timeout occurs, perform the following differentiated processing:

[0039] If a task at risk level R1 fails to respond within the timeout period, it will be subject to post-event spot checks or automatic downgrade execution.

[0040] If an R2 risk level task fails to respond within the timeout period, it will be placed in the manual review queue or its release will be restricted.

[0041] When an R3 risk level task times out without a response, it is prohibited to publish and notify higher-privilege roles.

[0042] Among them, R4 risk level tasks must not be automatically published due to timeout;

[0043] In scenarios involving emergency rescue or preventing further harm from the equipment, the method further includes:

[0044] Temporary authorization is initiated by a higher-privilege role. The temporary authorization only allows the execution of safety actions such as stopping, evacuating, slow navigation, safe rollback, clearing dangerous situations, and notifying personnel. It prohibits the addition of new diagnoses, prescriptions, treatment parameter adjustments, intrusion operations, or unverified control logic.

[0045] In one embodiment, the restricted release includes adding one or more of the following constraints: low-speed operation only, position control only, force control output prohibited, sandbox operation only, and physician presence required;

[0046] The provision that triggers a corresponding security response action based on the task risk level and current execution status when authorization expires includes:

[0047] New tasks are prohibited from starting if the authorization has expired and the task has not yet started.

[0048] If the authorization has expired and a low-risk, uninterruptible step is in progress, stop subsequent execution after completing the current step;

[0049] When authorization is revoked or R3 risk level task authorization expires, a safety stop or soft stop is triggered, and the system reverts to the previous safety control state or switches to doctor takeover.

[0050] When high-risk outputs involve contact actions, force-controlled actions, or physical therapy outputs, and the authorization has expired, stop the outputs, retain monitoring data, and generate alarm notifications.

[0051] In one embodiment, after obtaining the verification result, the method further includes:

[0052] If the verification result fails, the generation of authorization requests will be prohibited and the process will be terminated.

[0053] If the verification result is a warning, a restricted authorization request with limiting constraints is generated. These limiting constraints are used to attach to the dynamic control logic to be released in subsequent release steps.

[0054] Secondly, this application also provides an authorized distribution device for dynamic control logic of a medical robot, the device comprising:

[0055] The task access and boundary control module is used to receive medical robot task objects through a unified entry point, perform a total boundary check on the control parameters in the task objects, lock the control parameters that pass the check as the execution range, and generate a task identifier.

[0056] The control logic identification module is used to identify the source type and change granularity type of the dynamic control logic to be released, generate a unique identifier, version information and content hash value of the dynamic control logic, and determine whether to trigger re-sandbox verification and re-authorization based on the change granularity type.

[0057] The sandbox verification and admission module is used to perform sandbox verification on the dynamic control logic in an isolated simulation environment and obtain the verification results.

[0058] The risk assessment module is used to acquire assessment data from multiple dimensions when the verification result is passed, calculate a risk score based on the weighted assessment data from multiple dimensions, map the risk score to the corresponding risk level, match the corresponding authorization level according to the risk level, and generate an authorization application.

[0059] The authorization decision module is used to push the authorization application to the matching authorizing person's terminal, receive the authorization operation instruction returned by the authorizing person, and execute the corresponding operation according to the authorization operation instruction;

[0060] The review module is used to perform version access control review on the authorized dynamic control logic, verify the validity of the authorization, and generate a release conclusion that allows release, restricts release, or prohibits release based on the review results;

[0061] The release and execution control module is used to execute the corresponding release operation according to the release conclusion, and continuously verify the authorization status during the execution of the dynamic control logic. When the authorization fails, the corresponding security response action is triggered according to the task risk level and the current execution status.

[0062] The evidence chain and audit module is used to record key events and operation records of the entire process of the dynamic control logic from task access to execution completion, and to store them in an immutable format to form an evidence chain. The evidence chain is used to support audit traceability and version rollback management.

[0063] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0064] The system receives medical robot task objects through a unified entry point, performs a total boundary check on the control parameters in the task objects, locks the control parameters that pass the check as the execution range, and generates a task identifier.

[0065] Identify the source type and change granularity type of the dynamic control logic to be released, generate a unique identifier, version information and content hash value of the dynamic control logic, and determine whether to trigger re-sandbox verification and re-authorization based on the change granularity type;

[0066] The dynamic control logic is subjected to sandbox verification in an isolated simulation environment to obtain the verification results;

[0067] If the verification result is passed, assessment data from multiple dimensions is obtained, a risk score is calculated by weighting the assessment data from multiple dimensions, the risk score is mapped to the corresponding risk level, and the corresponding authorization level is matched according to the risk level to generate an authorization application.

[0068] The authorization request is pushed to the matching authorization user's terminal, the authorization operation instruction returned by the authorization user is received, and the corresponding operation is performed according to the authorization operation instruction;

[0069] The system performs version access control audits on the authorized dynamic control logic, verifies the validity of the authorization, and generates a release conclusion that allows, restricts, or prohibits release based on the audit results.

[0070] The corresponding release operation is executed according to the release conclusion, and the authorization status is continuously verified during the execution of the dynamic control logic. When the authorization fails, the corresponding security response action is triggered according to the task risk level and the current execution status.

[0071] The key events and operation records of the dynamic control logic from task access to execution completion are recorded and stored in an immutable format to form an evidence chain, which is used to support audit traceability and version rollback management.

[0072] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0073] The system receives medical robot task objects through a unified entry point, performs a total boundary check on the control parameters in the task objects, locks the control parameters that pass the check as the execution range, and generates a task identifier.

[0074] Identify the source type and change granularity type of the dynamic control logic to be released, generate a unique identifier, version information and content hash value of the dynamic control logic, and determine whether to trigger re-sandbox verification and re-authorization based on the change granularity type;

[0075] The dynamic control logic is subjected to sandbox verification in an isolated simulation environment to obtain the verification results;

[0076] If the verification result is passed, assessment data from multiple dimensions is obtained, a risk score is calculated by weighting the assessment data from multiple dimensions, the risk score is mapped to the corresponding risk level, and the corresponding authorization level is matched according to the risk level to generate an authorization application.

[0077] The authorization request is pushed to the matching authorization user's terminal, the authorization operation instruction returned by the authorization user is received, and the corresponding operation is performed according to the authorization operation instruction;

[0078] The system performs version access control audits on the authorized dynamic control logic, verifies the validity of the authorization, and generates a release conclusion that allows, restricts, or prohibits release based on the audit results.

[0079] The corresponding release operation is executed according to the release conclusion, and the authorization status is continuously verified during the execution of the dynamic control logic. When the authorization fails, the corresponding security response action is triggered according to the task risk level and the current execution status.

[0080] The key events and operation records of the dynamic control logic from task access to execution completion are recorded and stored in an immutable format to form an evidence chain, which is used to support audit traceability and version rollback management.

[0081] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0082] The system receives medical robot task objects through a unified entry point, performs a total boundary check on the control parameters in the task objects, locks the control parameters that pass the check as the execution range, and generates a task identifier.

[0083] Identify the source type and change granularity type of the dynamic control logic to be released, generate a unique identifier, version information and content hash value of the dynamic control logic, and determine whether to trigger re-sandbox verification and re-authorization based on the change granularity type;

[0084] The dynamic control logic is subjected to sandbox verification in an isolated simulation environment to obtain the verification results;

[0085] If the verification result is passed, assessment data from multiple dimensions is obtained, a risk score is calculated by weighting the assessment data from multiple dimensions, the risk score is mapped to the corresponding risk level, and the corresponding authorization level is matched according to the risk level to generate an authorization application.

[0086] The authorization request is pushed to the matching authorization user's terminal, the authorization operation instruction returned by the authorization user is received, and the corresponding operation is performed according to the authorization operation instruction;

[0087] The system performs version access control audits on the authorized dynamic control logic, verifies the validity of the authorization, and generates a release conclusion that allows, restricts, or prohibits release based on the audit results.

[0088] The corresponding release operation is executed according to the release conclusion, and the authorization status is continuously verified during the execution of the dynamic control logic. When the authorization fails, the corresponding security response action is triggered according to the task risk level and the current execution status.

[0089] The key events and operation records of the dynamic control logic from task access to execution completion are recorded and stored in an immutable format to form an evidence chain, which is used to support audit traceability and version rollback management.

[0090] The aforementioned authorization and release methods, devices, computer equipment, computer-readable storage media, and computer program products for the dynamic control logic of medical robots receive medical robot task objects through a unified entry point and perform a general boundary check. This forcibly incorporates all dynamic control logic into gateway management, effectively preventing AI-generated or unapproved control logic from bypassing doctors and directly entering the robot execution system, thus ensuring the safety of clinical decisions. By identifying the source type and change granularity type of the control logic, and determining whether to trigger re-sandbox verification and re-authorization based on the change granularity type, differentiated management of parameter-level, node-level, topology-level, behavior tree-level, security boundary-level, and version-level changes is achieved. Parameter-level changes are exempt from re-verification and re-authorization within the security domain. This improves efficiency, while major changes force a complete re-run of the process to eliminate safety hazards. Sandbox verification is performed in an isolated simulation environment, directly prohibiting the generation of authorization requests upon verification failure and generating restricted authorization requests with constraints upon verification warnings. This enforces the binding of technical verification results with the authorization entry point, eliminating invalid authorization requests at the source. After successful verification, assessment data from six dimensions—patient harm probability, task invasiveness, failure consequences, patient status, equipment contact level, and change granularity—are used to generate a weighted risk score, mapped to four risk levels from R1 to R4, and matched with corresponding authorization levels. This allows low-risk tasks to be automatically authorized or subject to post-event spot checks, medium-risk tasks to be confirmed by the responsible physician or therapist, and high-risk tasks to be authorized automatically. The system employs a pre-screening mechanism by the attending physician and a default ban on publishing for extremely high-risk tasks, achieving a precise match between authorization levels and clinical risks. Authorization requests are pushed to the matched authorizing user's terminal, and authorization operation instructions are received. When modifying authorizations, the system verifies whether the modifications exceed the preset boundaries between modifiable and unmodifiable items, preventing unauthorized modifications to safety-critical parameters such as hard limits, maximum speed / torque, and prohibited action libraries. If authorization is rejected, the reason for rejection is written into the evidence chain and fed back to the developer or physician, requiring modifications and re-verification through a sandbox before another authorization request can be generated, forming an unbypassable security loop. Version access control is applied to the control logic of authorized users, verifying the validity of the authorization and generating a conclusion allowing, restricting, or prohibiting publishing. When releasing a robot, restrictions can be imposed, such as limiting operation to low speed, limiting position control, prohibiting force control output, allowing sandbox operation only, or requiring a doctor to be present. This enables differentiated and precise release control. During execution, the authorization status is continuously verified. When the authorization expires, differentiated safety responses are triggered based on the task risk level and the current execution status: new task startup is prohibited when the authorization has expired and the task has not started; the robot stops after completing the current step when the authorization has expired and a low-risk, uninterrupted step is in progress; the robot immediately stops safely or reverts to the previous safety control state when the authorization is revoked or the R3 task fails; and high-risk outputs are immediately stopped and monitoring alarms are retained when contact actions, force control actions, or physiotherapy outputs are involved, ensuring that the robot remains within a controllable range in any authorization failure scenario.It records key events and operations throughout the entire process from task access to execution completion, using immutable formats such as blockchain, hash chains, or digital signatures to form an evidence chain. This supports audit traceability and version rollback management, achieving full auditability and traceability from control logic generation to deployment and execution. This fully meets the compliance requirements of medical device regulatory agencies regarding post-market change control and clinical deployment authorization for AI medical software. Attached Figure Description

