A Robust Safety Control Method for Robotic Arms Based on Input-State Safety
By adopting a robust safety control method for robotic arms based on input-state safety, the problem of safety constraint failure in traditional methods when facing unknown disturbances is solved, and smooth control and safety assurance of robotic arms in complex environments are achieved.
Patent Information
- Application Number
- CN202611522869.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-29
- Publication Date
- 2026-11-17
AI Technical Summary
Traditional control methods struggle to balance real-time safety constraints with control performance in robotic arm systems, especially when faced with external disturbances, uncertainties in system parameters, and dynamic environmental interactions, which can easily lead to system performance degradation and safety hazards.
A robust safety control method for robotic arms based on input-state safety is adopted. By establishing a dynamic model, designing safety functions and robust control barrier functions, and constructing virtual control laws and safety constraints, the system can be ensured to maintain safety under unknown bounded disturbances.
When faced with unknown bounded disturbances, it avoids the failure of safety constraints caused by observer estimation errors, ensures that the joint state of the robotic arm is always within the preset safety range, achieves smooth control command output, and avoids high-frequency vibration and boundary risks.
Smart Images

Figure CN123103673A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of robotic arm control technology, specifically relating to a robust safety control method for systems based on input-to-state safety. Background Technology
[0002] With the rapid development of robotic arm technology in various fields such as industrial automation, medical-assisted surgery, warehousing and logistics, intelligent manufacturing, and home services, its operating environment is becoming increasingly complex and its task objectives are becoming more diversified, posing unprecedented challenges to the safety of control systems. In practical applications, robotic arm systems often face a variety of uncertainties, such as external disturbances, uncertain system parameters, model nonlinearity errors, friction, load changes, and dynamic environmental interactions. If these factors are not effectively handled, they can easily lead to system performance degradation or even state out-of-bounds errors, resulting in serious safety hazards and affecting the accuracy of task completion and system stability, especially in scenarios with extremely high safety requirements, such as medical robotic arms and collaborative robotic arms.
[0003] Traditional control methods, such as PID control, robust control, optimal control, or adaptive control, can suppress system uncertainties to a certain extent, but they often struggle to balance control performance with ensuring real-time safety constraints. Furthermore, most of these methods rely on relatively ideal model assumptions, making it difficult to respond in real-time and proactively defend against discontinuous, spike, or high-frequency non-guided disturbances in real-world engineering scenarios. When the disturbance is non-differentiable, the observer estimation error is significantly amplified in the transient phase, leading to the failure of compensation-based safety constraints and even causing system state out-of-bounds errors. Therefore, this application proposes a robust safety control method for robotic arms based on input-to-state safety. Summary of the Invention
[0004] The purpose of this invention is to provide a robust safety control method for robotic arms based on input-to-state safety, so as to ensure the safety of the system under the influence of unknown bounded disturbances, and overcome the problems of strong assumptions about the differentiability of disturbances and the difficulty in handling unmatched disturbances in the prior art.
[0005] The specific technical solution adopted by this invention is as follows: A robust safety control method for a robotic arm based on input-to-state safety includes the following steps: Step S1: Establish a dynamic model of the robotic arm that includes joint states, control laws, and unknown bounded disturbances; Step S2: Design a safety function based on the actual safety task, and introduce an adjustment mechanism to construct an adjustable robust control obstacle function; Step S3: Based on the adjustable robust control obstacle function constructed in Step S2, and combining the desired control law of the robotic arm and the input-to-state safety theory, recursively construct the virtual control law and safety constraints that meet the safety requirements; using the desired control law as the nominal control, establish and solve the safety constraint optimization problem to obtain the virtual control law that meets the safety requirements, and then recursively derive the actual safety control law. And output; perform security analysis on the obtained actual security control law τ, prove the forward invariance of the extended security set of the system under unknown bounded disturbances, and prove that the system satisfies input-to-state security on the extended security set; Step S4: The actual safety control law output in step S3. The control commands are input to the underlying drive system of the robotic arm to drive it to perform tasks. This ensures the safety of the robotic arm in performing tasks when unknown bounded disturbances exist.
