Data-efficient domain adaptation with only a few labelled data is desired for many robotic applications, e.g., in grasping detection, the inference skill learned from a grasping dataset is not universal enough to directly apply on various other daily/industrial applications. This paper presents an approach enabling the easy domain adaptation through a novel grasping detection network with confidence-driven semi-supervised learning, where these two components deeply interact with each other. The proposed grasping detection network specially provides a prediction uncertainty estimation mechanism by leveraging on Feature Pyramid Network (FPN), and the mean-teacher semi-supervised learning utilizes such uncertainty information to emphasizing the consistency loss only for those unlabelled data with high confidence, which we referred it as the confidence-driven mean teacher. This approach largely prevents the student model to learn the incorrect/harmful information from the consistency loss, which speeds up the learning progress and improves the model accuracy. Our results show that the proposed network can achieve high success rate on the Cornell grasping dataset, and for domain adaptation with very limited data, the confidence- driven mean teacher outperforms the original mean teacher and direct training by more than 10% in evaluation loss especially for avoiding the overfitting and model diverging.

The industrial robot usually has high repeatability but relatively lower accuracy. Therefore, error compensation plays a pivotal role in many industrial robotic applications with high accuracy requirement. In this paper, we present a novel computational method that utilizes a hybrid model that consists of Local Product-Of-Exponential (POE) and Gaussian Process Regression (GPR) to compensate the positioning errors of the industrial robotic manipulator for high accuracy industrial robotic applications. Specifically in the proposed method, the Local POE calibration method is first applied to calibrate the robot forward kinematic model to reduce the geometric error. Then the GPR is applied to learn the inverse kinematic model to further compensate the residual error in task space. We also demonstrate the robustness and effectiveness of our proposed method by showing the reduction of norm pose error by up to 37.2%, compared to the existing methods with multiple datasets.

Environment understanding, object detection and recognition are crucial skills for robots operating in the real world. In this paper, we propose a Convolutional Neural Network with multi-task objectives: object detection and scene classification in one unified architecture. The proposed network reasons globally about an image to understand the scene, hypothesize object locations, and encodes global scene features with regional object features to improve object recognition. We evaluate our network on the standard SUN RGBD dataset. Experiments show that our approach outperforms state-of-the-arts. Network predictions are further transformed into continuous robot beliefs to ensure temporal coherence and extended to 3D space for robotics applications. We embed the whole framework in Robot Operating System, and evaluate its performance on a real robot for semantic mapping and grasp detection.

This paper presents a novel view planning method to generate suitable viewpoints for the reconstruction of the 3D shape of buildings, based on publicly available 2D map data. The proposed method first makes use of 2D map data, along with estimated height information, to generate a rough 3D model of the target building. Randomized sampling procedures are then employed to generate a set of initial candidate viewpoints for the reconstruction process. The most suitable viewpoints are selected from the candidate viewpoint set by first formulating a modified Set Covering Problem (SCP) which considers image registration constraints, as well as uncertainties present in the rough 3D model. A neighborhood greedy search algorithm is proposed to solve this SCP problem and select a series of individual viewpoints deemed most suitable for the 3D reconstruction task. The paper concludes with both computational and real-world field tests to demonstrate the overall effectiveness of the proposed method.

Robot calibration is critical for industrial robot applications that require high accuracy. This paper presents a novel calibration method that utilizes Product-Of-Exponential (POE) and Gaussian Process (GP) regression to compensate for both geometric and non-geometric errors within the robot manipulator. Effects of a payload at the end-effector is also considered in the GP regression model in order to further improve robot positioning accuracy in the task space. Simulation and experimental results demonstrate the effectiveness of the proposed method. The experimental results show that the proposed method reduces norm pose error by 65.5% and 50.2% on average compared to conventional base-tool calibration and POE calibration respectively.

Robot calibration is to improve the accuracy of the robot model so as to achieve better positioning accuracy within the robot work cell. Model based calibration approaches are in general limited to compensating for geometric errors and are unable to compensate for error sources that do not fit within the proposed robot model. In order to compensate for the unmodeled error sources, a Radial Basis Function (RBF) Neural Network (NN) augmented robot model is proposed together with a two stage calibration process for training the NN. A simulation and an experimental study are conducted to verify the effectiveness of the proposed solution.

Industrial robotic manipulators have excellent repeatability while accuracy is significantly poorer. Numerous error sources in the robotic workcell contributes to the accuracy problem. Modeling and identification of all the errors to achieve the required levels of accuracy may be difficult. To resolve the accuracy issues, a sensor based indirect error compensation approach is proposed in this paper where the errors are compensated online via measurements of the work object. The sensor captures a point cloud of the work object and with the CAD model of the work object, the actual relative pose of the sensor frame and work object frame can be established via a point cloud registration. Once this relationship has been established, the robot will be able to move the tool accurately relative to the work object frame near the point of compensation. A data pre-processing technique is proposed to reduce computation time and prevent a local minima solution during point cloud registration. A simulation study is presented to illustrate the effectiveness of the proposed solution.

This paper proposes a novel approach, Simultaneous Path Planning and Topological Mapping (SP2ATM), to address the problem of path planning by registering the topology of the perceived dynamic environment as opposed to the conventional grid representation. The local topology is encoded, concurrent and incremental with path planning, by extracting only the admissible free space. The resulting Admissible Space Topological Map (ASTM) then serves as the minimum information to facilitate path planning in the 3D configuration space. Experimental results obtained from our mobile robot X1 in a complex planar environment, validates completeness and optimality of the algorithm.

In this paper, coupled dynamics are presented for two cooperating mobile robotic manipulators manipulating an object with relative motion in the presence of uncertainties and external disturbances. Centralized robust adaptive controls are introduced to guarantee the motion, and force trajectories of the constrained object converge to the desired manifolds with prescribed performance. The stability of the closed-loop system and the boundedness of tracking errors are proved using Lyapunov stability synthesis. The tracking of the constraint trajectory/force up to an ultimately bounded error is achieved. The proposed adaptive controls are robust against relative motion disturbances and parametric uncertainties and are validated by simulation studies.