[0091] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0092] Figure 1 A schematic diagram illustrating an application scenario of the authorized release method for the dynamic control logic of a medical robot provided in an embodiment of this application;

[0093] Figure 2 This is a flowchart illustrating the authorization and release method for the dynamic control logic of the medical robot in the first embodiment;

[0094] Figure 3 This is a flowchart illustrating the authorization and release method for the dynamic control logic of the medical robot in the second embodiment;

[0095] Figure 4 This is a flowchart illustrating the authorization and release method for the dynamic control logic of the medical robot in the third embodiment;

[0096] Figure 5 This is a flowchart illustrating the authorization and release method for the dynamic control logic of the medical robot in the fourth embodiment;

[0097] Figure 6 This is a flowchart illustrating the authorization and release method for the dynamic control logic of the medical robot in the fifth embodiment;

[0098] Figure 7 This is a flowchart illustrating the authorization and release method for the dynamic control logic of the medical robot in the sixth embodiment;

[0099] Figure 8 This is a flowchart illustrating the authorization and release method for the dynamic control logic of the medical robot in the seventh embodiment;

[0100] Figure 9 This is a structural block diagram of the authorization and release device for the dynamic control logic of the medical robot in the eighth embodiment;

[0101] Figure 10This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0102] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0103] The terms "comprising" and "having," and any variations thereof, as used in this application, are intended to cover non-exclusive inclusion. The term "multiple" as used in this application refers to two or more. The term "and / or" as used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0104] For example, Figure 1 This is a schematic diagram illustrating an application scenario of the authorized release method for the dynamic control logic of a medical robot provided in an embodiment of this application, such as... Figure 1As shown, this application scenario includes multiple medical robots 101, a publishing system 102, and an authorized person terminal 103. The multiple medical robots 101 communicate with the publishing system 102 via a wireless network, and the authorized person terminal 103 communicates with the publishing system 102. In this application scenario, the publishing system 102 first receives medical robot task objects from different sources (such as rehabilitation training plans configured on doctor workstations, dynamic gait parameters generated by AI inference services, or auxiliary decision-making logic pushed by third-party algorithm platforms) through a unified entry point. It performs a total boundary check on the control parameters (such as movement speed, joint torque, auxiliary force, execution time, etc.) in the task object. Only when all parameters are within the preset safety boundary range can the check pass, and the parameters that pass the check are locked as the task execution range, while generating a unique task identifier. Subsequently, the publishing system 102 identifies the source type of the dynamic control logic to be published (including external AI real-time generation, rule engine generation, medical task template combination, parameter adjustment, doctor modification authorization, developer resubmission or user script configuration, etc.) and the change granularity type (including parameter level, node level, topology level, behavior tree level, safety boundary level or version level changes), generates a unique identifier, version information and content hash value of the dynamic control logic, and automatically determines whether to trigger re-sandbox verification and re-authorization based on the change granularity type.If the change is at the parameter level and falls within the preset safety domain, it is exempt; otherwise, the entire process is forcibly retried. After sandbox verification in an isolated simulation environment, if the verification passes, the system acquires assessment data from six dimensions: patient harm probability, task invasiveness, failure consequences, patient status, equipment contact level, and change granularity type. A weighted risk score is then calculated, and the score is mapped to four risk levels, R1 to R4. Based on the risk level, the corresponding authorization level is matched, and an authorization application is automatically generated. The publishing system 102 pushes the authorization application to the matched authorizing terminal 103 (e.g., the attending physician's mobile personal digital assistant, the attending physician's workstation, or the nurses' station's tablet). The authorizing person views the authorization application information through the authorizing terminal 103 and returns authorization operation instructions (including confirmation, restricted confirmation, modification authorization, or rejection). If the authorization is modified, the publishing system 102 verifies whether the modified content exceeds the preset authorization boundaries (modifiable items include execution time, execution location, low-risk training duration, number of actions, voice prompt content, reminder frequency, target ward, and low-risk...). The system defines the navigation target and authorization period; unmodifiable items include safety hard limits, maximum speed, maximum torque, maximum contact force, patient safety distance, prohibited action library, R4 tasks, nodes that have not passed sandbox verification, high-risk topology, emergency stop logic, doctor takeover logic, and registration verification boundaries. If a rejection is made, the reason for rejection is recorded and a feedback loop is initiated, requiring modification and re-passing sandbox verification before an authorization request can be generated again. After authorization is approved, the publishing system 102 performs version access control review, verifying whether the authorization has expired, expired, or been revoked, and whether the authorization scope matches the current task. It generates a conclusion allowing publishing, restricting publishing (with additional constraints such as low-speed operation only, position control only, prohibiting force control output, sandbox operation only, or requiring doctor supervision), or prohibiting publishing, and deploys the control logic to the corresponding medical robot 101 based on the conclusion. During the execution of the task by the medical robot 101, the publishing system 102 continuously verifies the authorization status. Once authorization revocation, expiration, or invalidation is detected, a differentiated safety response is triggered based on the task risk level and the current execution status. If the authorization expires and the task has not started, startup is prohibited. If the authorization expires and a low-risk, uninterruptible step is in progress, the process will stop after the current step is completed. If the authorization is revoked or the R3 task fails, a safety stop or soft stop will be triggered immediately, and the process will revert to the previous safety control state. If contact actions, force-controlled actions, or physiotherapy outputs are involved, high-risk outputs will be stopped immediately, while monitoring alarms will be retained. At the same time, the publishing system 102 records key events and operation records throughout the entire process from task access, boundary checks, change identification, sandbox verification, risk scoring, authorization application, authorization confirmation, deployment, execution monitoring to task termination. The records are stored in immutable formats such as blockchain, hash chain, or audit files with digital signatures to form an evidence chain, supporting full-process audit traceability and version rollback management. This achieves full-link automated authorization control from control logic generation to robot execution.

[0105] In the first exemplary embodiment, such as Figure 2 As shown, a method for licensing and publishing dynamic control logic for medical robots is provided, which can be applied to... Figure 1 Taking the publishing system in the example, the explanation includes the following steps S201 to S208. Wherein:

[0106] Step S201: Receive the medical robot task object through a unified entry point, perform a general boundary check on the control parameters in the task object, lock the control parameters that pass the check as the execution range, and generate a task identifier.

[0107] In this embodiment, the publishing system receives medical robot task objects from different sources through a unified entry point: the AdaptiveDistributed Medical Multi-Agent Intelligence Gateway (ADMMI Gateway). These task objects include, but are not limited to: rehabilitation training plans configured on the doctor's workstation, dynamic control logic generated in real-time by AI inference services, decision instructions output by the rule engine, auxiliary decision-making logic pushed by third-party algorithm platforms, or template combination tasks manually configured by therapists through a human-computer interaction interface. The specific data formats of the task objects include Clinical Robot Task Objects (structured clinical task objects, including patient information, diagnosis, contraindications, current vital signs, task description, action type, target parameters, execution duration, allowed speed / torque range, etc.) and ROS2 Task Graphs (task graph description files based on the ROS2 framework, including nodes, topics, action servers, and data flow topology relationships).

[0108] For example, the publishing system extracts all numerical control parameters (such as motion speed, joint torque, auxiliary force, execution time, joint angle range, etc.) from the received task object and compares them one by one with the system's predefined safety boundary database. This safety boundary database pre-stores the upper and lower limits of the device safety parameters for each registered medical robot, for example: maximum forward speed ≤ 0.8 m / s, maximum joint torque ≤ 50 Nm, maximum contact force ≤ 100 N, minimum safe distance ≥ 0.3 m, etc. If all parameters are within the safety boundary range, the overall boundary check passes. The publishing system writes the parameters that pass the check into the execution range locking data structure, locking the executable parameter boundaries of the task, and assigning a unique task ID to the task. If any control parameter exceeds the safety boundary range, the publishing system directly rejects the access of the task object, returns a boundary violation report to the task initiator, and does not proceed to any subsequent processes. Through this overall boundary check mechanism, it is ensured that all dynamic control logic operates within the predefined medical safety boundaries, preventing AI-generated or unapproved control logic from bypassing safety controls and directly entering the robot execution system.

[0109] Step S202: Identify the source type and change granularity type of the dynamic control logic to be released, generate a unique identifier, version information and content hash value of the dynamic control logic, and determine whether to trigger re-sandbox verification and re-authorization based on the change granularity type.

[0110] In this embodiment, the publishing system reads the source type attribute from the metadata field of the task object. This source type is declared by the caller when the task is generated, and includes, but is not limited to, any one or more of the following: external AI real-time generation (marked as AI_GENERATED), rule engine generation (marked as RULE_ENGINE), medical task template combination (marked as TEMPLATE_COMBO), parameter adjustment (marked as PARAM_TUNE), doctor modification authorization (marked as PHYSICIAN_MODIFY), developer resubmission (marked as DEV_RESUBMIT), and user script configuration (marked as USER_SCRIPT).

[0111] Furthermore, the release system identifies the change granularity type through the control logic difference tree algorithm. Specifically, the release system performs a structured comparison between the dynamic control logic to be released (new version) and the previously authorized version stored in the system. It uses the control logic difference tree algorithm to identify the scope of the change and categorizes the change granularity type into one of the following six types: parameter-level change (only involving modification of numerical parameters, without changing the structure and topology of the control logic, such as adjusting the step speed from 0.3m / s to 0.25m / s), node-level change (involving the addition, deletion, or replacement of a single behavior node in the ROS2 Task Graph), topology-level change (involving changes in the data flow or control flow connection relationships between nodes, such as adjusting the execution order or reconstructing parallel branches), behavior tree-level change (involving modifications to high-level logic structures such as selectors, sequencers, and condition nodes in the behavior tree), safety boundary-level change (involving modifications to safety-related logic such as safety monitoring nodes, emergency stop trigger conditions, and force / torque thresholds), and version-level change (replacing the entire control logic package with a complete new version, spanning more than any of the above levels).

[0112] Optionally, the publishing system further generates a unique identifier for the dynamic control logic, Control Logic ID (generated based on source type, template ID, and timestamp), version information Control Logic Version (semantic version number, such as V3.2.1), and content hash value Control Logic Hash (SHA-256 hash calculation performed on the entire control logic package).