[0006] The technical effects achieved by this invention are as follows: This invention does not require the assumption of first-order differentiability of external disturbances, nor does it rely on a disturbance observer for real-time reconstruction of disturbances. Therefore, it fundamentally avoids the risk of safety constraint failure caused by estimation lag or error amplification when the observer faces discontinuous disturbances. Real-world verification on the FrankaEmikaPanda robotic arm platform demonstrates that this method produces smooth control command outputs without high-frequency vibrations when subjected to discontinuous disturbances such as rigid impacts and sudden load changes, and the joint state is always strictly limited within a preset safety range. Attached Figure Description
[0007] Figure 1 Schematic diagram of the safe trajectory of a robotic arm under continuous smooth disturbance; Figure 2 Schematic diagram of robotic arm control input under continuous smooth disturbance; Figure 3 Schematic diagram of continuous smooth perturbation; Figure 4 Schematic diagram of the safe state trajectory of a robotic arm under discontinuous step disturbance; Figure 5 Schematic diagram of control input for a robotic arm under discontinuous step disturbance; Figure 6 Schematic diagram of discontinuous step disturbance; Figure 7 A schematic diagram of the safe state trajectory of a robotic arm joint under actual working conditions; Figure 8 A schematic diagram of the control inputs of a robotic arm under actual working conditions; Figure 9 Schematic diagram of step disturbance and smooth surging disturbance under actual working conditions. Detailed Implementation
[0008] To make the objectives and advantages of this invention clearer, the invention will be specifically described below with reference to embodiments. It should be understood that the following text is merely used to describe one or more specific embodiments of the invention and does not strictly limit the scope of protection specifically claimed by the invention.
[0009] like Figures 1-9 As shown, a robust safety control method for a robotic arm based on input-to-state safety includes the following steps: Step S1: Establish a dynamic model of the robotic arm that includes joint states, control laws, and unknown bounded disturbances. Specifically, the dynamic model of the robotic arm is established as follows: (1) in, Let the joint position vector be... For joint velocity vectors, This is the joint acceleration vector; The inertial matrix is symmetric and positive definite. The matrix represents the Coriolis force and the centrifugal force. Let this be the system's gravity vector; For joint control laws; An unknown impact torque exerted on the system by the external environment.
[0010] Rewrite equation (1) in state-space form: (2) in, This represents the joint velocity feedback vector without any artificial perturbation. The matrix of Coriolis force and centrifugal force is given by the joint position q and the velocity feedback v as independent variables. For the non-infinite perturbations of the velocity channel; Defined as: (3) To ensure the designability of the control law, based on the physical nature of the robotic arm, the design of this invention is based on the following system characteristics and reasonable assumptions: Assumption 1: Inertia Matrix It is always positive definite and bounded, and its inverse matrix is... It exists and is bounded. This property ensures that the control gain term always meets the controllability requirements.
[0011] Assumption 2: Coriolis force and centrifugal force matrix and gravity vector Its variables are locally Lipschitz continuous.
[0012] Assumption 3: At the first-order kinematic level, the velocity channel does not have its own nonlinear drift term (i.e., drift term). And the velocity gain matrix is the identity matrix. .
[0013] Assumption 4: Mismatched disturbances in the system and the impact torque applied by the external environment It is physically bounded. (This is in conjunction with Assumption 1.) Given the boundedness of the condition, the equivalent matching perturbation It is also bounded. Therefore, there exist unknown positive constants. , making and Established.
[0014] Step S2: Design a safety function based on the actual safety task, and introduce an adjustment mechanism to construct an adjustable robust control obstacle function; Define the location safety function as follows Its security set is: (4) Subsequently, a control barrier function based on the unperturbed condition is constructed. : (5) In the formula, The output signal of a first-order filter has the following dynamic equation: (6) in, The filtering time constant is For virtual velocity commands, the filter's initial conditions are set to... .
