Search

CN-122008198-A - Double-arm collaborative tail end tracking method integrating global visual angle information

CN122008198ACN 122008198 ACN122008198 ACN 122008198ACN-122008198-A

Abstract

The invention discloses a double-arm collaborative terminal tracking method for fusing global visual angle information, which is characterized in that state information of two mechanical arms is uniformly modeled into a joint state vector, an acceleration attenuation item and a coupling item are introduced to characterize dynamic relevance between the two ends of the two arms, a centralized Kalman filtering algorithm is used for fusing position measurement of different visual angles and establishing a global observation matrix, and finally, mechanical arm terminal tracking is completed according to fused information. Compared with the prior art, the method breaks through the limitation of the traditional independent filtering architecture, can deep dig to a certain extent by influencing the error covariance, improves the tail end tracking stability by utilizing the motion correlation between the two arm tail ends in the cooperative working state, realizes the reduction of the tail end tracking error under the global unified coordinate system, and enables the two arm cooperative tail end tracking algorithm to be applied to task scenes with stronger dynamic performance requirements.

Inventors

  • CAO XIANGHUI
  • SUN HAICHUAN
  • XU HAOCHEN

Assignees

  • 东南大学

Dates

Publication Date
20260512
Application Date
20260127

Claims (5)

  1. 1. A double-arm collaborative terminal tracking method for fusing global visual angle information is characterized in that a global visual angle camera is erected in a double-mechanical arm collaborative environment, global visual angle camera data and target position information obtained by a local visual angle camera equipped by each mechanical arm are fused, dynamic relevance between the tail ends of the double arms is described by introducing acceleration attenuation items and coupling items, a double-arm terminal motion state joint model based on a uniform acceleration motion model is constructed, fusion observation of multi-visual angle information is achieved through a centralized Kalman filtering algorithm, and position tracking of the tail ends of the double mechanical arms is achieved.
  2. 2. The method for tracking the collaborative dual-arm end with the integration of global view information according to claim 1, wherein a collaborative scene of two mechanical arms is constructed, the two mechanical arms are provided with cameras for obtaining visual information of respective views, the cameras are installed on the mechanical arms in such a way that the eyes are on hands, the cameras provided on the mechanical arms can only see the auxiliary positioning equipment of the end of the other mechanical arm, and a global view camera is provided on the top of an operation space of the mechanical arms for assisting in tracking targets, so that the target tracking of the end of each mechanical arm has two cameras as observation sources, and acceleration attenuation coefficients are particularly introduced And synergistic coupling strength Reflecting the cooperative motion relation of the end targets of the double mechanical arms: As an acceleration attenuation function, reflects the damping characteristic of the mechanical arm driving system, To characterize the kinetic effects between the mechanical arms for synergistic coupling strength, Definition mechanical arm Is the continuous time state vector of (1) Wherein the method comprises the steps of As a function of the position of the object, In order to be able to achieve a speed, For acceleration, the system dynamics can be described by the differential equation: analyzing and integrating the continuous model to further establish discretization dynamics modeling of mechanical arm tail end motion state coupling under a double-arm cooperative scene: by constructing a state vector containing the pose, the speed and the acceleration of the tail end of the double mechanical arms, the motion correlation among the mechanical arms is explicitly represented, the mathematical modeling of the cooperative characteristic is realized, Wherein, the ; Is a mechanical arm At a time step Is provided with a tip movement acceleration of (a), Is a mechanical arm At a time step Is provided with a tip movement acceleration of (a), Is a mechanical arm At a time step The speed of the movement of the tip, Is a mechanical arm At a time step End position; Is a fixed sampling period; ; Is a mechanical arm Acceleration dispersion noise of (a); Is a mechanical arm Is a velocity-discrete noise of (2); Is a mechanical arm Wherein, Is the acceleration noise covariance; is the velocity noise covariance; Is the position noise covariance; Is a unit matrix; Is the noise intensity.
  3. 3. The method for tracking the cooperative end of the double arms fused with global view information according to claim 2, wherein the continuous-time attenuation dynamics model and the mechanical arm end acceleration European model are accurately embedded into a discrete filtering framework based on physical constraint characteristics of the movement of the mechanical arm end target to form an improved centralized Kalman filtering framework, and the improved centralized Kalman filtering algorithm specifically comprises the following steps: Step 1, constructing initial state estimation Initial covariance Process noise covariance Observed noise covariance Time step , Step 2, at time step Recursive algorithm based on Kalman filtering Performing state prediction to obtain Time of day prior estimation , wherein, Is that A time system state vector; Is that The state transition matrix at moment represents the dynamic characteristics of the system; in order for the process to be noisy, Step 3, according to the state transition equation , Computing jacobian matrix Wherein the blocking matrix is , Step 4, linearizing the propagation equation according to the jacobian matrix Prediction Time prior covariance matrix , wherein, Is that The posterior covariance of the moment of time, Step 5, according to the block diagonal structure of the observation matrix , Obtaining an observation matrix , wherein, Represents the selective observation of the position component, and can further obtain an observation equation Wherein Is that The vector is observed at the moment of time, In order to observe the noise covariance matrix, Step 6, calculating an equation according to the Kalman gain Updating Kalman gain , Step 7, correcting the equation according to the state estimation Updating the state to obtain Time posterior estimation , wherein, In order to observe the matrix, Step 8, updating the equation according to the covariance Updating Time posterior covariance matrix , Step 9, repeating the steps 2-8 to finally obtain the motion state estimation sequence of the tail end of the double mechanical arms 。
  4. 4. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements a dual-arm collaborative end tracking method incorporating global view information according to any of claims 1-3 when executing the program.
  5. 5. A computer readable storage medium having stored thereon computer instructions which, when executed by a processor, implement a dual-arm collaborative end tracking method incorporating global view information according to any of claims 1-3.