[0113] For example, the release system determines whether to trigger re-sandbox verification and re-authorization based on the change granularity type. The specific judgment logic is as follows: If the change granularity type is a parameter-level change and the parameter-level change is within a preset security domain (e.g., the step speed adjustment is within the template's preset low-risk adjustable range of 0.2m / s to 0.4m / s), then re-sandbox verification and re-authorization are exempted, and the historical sandbox verification report is directly reused; if the change granularity type is a parameter-level change but the parameter-level change is not within a preset security domain, then re-sandbox verification and re-authorization are triggered; if the change granularity type is any of the following: node-level change, topology-level change, behavior tree-level change, security boundary-level change, or version-level change, then re-sandbox verification and re-authorization are forcibly triggered. Through the above source identification and change granularity identification mechanisms, the system can accurately determine which changes require re-verification and re-authorization, realizing the automatic binding of version updates and re-authorization trigger conditions.

[0114] Step S203: Perform sandbox verification on the dynamic control logic in an isolated simulation environment to obtain the verification results.

[0115] In this embodiment, the deployment system performs sandbox verification of the dynamic control logic in an isolated simulation environment. This isolated simulation environment constructs a digital twin simulation scene based on actual hospital physical environment data (such as ward dimensions, equipment layout, and patient models), and runs a ROS2 Task Graph using a robot simulation platform (such as Gazebo or Coppelia Sim). The deployment system deploys the dynamic control logic to be verified to a virtual robot in the simulation environment, inputting the patient's current state parameters (such as weight, joint range of motion, muscle strength level, vital signs, etc.) as initial simulation conditions, driving the virtual robot to execute the complete task flow. During simulation execution, the deployment system monitors various safety indicators in real time, including but not limited to: joint angles, speed, torque, end effector position, contact force, and trajectory deviation.

[0116] Optionally, the publishing system generates verification results based on the monitoring indicators throughout the simulation process, with the following judgment rules: If all monitoring indicators do not exceed the safety threshold throughout the task and the task results meet expectations (such as reaching the target joint angle, completing the predetermined number of steps, etc.), the verification result is passed; if the monitoring indicators do not exceed the hard safety threshold but are close to the threshold boundary (such as the peak torque reaching more than 85% of the threshold), or there are non-fatal abnormalities that can be tolerated under limited conditions (such as slight trajectory deviation but still within the patient's acceptable range), the verification result is a warning; if any monitoring indicator exceeds the hard safety threshold, or if a safety accident such as collision, loss of control, or emergency stop triggering occurs during the simulation, the verification result is failed.

[0117] Step S204: If the verification result is passed, obtain assessment data from multiple dimensions, calculate the risk score based on the weighted assessment data from multiple dimensions, map the risk score to the corresponding risk level, match the corresponding authorization level according to the risk level, and generate an authorization application.

[0118] In this embodiment, if the verification result is passed, the publishing system obtains evaluation data from multiple dimensions, including: patient harm probability, task invasiveness, failure consequences, patient status, device contact level, and change granularity type identified in step S202.

[0119] For example, each dimension is scored according to a preset scoring standard (0-100 points), and the specific scoring standard is as follows: The patient injury probability dimension is assessed based on the probability of injury occurring, with 0-20 points being almost impossible, 21-50 points being low probability, 51-80 points being moderate probability, and 81-100 points being high probability; The task invasiveness dimension is assessed based on the degree of physical invasiveness, with 0-20 points being non-contact / non-invasive, 21-50 points being skin contact / mild, 51-80 points being skin penetration / cavity penetration, and 81-100 points being invasive into body cavities / surgery; The failure consequences dimension is assessed based on the maximum harm caused by failure, with 0-20 points being mild discomfort, 21-50 points being temporary injury, 51-80 points being permanent mild to moderate injury, and 81-100 points being life-threatening. Life / Severe Disability; Patient Status dimension is assessed based on the patient's current physiological vulnerability: 0-20 points for healthy / stable, 21-50 points for mildly unstable, 51-80 points for moderately unstable, and 81-100 points for severely unstable / high-risk; Equipment Contact Degree dimension is assessed based on the interaction method between the robot and the patient: 0-20 points for no direct contact, 21-50 points for indirect contact (wearable), 51-80 points for direct contact but not force-controlled, and 81-100 points for force-controlled / invasive contact; Change Granularity Type dimension is assessed based on the scope of change's impact: 0-20 points for parameter level (within the security domain), 21-50 points for parameter level (superdomain) / node level, 51-80 points for topology level / behavior tree level, and 81-100 points for security boundary level / version level.

[0120] It should be understood that the publishing system in this embodiment calculates the Risk Score by weighting and summing the scores of each dimension according to preset weights. The weights for each dimension can be configured according to the hospital's actual policies. One possible default weight configuration is: patient harm probability weight 0.25, task invasiveness weight 0.20, failure consequences weight 0.20, patient status weight 0.15, equipment contact level weight 0.10, and change granularity type weight 0.10.

[0121] Optionally, the system maps the calculated risk score to the corresponding risk level. The mapping rules are as follows: 0-20 points are mapped to R1 (very low risk), 21-50 points to R2 (low-medium risk), 51-80 points to R3 (medium-high risk), and 81-100 points to R4 (very high risk). R1 ​​to R4 are four levels arranged from low risk to high risk.

[0122] The publishing system matches the corresponding authorization level based on the risk level: R1 level matches automatic system authorization or post-event random checks; R2 level matches authorization confirmed by the attending physician, rehabilitation therapist, or nurse station; R3 level matches pre-approval authorization by the attending physician or senior physician; R4 level matches a default prohibition on publishing. The publishing system automatically assembles the authorization application based on the matched authorization level. This authorization application includes: applicant information, patient information, task description, risk level, required authorization level, sandbox validation report identifier, control logic hash value, etc., and generates the authorization application.

[0123] It should be understood that through the above six-factor weighted risk scoring model, the system quantifies the control logic into four risk levels from R1 to R4 and automatically matches the corresponding authorization level to achieve hierarchical authorization. This allows low-risk tasks to be executed automatically, while medium- and high-risk tasks are approved by the responsible physician, attending physician, or regulatory body, respectively, ensuring that the authorization level is commensurate with the clinical risk.

[0124] Step S205: Push the authorization application to the matching authorized person's terminal, receive the authorization operation instruction returned by the authorized person, and execute the corresponding operation according to the authorization operation instruction.

[0125] In this embodiment, the publishing system sends the authorization request to the matched authorizing terminal via the hospital information system or message push service. The authorizing terminal displays relevant information about the authorization request to the authorizing party through an authorization interface, including: basic patient information, task description, control logic parameters, risk level, required authorization level, and sandbox validation report summary.

[0126] For example, when an authorizing user performs an authorization operation on the authorization interface, the returned authorization operation instructions include one or more of the following: confirmation instruction, restriction confirmation instruction, modification authorization instruction, and rejection instruction. The authorization interface clearly distinguishes between modifiable and non-modifiable items for the authorizing user: modifiable items include one or more of the following: execution time, execution location, low-risk training duration, number of actions, voice prompt content, reminder frequency, target ward, low-risk navigation target, and authorization period; non-modifiable items include one or more of the following: safety hard limits, maximum speed, maximum torque, maximum contact force, patient safety distance, prohibited action library, R4 risk level tasks, control logic nodes that have not passed sandbox verification, high-risk topology, emergency stop logic, doctor takeover logic, and registration verification boundaries.

[0127] Optionally, the publishing system receives authorization operation instructions returned by the authorizing person and performs corresponding operations according to the type of authorization operation instruction: if a confirmation instruction is received, version access control review is performed directly; if a restriction confirmation instruction is received, the additional constraints on the authorizing person carried in the restriction confirmation instruction (such as "only to be executed from 9:00 to 11:00 today", "must be accompanied by a therapist", etc.) are recorded, and version access control review is performed; if a modification authorization instruction is received, the modification content is verified to see if it exceeds the preset authorization boundary. If the modifications involve unmodifiable items, the modifications will be rejected and the authorized user's terminal will be notified of the specific violations, requiring the control logic to be modified or regenerated. If the modifications only involve modifiable items and are all within the permitted range, the modifications will be recorded and version access control will be implemented. Upon receiving a rejection instruction, the reason for rejection will be recorded and a feedback loop will be initiated. The reason for rejection (including specific parameter violations, risk level mismatch, lack of necessary attachments, etc.) will be written into the evidence chain and fed back to the developer, doctor, or task generator through the message channel. Modification of the control logic based on the reason for rejection (such as reducing the risk level, deleting high-risk nodes, adjusting parameters, supplementing explanations, or reselecting a template) is allowed, but the modified control logic must be re-verified through the sandbox. Only after re-verification through the sandbox is the authorization application allowed to be generated again.

[0128] It should be understood that through the aforementioned authorization boundary management mechanism, authorized personnel can only adjust modifiable items within the preset security domain, and cannot violate security hard limits, prohibited action libraries, high-risk topologies, or other security bottom lines, thus preventing security incidents caused by unauthorized modifications. Through the authorization rejection feedback closed-loop mechanism, the reasons for rejection are recorded and reported. After modification, a new sandbox verification is required before an authorization request can be generated again, effectively solving the problem of repeated invalid submissions caused by the traditional approval process's "rejection equals termination" approach.

[0129] Step S206: Perform version access control audit on the authorized dynamic control logic, verify the validity of the authorization, and generate a release conclusion that allows release, restricts release, or prohibits release based on the audit result.

[0130] In this embodiment, the publishing system performs version access control audits on the dynamic control logic that has been authorized. The publishing system first verifies the validity of the authorization, including: verifying whether the current authorization has timed out (during the authorization waiting period, the publishing system monitors the response time limit of the authorizing person; if it exceeds the preset response time limit, it is considered to have timed out), verifying whether the current authorization has expired (comparing the current time with the authorization validity period; if the current time exceeds the authorization validity period, it is considered to have expired), verifying whether the current authorization has been revoked (querying the authorization status in the hospital information system; if the authorizing person actively revoks or the system detects a change in the patient's status that triggers authorization invalidation, it is considered to have been revoked), and verifying whether the scope of the current authorization matches the current task (comparing whether the patient information, device number, execution location specified in the authorization are consistent with the actual parameters of the current task).

[0131] For example, while verifying the validity of authorization, the publishing system performs tiered authorization timeout handling: During the authorization waiting period, the publishing system monitors the response time of the authorizing person. If an authorization timeout occurs, the following differentiated handling is performed: If a task of risk level R1 fails to respond after timeout, it is transferred to post-event spot checks or automatically downgraded; if a task of risk level R2 fails to respond after timeout, it enters the manual review queue or is restricted from publishing; if a task of risk level R3 fails to respond after timeout, publishing is prohibited and higher-privilege roles are notified; among these, R4 risk level tasks cannot be automatically published due to timeout. Through the above tiered authorization timeout handling mechanism, low-risk tasks can be automatically downgraded, while medium- and high-risk tasks enter the manual queue or are prohibited from publishing, balancing the rigor of daily management with process efficiency.