[0015] According to hypothesis 4, and Bounded, meaning there exists an unknown positive constant D that satisfies Due to the presence of disturbances and Disturbances, and the control barrier function h established based on the nominal model, cannot guarantee the safety of the system. Therefore, an adjustable robust control barrier function is defined. for: (7) in, This is an item to be designed.
[0016] The extended security set defined by it is: (8) in, .
[0017] Define the following series of functions: (9) (10) Step S3: Based on the adjustable robust control obstacle function constructed in Step S2, and combining the desired control law of the robotic arm and the input-to-state safety theory, recursively construct the virtual control law and safety constraints that meet the safety requirements; using the desired control law as the nominal control, establish and solve the safety constraint optimization problem to obtain the virtual control law that meets the safety requirements, and then recursively derive the actual safety control law. And output; the resulting actual safety control law We conduct a security analysis, prove the forward invariance of the extended security set of the system under unknown bounded perturbations, and prove that the system satisfies input-to-state security on the extended security set. In step S3, the control law is designed: Under the condition of no disturbance, according to the position safety function There exists a function Makes the following inequality true: (11) in, An extension of the local Lipschitz continuity used for constructing safety constraints. Function-like.
[0018] However, to enhance the system's robustness under disturbances, a robust term is introduced based on the above equation to construct a robust safety control law. for: (12) Substituting equation (12) into equation (11) yields: (13) Using equation (13) as a constraint, the following optimization problem is established: (14) in, The expected virtual control law is expressed as follows: (15) in, and These represent the desired position and desired speed of the robotic arm, respectively. It is a positive definite proportional gain coefficient.
[0019] Define variables , for: (16) Using the Lagrange multiplier method, the explicit solution to the optimization problem (14) is expressed as: (17) Among them, Lagrange multipliers The expression is: (18) Therefore, equation (17) gives the virtual control law that satisfies the safety constraints. Further combining equations (13), (17), and (18), the actual control law can be derived. for: (19) in, For gain parameters, The designed dynamic adjustment function; the dynamic adjustment function belongs to the extended... A class function that satisfies the condition of being strictly monotonically increasing throughout its domain and possesses... The origin characteristics.
[0020] In step S3, security analysis: Define speed tracking error Combining equation (2), we can obtain: (20) Combining equation (20), and differentiating equation (5), we get: (twenty one) Combining equations (11), (12), and (13), we can obtain: (twenty two) This invention designs the actual control law for: (twenty three) Substituting equation (23) into equation (22), we get: (twenty four) The terms in equation (24) By scaling, we can obtain: (25) Similarly, we can conclude that: (26) Substituting the results of equations (25) and (26) into equation (24) and rearranging, we get: (27) in, for The local Lipschitz constant, , .
[0021] Then the extended security set (6) Defined as: (28) in, .
[0022] Based on this, differentiating equation (5) yields: (29) Obviously, The lower bound is determined by and Jointly decided. When the system state is within the extended security set. When the boundary (i.e.) ),have .Depend on and As can be seen from the definition, at this time: (30) From the above derivation, we can conclude that: (31) because And satisfy (32) Therefore, in On the boundary: (33) Furthermore, due to when hour, ,and It is about The function is given by: (34) Combining equation (33), by Nagumo's theorem and the comparison lemma, we can obtain the extended security set. It is forward invariant. Based on the input-to-state safety theory, the system equation (1) in the extended safety set can be obtained. The above is input to a state that is safe.
[0023] Step 4: Input the control law obtained in step S3 as a command into the underlying drive system of the robotic arm, and conduct experimental verification based on the FrankaROS robotic arm experimental platform to ensure the safety of the robotic arm when performing tasks under unknown bounded disturbances.
