The positioning accuracy of a robot is of great significance in advanced robotic manufacturing systems. This paper proposes a novel calibration method for improving robot positioning accuracy. First of all, geometric parameters are identified on the basis of the product of exponentials (POE) formula. The errors of the reduction ratio and the coupling ratio are identified at the same time. Then, joint stiffness identification is carried out by adding a load to the end-effector. Finally, residual errors caused by nongeometric parameters are compensated by a multilayer perceptron neural network (MLPNN) based on beetle swarm optimization algorithm. The calibration is implemented on a SIASUN SR210D robot manipulator. Results show that the proposed method possesses better performance in terms of faster convergence and higher precision.

Robot calibration is used to improve the accuracy of the kinematic model to achieve the higher positioning accuracy within the workspace. Due to some nongeometrical reasons such as joint and link flexibility, the errors are unevenly distributed in the workspace. In this case, it is difficult for the existing methods used to improve the absolute positioning accuracy to achieve good results in each region, especially for robots with large self‐weights. In this paper, a novel calibration method is proposed, which deals with joint deflection dependent errors to enhance the robot positioning accuracy in the whole workspace. Firstly, the joint angle workspace is divided into several local regions according to the mass distribution of the robot. Then, its geometric parameters are modeled and identified using the Denavit–Hartenberg (DH) model in each region and in the whole workspace separately. Since the nongeometric error sources are difficult to model correctly, an artificial neural network (ANN) is applied to compensate for the nongeometric errors. Finally, the experiments using an 8 degree‐of‐freedom (DOF) engineering robot are conducted to demonstrate the validity of the proposed method. The combination of the joint angle division and ANN could be an effective solution for the robot calibration, especially for one with a large self‐weight.

Kinematic calibration of manipulators is an efficient and fundamental way to ensure reliability and high performance of robots. Research on kinematic calibration has a long tradition, and a common strategy used to calibration is to guarantee the least errors in the sense of root mean square deviation. However, the absolute positioning accuracy is determined by the maximum error of manipulators, and it is a key indicator for evaluating performance. For example, using manipulators to print machine elements, obviously where the error is the most may likely cause inaccurate fit. Hence, it is crucial to study a robust calibration strategy. Considering the calibration problem, both positioning and orientation accuracy are ensured by minimizing the maximum positioning errors of three spherical mounted retro-reflectors (SMRs) on the end effector of manipulators. Unfortunately, traditional optimization methods based on gradient cannot be employed directly to solve the minimax problem. Due to the recent progresses on optimization, researchers found that the minimax can be transformed into sequence quadratic programming problems under inequality conditions, thus providing solutions for solving the robust calibration. This paper applied this method to convert the calibration problem into constrained quadratic subproblems, and the subproblems can be solved through primal-dual subgradient method. Then, convexity and robustness analysis is given to prove that these subproblems can quickly converge to local minimum. Finally, to verify the validity of the proposed algorithm, the experiments are conducted on an IRB 2600 manipulator, and the results show that, with the minimax search algorithm, both the positioning and orientation accuracy is enhanced by 67.34% and 73.14%, respectively, which is significantly higher than the performance of the single-SMR calibration algorithm widely used in the field of industry.

Calibration is a vital method for improving the performance of robots. While the Denavit-Hartenberg (DH) model is a widely used framework in robot calibration, challenges remain in accurately compensating for joint deformations caused by the robot’s weight or end payload, which negatively impact absolute positioning accuracy. Additionally, conventional methods can encounter convergence issues during optimization. This paper proposes a robust calibration algorithm that addresses these challenges by leveraging the unique characteristics of the Modified Denavit-Hartenberg (MDH) model, considering joint deformation, and the separable nonlinear least squares (SNLS) algorithm. The method utilizes the MDH model’s ability to separate linear and nonlinear parameters, transforming the calibration task into an SNLS optimization problem. This enhances robustness during parameter identification by eliminating partial linear model parameters. The SNLS algorithm also reduces the number of nonlinear parameters to be identified, resulting in fewer iterations and faster convergence, significantly decreasing dependency on initial values. Numerical and experimental results demonstrate the effectiveness of the proposed SNLS algorithm. The method significantly improves robot positioning accuracy, particularly under varying load conditions, from 9.43mm to 0.71mm, a 92.47% improvement. These findings highlight the substantial advances and importance of the proposed calibration algorithm in enhancing robot performance and reliability.

Adaptive control policies for robots often require balancing generalization from large offline datasets with efficient adaptation to specific deployment conditions. In this paper, we propose Adaptive Diffusion Policy (ADP), a two-stage framework that integrates temporal-aware diffusion models with parameter-efficient LoRA adaptation. First, in the learning stage, ADP imitates and generates actions based on image and video signals from a meager amount of expert demonstrations, considering both spatial and temporal information. This component contrasts with most existing works, which focus solely on spatial information. Second, in the practice stage, ADP incorporates a low-rank adaptation module into the policy, subsequently training it using residual reinforcement learning with minimal environment interaction. Experiments conducted on Meta-World benchmark demonstrate the efficiency of each ADP component and the superiority of ADP over representative baseline methods.