[0132] For example, in scenarios involving emergency rescue or preventing further harm from equipment, the publishing system also supports temporary authorization (emergency overriding) initiated by higher-privilege roles. This temporary authorization only allows the execution of safety actions such as stopping, evacuation, low-speed navigation, safe rollback, de-hazard removal, and notifying personnel. It prohibits the addition of new diagnoses, prescriptions, treatment parameter adjustments, intrusion operations, or unverified control logic. Through this emergency overriding mechanism, higher-privilege roles in emergency rescue scenarios can execute temporary authorizations limited to safety actions, while an action whitelist constraint ensures that "emergency authorization is not equivalent to unlimited authorization."

[0133] For example, after the authorization validity verification passes, the publishing system binds the authorization record to form an authorization chain and performs version access control audit. It compares the completeness and consistency of the control logic version, the sandbox verification report version, and the authorized version, and generates a publishing conclusion that allows, restricts, or prohibits publishing based on the audit results: if all conditions are met, a publishing conclusion that allows publishing is generated; if there are warning conditions or the authorization is subject to additional constraints, a publishing conclusion that restricts publishing is generated; if any condition is not met, a publishing conclusion that prohibits publishing is generated.

[0134] Step S207: Execute the corresponding release operation based on the release conclusion, and continuously verify the authorization status during the execution of the dynamic control logic. When the authorization fails, trigger the corresponding security response action based on the task risk level and the current execution status.

[0135] In this embodiment, the publishing system performs the corresponding publishing operation based on the publishing conclusion: if publishing is allowed, the dynamic control logic is deployed to the target medical robot system; if publishing is restricted, the system is deployed to the target medical robot system after applying restrictive constraints. Restricted publishing includes applying one or more of the following constraints: low-speed operation only (speed limit reduced to 50% of the safe value), position control only (force control mode disabled), force control output prohibited (force control gain set to zero), sandbox operation only (not deployed to the physical robot), and a doctor must be present (authorized person must be confirmed to be present through facial recognition or employee badge verification); if publishing is prohibited, deployment locking is performed to prevent the control logic from being deployed to the physical robot.

[0136] For example, during the execution of dynamic control logic, the publishing system continuously verifies the authorization status at a fixed frequency (e.g., 1Hz). Verification items include: whether the authorization has been revoked (real-time query of HIS authorization status), whether the authorization has expired (comparing the current time with the authorization validity period), and whether the task-patient-device match is still valid. Once authorization failure is detected, the publishing system triggers corresponding safety response actions based on the task risk level and the current execution status. Specifically: when authorization has expired and the task has not yet started, new task startup is prohibited; when authorization has expired and a low-risk, uninterruptible step is being executed, subsequent execution is stopped after the current step is completed; when authorization is revoked or R3 risk level task authorization fails, a safety stop or soft stop is triggered, and the system reverts to the previous safety control state or switches to doctor takeover; when contact actions, force-controlled actions, or physiotherapy outputs are involved and authorization has failed, high-risk outputs are stopped (force control reset to zero, movement stopped), monitoring data is retained, and alarm prompts are generated and pushed to the nurse station and doctor's terminal.

[0137] It should be understood that through the above-mentioned mechanism for continuously verifying the authorization status during execution, the system will perform differentiated safety responses based on the risk level and task status when the authorization fails. These responses include prohibiting new tasks, stopping after completing the current step, immediately stopping safely, or reverting to the previous safety control state, ensuring that the robot is always within a controllable range.

[0138] Step S208: Record key events and operation records of the entire process of dynamic control logic from task access to execution completion, and store them in an immutable format to form an evidence chain. The evidence chain is used to support audit traceability and version rollback management.

[0139] In this embodiment, the publishing system records key events and operation records of the entire process of dynamic control logic from task access to execution completion. The recording scope includes, but is not limited to: task version, model version, rule version, control logic version (including hash value), sandbox verification report version, authorization version (including digital signature of the authorizer), publishing version, modification records (including timestamp, operator, and modified content), execution status (started, running, paused, stopped, abnormal termination), etc.

[0140] Optionally, the publishing system preserves evidence in an immutable format to form an evidence chain, supporting, but not limited to, the following evidence formats: blockchain evidence preservation (encapsulating the evidence chain data into transactions and submitting them to the consortium blockchain, using the blockchain's consensus mechanism and block hash links to ensure immutability), encrypted logs (using AES-256 encryption to store log files, with the key held by multiple people), hash chain logs (each record block contains the hash value of the previous block, forming a chain structure; modification of any record will destroy the entire hash chain), and audit files with digital signatures (using the authorized person's private key to digitally sign the audit digest and export it as a PDF or XML file).

[0141] Optionally, the evidence chain supports audit traceability, such as: supporting the retrieval of historical authorization records by multiple dimensions such as Task ID, Patient ID, Authorizer ID, time range, and risk level, and fully replaying the entire event sequence from creation to completion of any task. The evidence chain also supports version rollback management: when a security vulnerability or authorization flaw is found in the published control logic, it supports rollback to any historical authorized version, and the rollback operation itself is recorded as a new evidence chain entry.

[0142] It should be understood that the above-mentioned evidence chain management mechanism enables the entire process from control logic generation to execution to be auditable and traceable, improving the security and regulatory transparency of the release of medical robot control logic, and fully meeting the compliance requirements of medical device regulatory agencies for post-market change control and clinical deployment authorization of AI medical software.

[0143] In the aforementioned method for authorizing and releasing the dynamic control logic of medical robots, a unified entry point is used to receive medical robot task objects and perform a general boundary check. All dynamic control logic is forcibly incorporated into gateway management, effectively preventing AI-generated or unapproved control logic from bypassing doctors and directly entering the robot's execution system. By identifying the source type and change granularity type of the control logic, and determining whether to trigger re-sandbox verification and re-authorization based on the change granularity type, differentiated management of six change granularities is achieved. Parameter-level changes within the security domain are exempt from re-verification and re-authorization, thus improving efficiency; major changes are forcibly triggered to re-run the entire process, thus eliminating safety hazards. Sandbox verification is performed in an isolated simulation environment. Only after successful verification does the risk assessment begin, achieving a mandatory binding between technical verification and the authorization process. Six-dimensional assessment data are obtained for weighted risk scoring, mapped to four risk levels from R1 to R4, and matched with corresponding authorization levels. It achieves precise matching between authorization levels and clinical risks; it pushes authorization requests to the authorizing user's terminal and verifies whether modifications exceed preset boundaries between modifiable and non-modifiable items when authorization is modified. When authorization is rejected, the reason for rejection is written into the evidence chain, and modification is required before authorization requests can be generated again after sandbox verification, forming an unbypassable security loop; it performs version access control audits on the control logic that has passed authorization, generates allowed, restricted, or prohibited release conclusions, and implements differentiated release control; it continuously verifies the authorization status during execution, and triggers differentiated security responses based on the task risk level and current execution status when authorization expires, ensuring that the robot remains within a controllable range in any authorization failure scenario; it records key events and operation records from task access to execution completion, storing them in an immutable format to form an evidence chain, supporting audit traceability and version rollback management, and achieving full auditability and traceability from control logic generation to deployment execution.

[0144] In a second exemplary embodiment, such as Figure 3 As shown, a method for licensing and publishing dynamic control logic for medical robots is provided, which can be applied to... Figure 1 Taking the publishing system in the example, the explanation includes the following steps S301 to S304. Wherein:

[0145] Step S301: Receive the medical robot task object through a unified entry point, perform a total boundary check on the control parameters in the task object, lock the control parameters that pass the check as the execution range, and generate a task identifier.

[0146] Step S302: Identify the source type and change granularity type of the dynamic control logic to be released, generate a unique identifier, version information and content hash value of the dynamic control logic, and determine whether to trigger re-sandbox verification and re-authorization based on the change granularity type.

[0147] Step S303: Perform sandbox verification on the dynamic control logic in an isolated simulation environment to obtain the verification results.

[0148] For the specific implementation process and technical effects of steps S301 to S303 in this embodiment, please refer to [link to documentation]. Figure 2 The relevant descriptions of steps S201 to S203 in the method embodiment shown will not be repeated here.

[0149] Step S304: If the verification result fails, prevent the generation of authorization requests and terminate the process.

[0150] In this embodiment, when the verification result obtained in step S303 is a failure, the release system directly prohibits the generation of any authorization request and terminates the entire authorization release process. Verification failure means that in the isolated simulation environment, the dynamic control logic has experienced at least one safety indicator exceeding the hard safety threshold during task execution (such as joint torque exceeding the limit, end position exceeding the safety boundary, virtual collision, etc.), or triggered an emergency stop condition, indicating that the control logic has a real safety risk in the current clinical environment and patient status.

[0151] Optionally, while prohibiting the generation of authorization requests, the deployment system can also generate a verification failure report to record the reason for the failure (including detailed data such as the name of the exceeded indicator, the exceeded value, the time of occurrence, and the triggering conditions), and feed this verification failure report back to the task initiator (developer, doctor, or task generator). The feedback includes a clear explanation of the failure reason and modification suggestions, so that the task initiator can modify the control logic, reduce the risk level, delete high-risk nodes, adjust parameters, or reselect the template based on the failure reason. Dynamic control logic that fails verification is not allowed to enter the risk assessment, authorization application, and deployment stages. Only after the modified control logic is resubmitted and passes sandbox verification can it enter the subsequent authorization process.

[0152] Through the above-mentioned hard-blocking mechanism for verification failure, the control logic that fails verification is directly rejected from generating an authorization request and cannot enter the authorization process, thus eliminating invalid authorization applications from the source and achieving a forced binding between sandbox verification and the authorization entry point.

[0153] In a third exemplary embodiment, such as Figure 4 As shown, a method for licensing and publishing dynamic control logic for medical robots is provided, which can be applied to... Figure 1 Taking the publishing system in the example, the explanation includes the following steps S401 to S404. Wherein:

[0154] Step S401: Receive the medical robot task object through a unified entry point, perform a general boundary check on the control parameters in the task object, lock the control parameters that pass the check as the execution range, and generate a task identifier.

[0155] Step S402: Identify the source type and change granularity type of the dynamic control logic to be released, generate a unique identifier, version information and content hash value of the dynamic control logic, and determine whether to trigger re-sandbox verification and re-authorization based on the change granularity type.

[0156] Step S403: Perform sandbox verification on the dynamic control logic in an isolated simulation environment to obtain the verification results.

[0157] For the specific implementation process and technical effects of steps S301 to S303 in this embodiment, please refer to [link to documentation]. Figure 2 The relevant descriptions of steps S201 to S203 in the method embodiment shown will not be repeated here.

[0158] Step S404: If the verification result is a warning, generate a restricted authorization request with limiting constraints. The limiting constraints are used to be attached to the dynamic control logic to be released in subsequent release steps.

[0159] In this embodiment, when the verification result obtained in step S403 is a warning, the publishing system generates a restricted authorization request with additional constraints. A warning verification result means that, in the isolated simulation environment, although the dynamic control logic does not exceed the hard safety threshold during task execution, one or more of the following situations exist: monitoring indicators are close to the threshold boundary (e.g., peak torque reaches more than 85% of the threshold), there are non-fatal anomalies that can be tolerated under constraints (e.g., slight trajectory deviation but still within the patient's acceptable range), there are unverified low-risk alternative paths in the control logic, or the patient's state is slightly unstable. These situations indicate that the control logic itself does not have hard safety defects, but there are potential risks under specific conditions, which need to be controlled by adding constraints.