[0024] This invention considers the following dynamic equations for a two-degree-of-freedom robotic arm: ; in, Let the joint position vector be... For joint velocity vectors, This is the joint acceleration vector; The inertial matrix is symmetric and positive definite. The matrix represents the Coriolis force and the centrifugal force. Let this be the system's gravity vector; This refers to the joint control law for the actual servo motor; An unknown impact torque is exerted on the system by the external environment. The above dynamics are then transformed into a system with mismatched disturbances. Matching perturbation Format: ; Desired motion trajectory setting: In this experiment, the basic desired working trajectory of each joint of the two-degree-of-freedom robotic arm is set as a continuous periodic signal, specifically represented as follows: ; Perturbation scenarios: To evaluate the robustness of the proposed method under different perturbations, the following three perturbation scenarios are considered: Case 1 (Continuous smooth disturbance): ; This scenario is used to simulate the continuous disturbance rejection capability of a robotic arm under periodic external interference; Case 2 (Discontinuous step disturbance): ; Where k is a non-negative integer; This scenario is used to simulate non-differentiable, non-continuous abrupt changes in working conditions, such as sudden external resistance or periodic hard collisions encountered by a robotic arm during operation. Scenario 3 (Actual composite operating condition interference): ; The composite disturbance is: ; This scenario is used to simulate the comprehensive anti-disturbance and safety protection capabilities of a robotic arm under multi-field coupled nonlinear time-varying conditions, including continuous high-intensity operations, simultaneous encounters with thermo-mechanical transient disturbances (such as joint temperature rise, changes in lubrication status, and the gradual attenuation of frictional torque as heat dissipation and load release caused by phased high-load operations) and sudden strong collisions.
[0025] The control objective is to adjust the joint position of the robotic arm. In a dynamically changing operating environment, it always tends to remain at the upper boundary. and lower boundary Defined dynamic safety zone Inside; Wherein, the dynamic boundary changes with time Evolution, specifically, is represented as: Upper boundary: ; Lower boundary: ; The location security function is: Upper boundary: ; Lower boundary: ; The final control law is: Scenario 1: ; in, ; Scenario 2: ; in, ; Scenario 3: ; in, ; Experimental verification and result analysis of this invention: This experiment was conducted on the FrankaEmikaPanda robotic arm platform to analyze the effectiveness of the proposed adjustable robust safety control method. To simulate the complex periodic motion of a wide-field-of-view sweep with the end-effector following a smooth elliptical trajectory in actual industrial inspection, all experiments drove the robotic arm's base joints and wrist pitch joints to execute orthogonal periodic reference trajectories of sine and cosine respectively, thus constructing a typical "wide-field dynamic scanning" basic operating condition. The overall experimental results are as follows: Figures 1 to 9 As shown.
[0026] 1. Continuous smooth perturbation experiment Under the above scanning conditions, a continuous smoothing disturbance, scenario 1, is introduced. and (like Figure 3 As shown), the safe state trajectory of the robotic arm and the control input are respectively as follows: Figure 1 and Figure 2 As shown. Experimental results show that when performing dynamic sweeping tasks, facing continuous unknown fluctuations common in industrial settings, the system state can always remain within the preset safety set under the regulation of the adjustable robust control barrier function (TRCBF), and the generated control input is smooth and continuous without drastic control abrupt changes.
[0027] 2. Discontinuous step disturbance experiment To address the non-differentiable abrupt changes in conditions that may be encountered during scanning operations, such as sudden external resistance or hard collisions, the experiment introduces scenario 2: a discontinuous step-change disturbance. and (like Figure 6 As shown in the figure), the safe trajectory of the robotic arm and the control input at this time are respectively as follows: Figure 4 and Figure 5 As shown in the figure, the experimental results demonstrate that even during the moment of a step change in disturbance during dynamic operation, and in the subsequent steady-state process, the system state is firmly confined within the safety boundary, with no boundary breaches occurring throughout the process. This effectively verifies the strong robustness and safety of the proposed method in dealing with sudden disturbances.
[0028] 3. Actual composite working condition interference test To further verify the practical value of the method in real-world long-term engineering scenarios, the experiment, based on wide-span dynamic scanning, further introduced a composite disturbance condition combining step disturbance and smooth surging disturbance. Figure 9 ).