Description

Double-arm collaborative tail end tracking method integrating global visual angle information Technical Field The invention relates to the field of mechanical arm motion planning methods, in particular to a double-arm collaborative tail end tracking method integrating global visual angle information. Background In the process of the evolution from industrial automation to intelligent, the cooperative operation of multiple mechanical arms has become a core technology of the scenes of complex assembly, flexible manufacturing and the like. However, the continuous tracking precision of the target at the tail end of the mechanical arm is severely restricted by depending on visual data in the cooperative operation of multiple mechanical arms, such as dynamic shielding, visual blurring caused by high-speed movement and visual sensor noise. This essentially stems from the inherent limitations of the vision system at a single viewing angle, the narrowness of the viewing angle resulting in dynamic changes in target visibility with the pose of the robotic arm. Kalman filtering is used as a theoretical foundation stone for dynamic system state estimation and plays a key role continuously in the field of robot perception. The algorithm realizes optimal fusion of sensor observation and a dynamics model through a time domain recursion frame, and shows excellent noise suppression capability and calculation efficiency in tasks such as target tracking, pose estimation, multi-sensor fusion and the like. In a mechanical arm visual servo control scene, the conventional method usually relies on a single visual sensor to realize target positioning, however, when faced with challenges such as multi-machine cooperation, high-speed motion, local shielding and the like in a complex industrial environment, the limitation of the single sensor can lead to the significant reduction of the reliability and the continuity of state estimation. In particular, the limited viewing angle is prone to target loss, and motion blur can reduce image feature extraction accuracy. The additive effect of these problems severely restricts the operability of the robotic arm in high dynamic scenarios. In order to break through the performance bottleneck of a single sensor, the multi-source heterogeneous information fusion technology gradually becomes a research focus in the field of robot perception. The existing solution mainly evolves along two directions, namely, one is to configure the state estimation consistency of the tracking system through multi-sensor redundancy, such as fusing vision, inertial navigation and joint encoder data, and the other is to construct a cooperative motion model to capture the state coupling characteristics among the mechanical arms. However, how to effectively coordinate the spatiotemporal asynchrony of multi-source data and how to establish accurate collaborative motion characterization is still a technical problem to be solved. Disclosure of Invention The invention aims to solve the problem of joint modeling and observation fusion of a multi-view sensor in the process of double mechanical arms collaborative operation, and provides a double-arm collaborative tail end tracking method fused with global view information. According to the invention, a global view camera is erected in a double-mechanical-arm cooperation environment, global view camera data is fused with target position information obtained by a local view camera equipped by each mechanical arm, dynamic relevance between the tail ends of the two arms is described by introducing an acceleration attenuation item and a coupling item, a double-arm tail end motion state joint model based on a uniform acceleration motion model is constructed, fusion observation of multi-view information is realized by a centralized Kalman filtering algorithm, and position tracking of the tail ends of the two mechanical arms is realized. The improved centralized Kalman filtering algorithm specifically comprises the following steps: Step 1, constructing initial state estimation Initial covarianceProcess noise covarianceObserved noise covarianceTime step。 Step 2, at time stepRecursive algorithm based on Kalman filteringPerforming state prediction to obtainTime of day prior estimation, wherein,Is thatA time system state vector; Is that The state transition matrix at moment represents the dynamic characteristics of the system; Is process noise. Step 3, according to the state transition equation , Computing jacobian matrixWherein the blocking matrix is , Step 4, linearizing the propagation equation according to the jacobian matrixPredictionTime prior covariance matrix, wherein,Is thatPosterior covariance of time of day. Step 5, according to the block diagonal structure of the observation matrix , Obtaining an observation matrix, wherein,Represents the selective observation of the position component, and can further obtain an observation equationWhereinIs thatThe vector is observed at the mo