[0160] For example, the deployment system automatically generates corresponding constraints based on the type and severity of warnings identified during the simulation. For instance: if the warning is due to a speed index approaching the upper threshold, a "low-speed operation only" constraint is added (reducing the speed limit to 50% of the safe value); if the warning is due to large fluctuations in contact force in force control mode, "force control output prohibited" and "position control only" constraints are added; if the warning is due to the existence of unverified alternative paths in the task, a "sandbox operation only" constraint is added (not deployed to the physical robot); if the warning is due to a slightly unstable patient condition, a "doctor must be present" constraint is added (execution can only proceed after authorized personnel are verified through facial recognition). These constraints can be added individually or in combination based on multiple warning types.

[0161] Optionally, after generating a restricted authorization request with attached constraints, the authorization request enters the subsequent authorization process (including risk assessment, authorization decision, version access control review, release execution, and evidence chain storage in steps S204 to S208). During the authorization decision stage, the authorizing user's terminal will prominently display sandbox verification warning information and corresponding constraints. After fully understanding the warning information, the authorizing user can make a decision to confirm authorization (with attached constraints), restrict confirmation (adding additional constraints), modify authorization, or reject the authorization. During the release execution stage, the constraints in the restricted authorization request are attached to the dynamic control logic to be released and are forcibly applied when deployed to the target medical robot system.

[0162] Through the aforementioned verification warning branch mechanism, the system does not reject control logic with potential risks but non-fatal defects outright. Instead, it allows the logic to be released and executed under controlled conditions by adding restrictive constraints. This ensures security while avoiding the problem of reduced availability due to excessive conservatism, thus achieving a balance between security and utility.

[0163] In the fourth exemplary embodiment, such as Figure 5 As shown, a method for licensing and publishing dynamic control logic for medical robots is provided, which can be applied to... Figure 1 Taking the publishing system in the example, the explanation includes the following steps S501 to S506. Wherein:

[0164] Step S501: Receive the medical robot task object through a unified entry point, perform a general boundary check on the control parameters in the task object, lock the control parameters that pass the check as the execution range, and generate a task identifier.

[0165] Step S502: Identify the source type and change granularity type of the dynamic control logic to be released, generate a unique identifier, version information and content hash value of the dynamic control logic, and determine whether to trigger re-sandbox verification and re-authorization based on the change granularity type.

[0166] Step S503: Perform sandbox verification on the dynamic control logic in an isolated simulation environment to obtain the verification results.

[0167] Step S504: If the verification result is passed, obtain assessment data from multiple dimensions, calculate the risk score based on the weighted assessment data from multiple dimensions, map the risk score to the corresponding risk level, match the corresponding authorization level according to the risk level, and generate an authorization application.

[0168] For the specific implementation process and technical effects of steps S501 to S504 in this embodiment, please refer to [link to documentation]. Figure 2 The relevant descriptions of steps S201 to S204 in the method embodiment shown will not be repeated here.

[0169] Step S505: Push the authorization request to the matching authorized person's terminal and monitor the authorized person's response time during the authorization waiting period.

[0170] In this embodiment, after the publishing system sends the authorization request to the matched authorized person's terminal through the hospital information system or message push service, it starts a response timeout timer. Different risk levels of tasks correspond to different preset response timeouts. For example, R1 risk level tasks can be set with a longer response timeout (e.g., 60 minutes) or no timeout; R2 risk level tasks are set with a medium response timeout (e.g., 30 minutes); R3 risk level tasks are set with a shorter response timeout (e.g., 15 minutes); and R4 risk level tasks do not have automatic timeout processing (always waiting for explicit authorization). The preset response timeouts can be flexibly configured according to the hospital's actual management policies.

[0171] For example, during the authorization waiting period, the publishing system monitors the authorized person's response time limit at a fixed frequency (e.g., once per minute) to check whether the current waiting time has exceeded the preset response time limit. At the same time, the publishing system can send reminder notifications to the authorized person's terminal during the authorization waiting period (e.g., send reminders once after 5 minutes and once after 10 minutes) to prevent the task from being suspended for a long time due to forgetfulness or being busy.

[0172] In step S506, if an authorization timeout occurs, the corresponding differentiated processing is performed according to the risk level.

[0173] In this embodiment, when the authorizing person's response time exceeds the preset response time limit, the publishing system determines it as an authorization timeout and performs corresponding differentiated processing based on the risk level:

[0174] When an R1 risk level task fails to respond within the timeout period, the system will either initiate a post-event spot check or automatically downgrade the execution. R1 risk level tasks are considered extremely low-risk (such as simple passive joint movement training, low-intensity repetitive movements, etc.). Provided the sandbox validation is successful, the system will automatically grant temporary execution permission and begin execution, but the task will be marked as a "post-event spot check" task. The authorized person will then conduct a retrospective audit of the execution record. If the post-event audit finds that the task does not meet the authorization conditions, it will be recorded as an abnormal event, and a corrective process will be initiated.

[0175] When an R2 risk level task fails to respond within the timeout period, the system will either place the request in a manual review queue or restrict its release. For R2 risk level tasks (low to medium risk, such as lower limb gait training and upper limb rehabilitation training), the system will transfer the authorization request to a manual review queue for secondary allocation by the head nurse or on-call therapist. Alternatively, the system can restrict release by adding constraints such as "sandbox only" or "physician must be present" before granting execution, ensuring that the task can still proceed under controlled conditions even if the authorizer fails to respond promptly.

[0176] If a task at risk level R3 fails to respond within the timeout period, the system will prohibit its publication and notify higher-authority roles. For tasks at risk level R3 that are considered medium to high risk (such as invasive rehabilitation procedures or force-controlled assisted movements), the system will automatically refuse to publish them and send a notification to higher-authority roles (such as department heads or on-duty staff in the medical affairs office) for intervention or reassignment of authorized personnel.

[0177] Among them, R4 risk level tasks are extremely high risk tasks. The system is set not to be automatically released due to timeout. No matter how long the wait is, explicit authorization must be obtained before execution. There is no automatic downgrade or automatic approval mechanism.

[0178] Through the above-mentioned authorization timeout tiered processing mechanism, low-risk tasks can be automatically downgraded to improve process efficiency, while medium- and high-risk tasks are placed in a manual queue or prohibited from being published to prevent security risks. This balances the rigor of daily management with the flexibility of emergency scenarios, effectively solving the problem of tasks being suspended for a long time due to the authorizer being busy or forgetting.

[0179] In the fifth exemplary embodiment, such as Figure 6 As shown, a method for licensing and publishing dynamic control logic for medical robots is provided, which can be applied to... Figure 1 Taking the publishing system in the example, the explanation includes the following steps S601 to S606. Wherein:

[0180] Step S601: Receive the medical robot task object through a unified entry point, perform a general boundary check on the control parameters in the task object, lock the control parameters that pass the check as the execution range, and generate a task identifier.

[0181] Step S602: Identify the source type and change granularity type of the dynamic control logic to be released, generate a unique identifier, version information and content hash value of the dynamic control logic, and determine whether to trigger re-sandbox verification and re-authorization based on the change granularity type.

[0182] Step S603: Perform sandbox verification on the dynamic control logic in an isolated simulation environment to obtain the verification results.

[0183] Step S604: If the verification result is passed, obtain assessment data from multiple dimensions, calculate the risk score based on the weighted assessment data from multiple dimensions, map the risk score to the corresponding risk level, match the corresponding authorization level according to the risk level, and generate an authorization application.

[0184] For the specific implementation process and technical effects of steps S601 to S604 in this embodiment, please refer to [link to documentation]. Figure 2 The relevant descriptions of steps S201 to S204 in the method embodiment shown will not be repeated here.

[0185] Step S605: In a scenario involving emergency rescue or preventing further damage from the device, an emergency coverage request initiated by a higher-privilege role is detected.

[0186] In this embodiment, the publishing system continuously listens for emergency coverage requests sent from the authorized user's terminal. Typical scenarios that trigger emergency coverage include, but are not limited to: emergency situations such as a patient suddenly experiencing epilepsy, respiratory arrest, or severe arrhythmia; equipment malfunctions during robot execution that may cause harm to the patient, such as abnormal vibration, overheating, or unusual noises; patients experiencing adverse reactions (such as pain, spasms, or a sudden increase in blood pressure) requiring immediate termination of treatment; and other emergency situations where the authorized user determines that the risks of continuing the task outweigh the expected benefits.

[0187] For example, the initiator of an emergency coverage request is a higher-authority role, including but not limited to department heads, on-call chief physicians, and emergency department consulting physicians, who possess higher clinical decision-making authority. When the publishing system receives an emergency coverage request, it first verifies the initiator's identity and permissions (through multi-factor authentication methods such as digital certificates, employee badge authentication, and biometrics). Only after confirming that the initiator has authorization levels higher than those required for the current task can the emergency coverage request be accepted.

[0188] Step S606: A temporary authorization is initiated by a higher-privilege role. The temporary authorization only allows the execution of safety actions such as stopping, evacuating, low-speed navigation, safe rollback, clearing dangerous situations, and notifying personnel. It prohibits the addition of new diagnoses, prescriptions, treatment parameter adjustments, intrusion operations, or unverified control logic.

[0189] In this embodiment, after receiving an emergency coverage request, the publishing system initiates temporary authorization by a higher-privilege role. The core feature of this temporary authorization is the action whitelist constraint, which only allows the following predefined safety actions to be executed: stop (immediately stop all robot movements), evacuation (move the robot's end effector away from the patient to a safe position), low-speed navigation (move the robot to the designated position at a speed 50% below the safety limit), safe retreat (slowly retreat the robot's joints or end effector to the initial safe position), de-danger (release the gripping force, unlock the state, etc.), and notify personnel (send alarm notifications to the nurse station, doctor's terminal, and duty room).

[0190] The temporary authorization explicitly prohibits the following actions: adding new diagnoses (new diagnostic procedures cannot be initiated using emergency authorization), prescribing (medication prescriptions cannot be issued or modified), adjusting treatment parameters (treatment-related parameters such as force control gain, speed limit, and torque limit cannot be adjusted), invasive procedures (any puncture, cutting, or insertion procedures cannot be performed), and unverified control logic (any dynamic control logic that has not passed sandbox validation cannot be executed). This whitelist of actions ensures that emergency authorization does not introduce new security risks, and emergency coverage is used solely for risk control, not for executing new tasks.

[0191] Optionally, after the aforementioned security actions are performed, the emergency coverage temporary authorization is marked as "emergency terminated." The publishing system records the complete emergency coverage process (including the initiator's identity, trigger time, security actions performed, device status at termination, patient vital signs, etc.), and writes it into the chain of evidence to form an immutable audit log for post-event review and root cause analysis.