[0029] This scenario simulates a robotic arm undergoing continuous high-intensity operation, simultaneously encountering gradual transient disturbances such as joint temperature rise and frictional torque changes caused by periodic high loads, and discontinuous disturbances caused by sudden external collisions. Under this combined condition, the safe state trajectory of the robotic arm joints and the control input response are as follows: Figure 7 and Figure 8 As shown in the figure, the observation results indicate that during the rising, peak, and decay phases of smooth surging disturbances, as well as during the abrupt changes in step disturbances, the control input of this method remains smooth throughout, avoiding high-frequency chattering of the system caused by disturbance rejection. Simultaneously, the joint states are strictly constrained within the preset safety space without exceeding the limits. This verifies that this method possesses both robust safety protection capabilities and engineering applicability when facing complex real-world operations and time-varying disturbances.
[0030] The above description is merely a preferred embodiment of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention. Structures, devices, and operating methods not specifically described or explained in this invention are implemented according to conventional methods in the art unless otherwise specified or limited.
Claims
1. A robust safety control method for a robotic arm based on input-to-state safety, characterized in that: Includes the following steps: Step S1: Establish a dynamic model of the robotic arm that includes joint states, control laws, and unknown bounded disturbances; Step S2: Design a safety function based on the actual safety task, and introduce an adjustment mechanism to construct an adjustable robust control obstacle function; Step S3: Based on the adjustable robust control obstacle function constructed in Step S2, and combining the desired control law of the robotic arm and the input-to-state safety theory, recursively construct a virtual control law and safety constraints that meet the safety requirements; using the desired control law as the nominal control, establish and solve the safety constraint optimization problem to obtain the virtual control law that meets the safety requirements, and recursively obtain the actual safety control law τ and output it; perform a safety analysis on the obtained actual safety control law τ, prove the forward invariance of the extended safety set of the system under unknown bounded disturbances, and prove that the system satisfies input-to-state safety on the extended safety set; Step S4: Input the actual safety control law τ output in step S3 as a control command into the underlying drive system of the robotic arm to drive the robotic arm to perform the task.
2. The robust safety control method for a robotic arm based on input-to-state safety according to claim 1, characterized in that: In step S1, a dynamic model of the robotic arm is established, including joint states, control laws, and unknown bounded disturbances; specifically, the dynamic model of the robotic arm is established as follows: (1) in, Let the joint position vector be... For joint velocity vectors, This is the joint acceleration vector; The inertial matrix is symmetric and positive definite. The matrix represents the Coriolis force and the centrifugal force. Let this be the system's gravity vector; For joint control laws; An unknown impact torque exerted on the system by the external environment; Rewrite equation (1) in state-space form: (2) in, This represents the joint velocity feedback vector without any artificial perturbation. The matrix of Coriolis force and centrifugal force is given by the joint position q and the velocity feedback v as independent variables. For the non-infinite perturbations of the velocity channel, Inertia matrix The inverse matrix; For equivalent matching perturbation, Defined as: (3) The design is based on the following system characteristics and reasonable assumptions: Assumption 1: Inertia Matrix It is always positive definite and bounded, and its inverse matrix is... It exists and is bounded; this property ensures that the control gain term always meets the controllability requirements. Assumption 2: Coriolis force and centrifugal force matrix and gravity vector Regarding its variables, they are locally Lipschitz continuous; Assumption 3: At the first-order kinematic level, the velocity channel does not have its own nonlinear drift term, i.e., drift term. And the velocity gain matrix is an identity matrix. ; Assumption 4: Mismatched disturbances in the system and the impact torque applied by the external environment It is physically bounded; combined with hypothesis 1 Given the boundedness of the condition, the equivalent matching perturbation It is also bounded; therefore, there exist unknown positive constants. , making and Established.