[0192] Through the aforementioned emergency coverage mechanism, higher-privilege roles in emergency situations can perform temporary authorizations limited to safety actions. This ensures the timeliness of emergency treatment while preventing the security vulnerability of "emergency authorization equals unlimited authorization" through action whitelist constraints, thus balancing the flexibility of emergency response with the rigor of patient safety protection.

[0193] In the sixth exemplary embodiment, such as Figure 7 As shown, a method for licensing and publishing dynamic control logic for medical robots is provided, which can be applied to... Figure 1 Taking the publishing system in the example, the explanation includes the following steps S701 to S708. Wherein:

[0194] Step S701: Receive the medical robot task object through a unified entry point, perform a total boundary check on the control parameters in the task object, lock the control parameters that pass the check as the execution range, and generate a task identifier.

[0195] Step S702: Identify the source type and change granularity type of the dynamic control logic to be released, generate a unique identifier, version information and content hash value of the dynamic control logic, and determine whether to trigger re-sandbox verification and re-authorization based on the change granularity type.

[0196] Step S703: Perform sandbox verification on the dynamic control logic in an isolated simulation environment to obtain the verification results.

[0197] Step S704: If the verification result is passed, obtain assessment data from multiple dimensions, calculate the risk score based on the weighted assessment data from multiple dimensions, map the risk score to the corresponding risk level, match the corresponding authorization level according to the risk level, and generate an authorization application.

[0198] Step S705: Push the authorization application to the matching authorized person's terminal, receive the authorization operation instruction returned by the authorized person, and perform the corresponding operation according to the authorization operation instruction.

[0199] Step S706: Perform version access control audit on the authorized dynamic control logic, verify the validity of the authorization, and generate a release conclusion that allows release, restricts release, or prohibits release based on the audit result.

[0200] Step S707: Execute the corresponding release operation based on the release conclusion, deploy the dynamic control logic to the target medical robot system and start execution.

[0201] For the specific implementation process and technical effects of steps S701 to S707 in this embodiment, please refer to [link to documentation]. Figure 2 The relevant descriptions of steps S201 to S207 in the method embodiment shown will not be repeated here.

[0202] Step S708: During execution, continuously verify the authorization status. When authorization failure is detected, trigger the corresponding security response action based on the task risk level and the current execution status.

[0203] In this embodiment, while the dynamic control logic is deployed and executing on the medical robot, the publishing system continuously verifies the authorization status at a fixed frequency (e.g., 1Hz). Verification includes, but is not limited to: whether the authorization has been revoked (real-time query of the authorization status in the hospital information system; if the authorizer actively revokes it, a change in patient status triggers the authorization expiration condition, or a higher-level authority forcibly terminates the authorization, it is considered revoked); whether the authorization has expired (the system compares the current timestamp with the authorization expiration time in the authorization record; if the current time exceeds the expiration time, it is considered expired); and whether the authorization scope still matches the current task (comparing the currently executing patient, device, and location with the scope specified in the authorization record).

[0204] For example, once the publishing system detects an authorization failure, it immediately triggers a corresponding security response action based on the task's risk level and current execution status. Specific scenarios may include:

[0205] Scenario 1: When the authorization has expired and the task has not yet started, the publishing system prohibits the initiation of new tasks. The task remains in the pending state, and the system sends a notification message to the authorizer and the task initiator, explaining that the authorization has expired and that re-authorization is required before execution can begin.

[0206] Scenario 2: When authorization expires and a low-risk, uninterruptible step is in progress, the publishing system stops subsequent execution after completing the current step. A low-risk, uninterruptible step refers to a continuous sequence of movements that cannot be safely interrupted after initiation (such as a joint flexion movement that has completed more than 80% of its range, or the current gait cycle in an ongoing multi-gait cycle). Forced interruption may cause greater mechanical shock or patient discomfort. The system allows the current step to complete naturally, then performs a soft stop at the step boundary, preventing the initiation of subsequent steps, and simultaneously sends an authorization expiration notification to the authorizing user.

[0207] Scenario 3: When authorization is revoked or R3 risk level task authorization expires, the publishing system triggers a safe stop or soft stop, then reverts to the previous safe control state or switches to doctor takeover. A safe stop is a rapid but controlled deceleration stop, while a soft stop is a slow deceleration stop to reduce impact on the patient. Reverting to the previous safe control state means restoring the robot's joint angles, end effector positions, and other states to the safe baseline state before the task started. Doctor takeover means switching the robot to manual control mode or zero-force mode, allowing the doctor to directly control the robot to move it to a safe position.

[0208] Scenario 4: When contact-based movements, force-controlled movements, or physiotherapy outputs are involved, and authorization has expired, the issuing system immediately stops high-risk outputs (force control reset to zero, movement stopped, energy output cut off), while retaining monitoring data (heart rate, blood pressure, electromyography signals, joint angles, contact force, and other sensor data), and generating alarm notifications to be pushed to the nurse station and doctor's terminals. After stopping high-risk outputs, the system remains in monitoring mode, awaiting manual intervention from doctors or technicians.

[0209] Through the aforementioned differentiated response mechanism for authorization failure during execution, the system performs differentiated safety responses based on the task risk level and the current execution status when authorization is revoked, expired, or invalidated. When authorization expires and the task has not started, the system prohibits the initiation of new tasks. When authorization expires and a low-risk, uninterrupted step is in progress, the system stops after completing the current step. When authorization is revoked or R3 task authorization fails, the system immediately stops safely or reverts to the previous safety control state. When contact actions, force-controlled actions, or physiotherapy outputs are involved, high-risk outputs are immediately stopped while monitoring alarms are retained, ensuring that the robot remains within a controllable range in any authorization failure scenario.

[0210] In the seventh exemplary embodiment, such as Figure 8 As shown, a method for licensing and publishing dynamic control logic for medical robots is provided, which can be applied to... Figure 1 Taking the publishing system in the example, the explanation includes the following steps S801 to S807. Wherein:

[0211] Step S801: Receive the medical robot task object through a unified entry point, perform a total boundary check on the control parameters in the task object, lock the control parameters that pass the check as the execution range, and generate a task identifier.

[0212] Step S802: Identify the source type and change granularity type of the dynamic control logic to be released, generate a unique identifier, version information and content hash value of the dynamic control logic, and determine whether to trigger re-sandbox verification and re-authorization based on the change granularity type.

[0213] Step S803: Perform sandbox verification on the dynamic control logic in an isolated simulation environment to obtain the verification results.

[0214] Step S804: If the verification result is passed, obtain assessment data from multiple dimensions, calculate the risk score by weighting the assessment data from multiple dimensions, map the risk score to the corresponding risk level, match the corresponding authorization level according to the risk level, and generate an authorization application.

[0215] Step S805: Push the authorization request to the matching authorized person's terminal.

[0216] For the specific implementation process and technical effects of steps S801 to S805 in this embodiment, please refer to [link to documentation]. Figure 2 The relevant descriptions of steps S201 to S205 in the method embodiment shown will not be repeated here.

[0217] Step S806: Receive the rejection instruction returned by the authorizer, record the reason for rejection and start the feedback loop, write the reason for rejection into the evidence chain and feed it back to the developer, doctor or task generator.

[0218] In this embodiment, after reviewing the authorization application on the authorizing terminal, if the authorizing person believes that the dynamic control logic is unsuitable for the current patient, the parameter settings do not meet clinical requirements, the risk level assessment is inaccurate, or there are other circumstances that do not meet the authorization conditions, a rejection instruction can be returned. When rejecting, the authorizing person needs to select or fill in the rejection reason on the authorization interface. The categories of rejection reasons include, but are not limited to: parameter violation (a certain control parameter exceeds the clinical safety range recognized by the authorizing person), risk level mismatch (the authorizing person believes that the risk level assessed by the system is too low or too high), lack of necessary attachments (such as lack of patient informed consent, lack of preoperative assessment report, etc.), clinical contraindication (the operation involved in the control logic conflicts with the patient's specific contraindications), and other rejection reasons defined by the authorizing person.

[0219] For example, after receiving a rejection instruction, the publishing system writes the complete rejection record into the evidence chain, including: rejection timestamp, authorizing person's identity information, rejection reason category, specific rejection description text, current task status, patient information summary, etc. At the same time, the publishing system initiates a feedback loop, feeding back the rejection reason to the task initiator (developer, doctor, or task generator) through a message channel.

[0220] Step S807 allows modification of the control logic based on the reason for rejection. The modified control logic needs to be re-verified through the sandbox. Only after the sandbox verification is passed can the authorization request be generated again.

[0221] In this embodiment, after receiving feedback on the reason for rejection, the task initiator can modify the control logic based on the reason for rejection. Modification methods include, but are not limited to: adjusting control parameters (such as reducing speed, reducing torque, shortening training time), reducing risk level (such as deleting high-risk nodes, turning off force control mode, increasing safety boundaries), supplementing explanations (such as supplementing patient indications, adding operation comments), and reselecting a template (changing to a lower-risk task template), etc.

[0222] Optionally, after the task initiator completes the modifications, it resubmits the modified control logic to the release system. Upon receiving the resubmitted control logic, the release system mandates that the control logic undergo sandbox verification again. This means re-executing the sandbox verification process in step S803, performing a complete simulation verification of the modified control logic in an isolated simulation environment, without skipping verification on the grounds of "minor changes." The judgment rules for re-sandbox verification are completely consistent with the initial verification: if the verification result is a failure, the generation of an authorization request is prohibited and the process is terminated; if the verification result is a warning, a restricted authorization request with limiting constraints is generated; if the verification result is a pass, re-entry into the risk assessment and authorization application process is allowed.

[0223] It should be emphasized that the release system will only allow the generation of authorization requests and re-enter the process of steps S804 to S806 after the sandbox verification is successful.

[0224] Through the aforementioned mandatory path of "reject, modify, re-verify, re-authorize," it is ensured that any rejected control logic must undergo complete security verification upon resubmission, and there is no way to bypass verification and resubmit directly, thus forming an unbypassable security loop.

[0225] In this embodiment, through the aforementioned authorization rejection feedback closed-loop mechanism, the reason for rejection is recorded and fed back to the developer or doctor, allowing them to modify and resubmit for sandbox verification. This effectively solves the problem of repeated invalid submissions caused by the traditional approval process of "rejection is termination". At the same time, the security of the modified control logic is ensured by forcing re-sandbox verification.

[0226] The methods in the embodiments of this application will be illustrated below with specific application scenarios.

[0227] For example, a specific application scenario is the lower limb rehabilitation training performed on a patient who underwent right total knee replacement at a tertiary hospital's Department of Rehabilitation Medicine. The patient is a 65-year-old male with stable vital signs but mild hypertension, posing a risk of falls; he is classified as a medium-to-high-risk patient. The medical device used is a lower limb rehabilitation robot, which supports both force control and position control modes and is connected to a doctor-authorized robot dynamic control logic publishing system (hereinafter referred to as the system). The specific process is as follows:

[0228] Step 1: Task access and boundary locking.