3. The robust safety control method for a robotic arm based on input-to-state safety according to claim 1, characterized in that: In step S2, the location security function is defined as follows: and its security set for: (4) Subsequently, a control barrier function based on the unperturbed condition is constructed. : (5) In the formula, The output signal of a first-order filter has the following dynamic equation: (6) in, The filtering time constant is For virtual velocity commands, the filter's initial conditions are set to... ; According to hypothesis 4, and Bounded, meaning there exists an unknown positive constant D that satisfies Due to the presence of disturbances and Disturbance; therefore, an adjustable robust control barrier function is defined. for: (7) in, For items to be designed; Extended security set defined by it for: (8) in, ; Define the following series of functions: (9) (10)。 4. The robust safety control method for a robotic arm based on input-to-state safety according to claim 1, characterized in that: In step S3, the specific process of designing the control law is as follows: Under the condition of no disturbance, according to the position safety function There exists a function Makes the following inequality true: (11) in, An extension of the local Lipschitz continuity used for constructing safety constraints. Class function; Based on equation (11), a robust term is introduced to construct a robust safety control law. for: (12) Substituting equation (12) into equation (11) yields: (13) Using equation (13) as a constraint, the following optimization problem is established: (14) in, The expected virtual control law is expressed as follows: (15) in, and These represent the desired position and desired speed of the robotic arm, respectively. The proportional gain coefficient is positive definite. Define variables , for: (16) Using the Lagrange multiplier method, the explicit solution to the optimization problem (14) is expressed as: (17) Among them, Lagrange multipliers The expression is: (18) Therefore, equation (17) gives the virtual control law that satisfies the safety constraints; combining equations (13), (17) and (18), the actual control law can be derived. for: (19) in, For gain parameters, The designed dynamic adjustment function; the dynamic adjustment function belongs to the extended... A class function that satisfies the condition of being strictly monotonically increasing throughout its domain and possesses... The origin characteristics.
5. A robust safety control method for a robotic arm based on input-to-state safety according to claim 1, characterized in that: In step S3, the specific process of security analysis is as follows: Define speed tracking error Combining equation (2), we can obtain: (20) Combining equation (20), and differentiating equation (5), we get: (21) Combining equations (11), (12), and (13), we can obtain: (22) Design the actual control law for: (23) Substituting equation (23) into equation (22), we get: (24) The terms in equation (24) By scaling, we can obtain: (25) Similarly, we can conclude that: (26) Substituting the results of equations (25) and (26) into equation (24) and rearranging, we get: (27) in, for The local Lipschitz constant, , ; Then the extended security set (6) Defined as: (28) in, ; Based on this, differentiating equation (5) yields: (29) Obviously, The lower bound is determined by and Jointly decided; when the system state is in the extended security set When the boundary (i.e.) ),have ;Depend on and As can be seen from the definition, at this time: (30) From the above derivation, we can conclude that: (31) because And satisfy (32) Therefore, in On the boundary: (33) Furthermore, due to when hour, ,and It is about The function is given by: (34) Combining equation (33), by Nagumo's theorem and the comparison lemma, we can obtain the extended security set. It is forward invariant. Based on the input-to-state safety theory, the system equation (1) in the extended safety set is obtained. The above is input to a state that is safe.
6. A robust safety control method for a robotic arm based on input-to-state safety according to claim 5, characterized in that: Consider the following dynamic equations for a two-degree-of-freedom robotic arm: (35) in, The inertia matrix, The matrix represents the Coriolis force and centrifugal force. The gravity term vector, For control laws, An unknown impact torque exerted on the system by the external environment. To increase the speed after the disturbance; Its specific dynamic parameters are: (36) in, It is gravitational acceleration; The safety task is: to ensure the joint position of the robotic arm is... In a dynamically changing operating environment, it always tends to remain at the upper boundary. and lower boundary Defined dynamic safety zone Inside; The location security function is: Safe distance from the lower bound: ; Safe distance from the upper bound: ; The derivative of the location security function is: The derivative of the safe distance from the lower bound: ; The derivative of the safe distance from the upper bound: ; The dynamically adjustable weight function is: (37) in, This represents the minimum real-time margin between the current state and the upper and lower boundaries. and These are the basic strength coefficient and the dynamic attenuation coefficient, respectively. The control law that ultimately acts on the joints of the robotic arm Designed as follows: (38) in, For feedback gain, Given the known basic disturbance compensation value; To utilize the dynamically adjustable weight function Construct a robust compensation hedging term.