[0229] In this step, rehabilitation therapist Wang selects the lower limb gait training template V3.2 through the system interface and sets the training parameters: gait speed 0.3m / s, stride length 0.25m, training duration 20 minutes, and assistive force 80%. The system generates a clinical robot task object, which is received through the ADMMI Gateway unified entry point.

[0230] In this step, the system performs a total boundary check: walking speed of 0.3 m / s is below the safe upper limit of 0.5 m / s, stride length of 0.25 m is within the safe range of 0.15 m to 0.35 m, assisted force of 80% is below the upper limit of 90%, and training duration of 20 minutes is below the upper limit of 30 minutes. After the check passes, the system generates a task identifier REHAB-20260701-003 and locks the execution range within the above parameter boundaries.

[0231] Step 2: Identify the source and granularity of control logic changes.

[0232] In this step, the control logic identification module determines that the control logic originates from the task template combination. It detects that Dr. Wang only adjusted the walking speed and training duration based on the template, without adding or removing nodes or reconstructing the topology, thus classifying the change as a parameter-level change. The system generates a control logic identifier CL-REHAB-001, version number V3.2.1, and hash value 0x7F3A9C2B. Because the change is only at the parameter level and all parameters are within the template's preset safety range (walking speed of 0.3m / s is within the low-risk adjustable range of 0.2m / s to 0.4m / s, and training duration of 20 minutes is within the safe range), the system determines that no re-sandbox verification and authorization is required, and directly reuses the previously verified sandbox report SB-REPORT-20260615-008.

[0233] Step 3: Sandbox verification and execution admission determination.

[0234] In this step, the system retrieves the stored sandbox verification report, which shows a "passed" status, confirming that the control logic runs stably in the simulation environment. With verification successful, the system is authorized to proceed to the risk assessment and authorization process.

[0235] Step 4: Risk assessment and authorization application generation.

[0236] In this step, the risk assessment module uses a six-factor weighted scoring system. The factors are: patient harm probability (35 points), task invasiveness (20 points), failure consequences (40 points), patient condition (45 points), equipment contact level (30 points), and control logic change granularity (5 points). The weighted calculation is: 35 × 0.25 = 8.75, 20 × 0.20 = 4.00, 40 × 0.20 = 8.00, 45 × 0.15 = 6.75, 30 × 0.10 = 3.00, 5 × 0.10 = 0.50, for a total Risk Score of 31.00. According to the mapping rules, 0-20 points are R1, 21-50 points are R2, 51-80 points are R3, and 81-100 points are R4. A score of 31.00 falls under R2, corresponding to authorization confirmed by the responsible physician. The system automatically generates an authorization request, including the applicant, task description, risk level R2, and the required authorization level confirmed by the responsible physician.

[0237] Step 5: Doctor authorization and boundary verification.

[0238] In this step, the system pushes the authorization request to the terminal of the attending physician, Director Li. The authorization interface distinguishes between modifiable and non-modifiable items: Modifiable items include execution time, execution location, low-risk training duration, number of actions, voice prompt content, reminder frequency, target ward, low-risk navigation target, and authorization period; Non-modifiable items include safety hard limits, maximum speed, maximum torque, maximum contact force, patient safety distance, prohibited action library, R4 tasks, nodes that have not passed sandbox verification, high-risk ROS2 Task Graph topology, emergency stop logic, doctor takeover logic, and registration verification boundaries.

[0239] In this step, Director Li changed the execution time from 9:00 to 10:30, the training duration from 20 minutes to 15 minutes, confirmed the authorization, and added the on-site monitoring condition. After the system verification passed, the following information was recorded: authorized personnel number D00897, authorization time 9:15, digital signature 0xE4F9A1C3, authorization type as restricted confirmation authorization, and authorization period valid for today.

[0240] Step 6: Authorization validity and version access control verification.

[0241] In this step, the system verifies that the authorization is completed within 15 minutes without timeout, is valid for the day and has not expired, has not been revoked, and matches the scope and task. After all verifications are passed, the version access control audit is performed to confirm that the control logic version V3.2.1, the sandbox report SB-REPORT-20260615-008 is in a passed state, and the authorization version AUTH-20260701-003 is signed by Director Li. A conclusion allowing publication is generated and attached with the conditions for on-site monitoring, and an authorization chain record is generated simultaneously.

[0242] Step 7: Differentiated release and control during execution.

[0243] In this step, the system implements a restricted deployment, assigning control logic to the robot with the following constraints: execution limited to 10:30-11:00 AM today, location control mode, doctor's presence required, and execution only permitted after verification of Director Li's presence via camera and facial recognition. Director Li arrived at 10:30 AM to confirm, and the robot began training. After 8 minutes of training, the system detected that the patient's heart rate increased from 72 beats / min to 98 beats / min, and blood pressure increased from 135 / 85 mmHg to 155 / 95 mmHg, assessing the risk as R3. Director Li simultaneously paused the operation. The system immediately executed a soft stop, halting all movement within 2 seconds, slowly returning the patient's lower limbs to their initial position, and issuing alarms to the nurses' station and the doctor's terminal.

[0244] Step 8: Evidence chain preservation and audit traceability.

[0245] In this step, the system generates a complete chain of evidence, recording 11 key events in sequence: task creation, boundary check, change granularity identification, risk assessment, authorization application, authorization confirmation, restricted release, start of execution, anomaly monitoring, soft stop, and task termination. Each block is linked through a hash chain to form an immutable audit record, which supports exporting audit files with digital signatures for review and traceability.

[0246] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0247] Based on the same inventive concept, this application also provides a licensing and publishing device for the dynamic control logic of a medical robot, which implements the licensing and publishing method for the dynamic control logic of the medical robot described above. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the licensing and publishing device for the dynamic control logic of a medical robot provided below can be found in the limitations of the licensing and publishing method for the dynamic control logic of a medical robot described above, and will not be repeated here.

[0248] In the eighth exemplary embodiment, such as Figure 9 As shown, an authorization and release device for dynamic control logic of a medical robot is provided, including: a task access and boundary control module 901, a control logic identification module 902, a sandbox verification and access module 903, a risk assessment module 904, an authorization decision module 905, an audit module 906, a release and execution control module 907, and an evidence chain and audit module 908, wherein:

[0249] The task access and boundary control module 901 is used to receive medical robot task objects through a unified entry point, perform a total boundary check on the control parameters in the task object, lock the control parameters that pass the check as the execution range, and generate a task identifier.

[0250] The control logic identification module 902 is used to identify the source type and change granularity type of the dynamic control logic to be released, generate a unique identifier, version information and content hash value of the dynamic control logic, and determine whether to trigger re-sandbox verification and re-authorization based on the change granularity type.

[0251] The sandbox verification and admission module 903 is used to perform sandbox verification on dynamic control logic in an isolated simulation environment and obtain verification results.

[0252] The risk assessment module 904 is used to obtain assessment data from multiple dimensions when the verification result is passed, calculate a risk score based on the weighted assessment data from multiple dimensions, map the risk score to the corresponding risk level, match the corresponding authorization level according to the risk level, and generate an authorization application.

[0253] The authorization decision module 905 is used to push the authorization application to the matching authorizing terminal, receive the authorization operation instructions returned by the authorizing person, and execute the corresponding operation according to the authorization operation instructions;

[0254] The audit module 906 is used to perform version access control audits on the authorized dynamic control logic, verify the validity of the authorization, and generate a release conclusion that allows, restricts, or prohibits release based on the audit results.

[0255] The release and execution control module 907 is used to execute the corresponding release operation based on the release conclusion, and continuously verify the authorization status during the execution of dynamic control logic. When the authorization expires, the corresponding security response action is triggered according to the task risk level and the current execution status.

[0256] The evidence chain and audit module 908 is used to record key events and operation records of the entire process of dynamic control logic from task access to execution completion, and to store them in an immutable format to form an evidence chain. The evidence chain is used to support audit traceability and version rollback management.

[0257] In some embodiments, the source type includes any one or more of the following: external AI real-time generation, rule engine generation, medical task template combination, parameter adjustment, doctor modification authorization, developer resubmission, and user script configuration.

[0258] Change granularity types include parameter-level changes, node-level changes, topology-level changes, behavior tree-level changes, security boundary-level changes, and version-level changes;

[0259] Determine whether to trigger re-sandbox verification and re-authorization based on the type of change granularity, including:

[0260] If the change granularity type is a parameter-level change and the parameter-level change is within the preset security domain, re-sandbox verification and re-authorization are exempted; if the change granularity type is a parameter-level change and the parameter-level change is not within the preset security domain, re-sandbox verification and re-authorization are triggered; if the change granularity type is any of the node-level change, topology-level change, behavior tree-level change, security boundary-level change, or version-level change, re-sandbox verification and re-authorization are triggered.

[0261] In some embodiments, the multi-dimensional assessment data includes six dimensions: patient injury probability, task invasiveness, failure consequences, patient status, device access level, and change granularity type.

[0262] The risk levels are divided into four levels: R1, R2, R3, and R4, arranged from low to high risk.

[0263] The system matches the corresponding authorization level based on the risk level, including: R1 level matches automatic authorization by the system or random checks afterward; R2 level matches authorization confirmed by the attending physician, rehabilitation therapist, or nurse station; R3 level matches pre-approval authorization by the attending physician or senior physician; and R4 level matches publication is prohibited by default.

[0264] In some embodiments, the authorization operation instruction includes one or more of the following: a confirmation instruction, a restriction confirmation instruction, a modification authorization instruction, and a denial instruction;

[0265] The system executes corresponding operations based on authorized operation instructions, including: upon receiving a confirmation instruction, performing version access control review; upon receiving a restriction confirmation instruction, recording the additional constraints imposed by the authorizing party in the restriction confirmation instruction and performing version access control review; upon receiving a modification authorization instruction, verifying whether the modified content exceeds the preset authorization boundaries; if it does, rejecting the modification and prompting a resubmission; if it does not exceed the boundaries, recording the modified content and performing version access control review; upon receiving a rejection instruction, recording the rejection reason and initiating a feedback loop, writing the rejection reason into the evidence chain and feeding it back to the developer, doctor, or task generator; it allows modification of the control logic based on the rejection reason, and the modified control logic needs to be re-verified through a sandbox; only after passing the sandbox verification is authorization requests allowed to be generated again.

[0266] In some embodiments, the preset authorization boundaries include: modifiable items and non-modifiable items; the modifiable items include one or more of the following: execution time, execution location, low-risk training duration, number of actions, voice prompt content, reminder frequency, target ward, low-risk navigation target, and authorization period; the non-modifiable items include one or more of the following: safety hard limit, maximum speed, maximum torque, maximum contact force, patient safety distance, prohibited action library, R4 risk level task, control logic node that has not passed sandbox verification, high-risk topology, emergency stop logic, doctor takeover logic, and registration verification boundaries.

[0267] In some embodiments, verifying the validity of authorization includes at least one of the following: verifying whether the current authorization has timed out, verifying whether the current authorization has expired, verifying whether the current authorization has been revoked, and verifying whether the scope of the current authorization matches the current task. During the authorization waiting period, the response time of the authorizing person is monitored. If authorization timeout occurs, the following differentiated processing is performed: if an R1 risk level task does not respond after timeout, it is transferred to post-event spot check or automatically downgraded for execution; if an R2 risk level task does not respond after timeout, it is entered into the manual review queue or its release is restricted; if an R3 risk level task does not respond after timeout, its release is prohibited and higher-privilege roles are notified; wherein, R4 risk level tasks shall not be automatically released due to timeout.

[0268] In situations requiring first aid or to prevent further harm from equipment, methods also include:

[0269] Temporary authorization is initiated by a higher-privilege role. Temporary authorization only allows the execution of safety actions such as stopping, evacuating, slow navigation, safe rollback, clearing dangerous situations, and notifying personnel. It prohibits the addition of new diagnoses, prescriptions, treatment parameter adjustments, intrusion operations, or unverified control logic.

[0270] In some embodiments, restricted release includes adding one or more of the following constraints: low-speed operation only, position control only, force control output prohibited, sandbox operation only, and physician presence required.

[0271] When authorization expires, corresponding safety response actions are triggered based on the task risk level and current execution status, including: prohibiting the initiation of new tasks when authorization has expired and the task has not yet started; stopping subsequent execution after the current step is completed when authorization has expired and a low-risk, uninterruptible step is being executed; triggering a safety stop or soft stop after authorization is revoked or R3 risk level task authorization expires, and then reverting to the previous safety control state or switching to doctor takeover; stopping high-risk outputs, retaining monitoring data, and generating alarm prompts when contact actions, force-controlled actions, or physiotherapy outputs are involved and authorization has expired.

[0272] In some embodiments, the sandbox verification and admission module 903 is further configured to prohibit the generation of authorization requests and terminate the process if the verification result is a failure; and to generate a restricted authorization request with limiting constraints if the verification result is a warning, the limiting constraints being used to be attached to the dynamic control logic to be released in subsequent release steps.

[0273] Each module in the aforementioned authorization and distribution device for the dynamic control logic of the medical robot can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0274] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 10 As shown, the computer device includes a processor, memory, input / output interfaces, a communication interface, a display unit, and an input device. The processor, memory, and input / output interfaces are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements a licensing method for dynamic control logic of a medical robot. The display unit of the computer device is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0275] Those skilled in the art will understand that Figure 10 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0276] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method steps of the various embodiments described above.

[0277] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the method steps of the various embodiments described above.

[0278] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the method steps of the various embodiments described above.

[0279] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0280] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0281] 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 application.

[0282] The above embodiments are merely illustrative of several implementation methods of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of this 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 application should be determined by the appended claims.

Claims

1. A method for licensing and releasing dynamic control logic for a medical robot, characterized in that, The method includes: The system receives medical robot task objects through a unified entry point, performs a total boundary check on the control parameters in the task objects, locks the control parameters that pass the check as the execution range, and generates a task identifier. Identify the source type and change granularity type of the dynamic control logic to be released, generate a unique identifier, version information and content hash value of the dynamic control logic, and determine whether to trigger re-sandbox verification and re-authorization based on the change granularity type; The dynamic control logic is subjected to sandbox verification in an isolated simulation environment to obtain the verification results; If the verification result is passed, assessment data from multiple dimensions is obtained, a risk score is calculated by weighting the assessment data from multiple dimensions, the risk score is mapped to the corresponding risk level, and the corresponding authorization level is matched according to the risk level to generate an authorization application. The authorization request is pushed to the matching authorization user's terminal, the authorization operation instruction returned by the authorization user is received, and the corresponding operation is executed according to the authorization operation instruction; The system performs version access control audits on the authorized dynamic control logic, verifies the validity of the authorization, and generates a release conclusion that allows, restricts, or prohibits release based on the audit results. The corresponding release operation is executed according to the release conclusion, and the authorization status is continuously verified during the execution of the dynamic control logic. When the authorization fails, the corresponding security response action is triggered according to the task risk level and the current execution status. The key events and operation records of the dynamic control logic from task access to execution completion are recorded and stored in an immutable format to form an evidence chain, which is used to support audit traceability and version rollback management.

2. The method according to claim 1, characterized in that, The source types include any one or more of the following: external AI real-time generation, rule engine generation, medical task template combination, parameter adjustment, doctor modification authorization, developer resubmission, and user script configuration. The change granularity types include parameter-level changes, node-level changes, topology-level changes, behavior tree-level changes, security boundary-level changes, and version-level changes. The step of determining whether to trigger re-sandbox verification and re-authorization based on the change granularity type includes: If the change granularity type is a parameter-level change, and the parameter-level change is within a preset security domain, then re-sandbox verification and re-authorization are exempted. If the change granularity type is a parameter-level change, and the parameter-level change is not within the preset security domain, then re-sandbox verification and re-authorization are triggered. If the change granularity type is any of node-level change, topology-level change, behavior tree-level change, security boundary-level change, or version-level change, re-sandbox verification and re-authorization will be triggered.

3. The method according to claim 1, characterized in that, The assessment data across multiple dimensions includes six dimensions: patient risk, task invasiveness, consequences of failure, patient status, degree of device access, and the type of change granularity. The risk levels include four levels: R1, R2, R3, and R4, arranged from low risk to high risk. The process of matching the corresponding authorization level based on the risk level includes: R1 level matching system automatically authorizes or conducts random checks afterward; R2 level matching requires authorization from the attending physician, rehabilitation therapist, or nursing station. R3 level matching pre-approval authorization from attending physician or senior physician; R4 level matching is prohibited from publishing by default.

4. The method according to claim 1, characterized in that, The authorization operation instructions include one or more of the following: confirmation instructions, restriction confirmation instructions, modification authorization instructions, and denial instructions; The step of executing the corresponding operation according to the authorized operation instruction includes: Upon receiving the confirmation instruction, perform version access control verification; Upon receiving the restriction confirmation instruction, record the additional constraints imposed on the authorized person carried in the restriction confirmation instruction, and perform version access control audit; Upon receiving the modification authorization instruction, the system checks whether the modified content exceeds the preset authorization boundary. If it does, the modification is rejected and a resubmission is prompted. If it does not exceed the boundary, the modified content is recorded and version access control review is performed. Upon receiving the rejection instruction, the reason for rejection is recorded and a feedback loop is initiated. The reason for rejection is written into the evidence chain and fed back to the developer, doctor, or task generator. It is allowed to modify the control logic based on the reason for rejection. The modified control logic needs to be re-verified through the sandbox. Only after the sandbox verification is passed can the authorization request be generated again.

5. The method according to claim 4, characterized in that, The preset authorization boundaries include: modifiable items and non-modifiable items; The modifiable items include one or more of the following: execution time, execution location, low-risk training duration, number of actions, voice prompt content, reminder frequency, target ward, low-risk navigation target, and authorization period; The unmodifiable items include one or more of the following: safety hard limits, maximum speed, maximum torque, maximum contact force, patient safety distance, prohibited action library, R4 risk level tasks, control logic nodes that have not passed sandbox verification, high-risk topology, emergency stop logic, doctor takeover logic, and registration verification boundaries.

6. The method according to claim 1, characterized in that, The verification of authorization validity includes at least one of the following: verifying whether the current authorization has timed out, verifying whether the current authorization has expired, verifying whether the current authorization has been revoked, and verifying whether the scope of the current authorization matches the current task. During the authorization waiting period, monitor the response time of the authorizing party. If an authorization timeout occurs, perform the following differentiated processing: If a task at risk level R1 fails to respond within the timeout period, it will be subject to post-event spot checks or automatic downgrade execution. If an R2 risk level task fails to respond within the timeout period, it will be placed in the manual review queue or its release will be restricted. When an R3 risk level task times out without a response, it is prohibited to publish and notify higher-privilege roles. Among them, R4 risk level tasks must not be automatically published due to timeout; In scenarios involving emergency rescue or preventing further harm from the equipment, the method further includes: Temporary authorization is initiated by a higher-privilege role. The temporary authorization only allows the execution of safety actions such as stopping, evacuating, slow navigation, safe rollback, clearing dangerous situations, and notifying personnel. It prohibits the addition of new diagnoses, prescriptions, treatment parameter adjustments, intrusion operations, or unverified control logic.

7. The method according to claim 1, characterized in that, The restricted release includes one or more of the following constraints: low-speed operation only, position control only, force control output prohibited, sandbox operation only, and physician presence required. The provision that triggers a corresponding security response action based on the task risk level and current execution status when authorization expires includes: Prevent new tasks from starting if the authorization has expired and the task has not yet started; If the authorization has expired and a low-risk, uninterruptible step is in progress, stop subsequent execution after completing the current step; When authorization is revoked or R3 risk level task authorization expires, a safety stop or soft stop is triggered, and the system reverts to the previous safety control state or switches to doctor takeover. When high-risk outputs involve contact actions, force-controlled actions, or physical therapy outputs, and the authorization has expired, stop the outputs, retain monitoring data, and generate alarm notifications.

8. The method according to any one of claims 1 to 7, characterized in that, After obtaining the verification result, the method further includes: If the verification result fails, the generation of authorization requests will be prohibited and the process will be terminated. If the verification result is a warning, a restricted authorization request with limiting constraints is generated. These limiting constraints are used to attach to the dynamic control logic to be released in subsequent release steps.

9. An authorization and distribution device for dynamic control logic of a medical robot, characterized in that, The device includes: The task access and boundary control module is used to receive medical robot task objects through a unified entry point, perform a total boundary check on the control parameters in the task objects, lock the control parameters that pass the check as the execution range, and generate a task identifier. The control logic identification module is used to identify the source type and change granularity type of the dynamic control logic to be released, generate a unique identifier, version information and content hash value of the dynamic control logic, and determine whether to trigger re-sandbox verification and re-authorization based on the change granularity type. The sandbox verification and admission module is used to perform sandbox verification on the dynamic control logic in an isolated simulation environment and obtain the verification results. The risk assessment module is used to acquire assessment data from multiple dimensions when the verification result is passed, calculate a risk score based on the weighted assessment data from multiple dimensions, map the risk score to the corresponding risk level, match the corresponding authorization level according to the risk level, and generate an authorization application. The authorization decision module is used to push the authorization application to the matching authorizing person's terminal, receive the authorization operation instruction returned by the authorizing person, and execute the corresponding operation according to the authorization operation instruction; The review module is used to perform version access control review on the authorized dynamic control logic, verify the validity of the authorization, and generate a release conclusion that allows release, restricts release, or prohibits release based on the review results; The release and execution control module is used to execute the corresponding release operation according to the release conclusion, and continuously verify the authorization status during the execution of the dynamic control logic. When the authorization fails, the corresponding security response action is triggered according to the task risk level and the current execution status. The evidence chain and audit module is used to record key events and operation records of the entire process of the dynamic control logic from task access to execution completion, and to store them in an immutable format to form an evidence chain. The evidence chain is used to support audit traceability and version rollback management.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 8.