Hongtai Cheng

Topological Indoor Localization and Navigation for Autonomous Mobile Robot

Mobile robot typically has limited on-board resources and may be applied in different indoor environment. Thus, it is necessary that they can learn a map and navigate themselves autonomously with lightweight algorithms. A novel topological map-building-based localization and navigation method is proposed in this paper. Based on the depth curve provided by a 3D sensor, a progressive Bayesian classifier is developed to realize direct corridor type identification. Instead of extracting features from single observation, information from multi-observations are fused to achieve a more robust performance. A topological map generation and loop closing method are proposed to build the environment map through autonomous exploration. Based on the derived map and the Markov localization method, the robot can then localize itself and navigate freely in the indoor environment. Experiments are performed on a recently built mobile robot system, and the results verify the effectiveness of the proposed methodology.

Accuracy Analysis of Dynamic-Wafer-Handling Robotic System in Semiconductor Manufacturing

Wafer-handling robots are widely used to transfer wafers in semiconductor manufacturing. To improve the wafer transfer efficiency, optic sensors are installed at each station to estimate the wafer eccentricity on the fly. However, due to nonideal factors such as light beam radius, robot motion error, and system nonlinearity and uncertainties, it is difficult to achieve high-accuracy eccentricity estimation to satisfy the demanding requirements in semiconductor manufacturing. To further improve the eccentricity estimation accuracy, the relationship among the robot kinematic error, sensor calibration error, and eccentricity identification error is analyzed. The analytic results show that both the error modeling method and the global data sampling method can greatly improve the wafer eccentricity estimation accuracy. The proposed eccentricity estimation techniques are verified by experiments performed on a wafer-handling robotic system. The results demonstrate that the developed methods can be used to improve the wafer-handling accuracy and reduce the wafer-handling cycle time in semiconductor manufacturing.

Online parameter optimization in robotic force controlled assembly processes.

In the high precision robotic assembly processes, the process parameters have to be tuned in order to adapt to variations and satisfy the performance requirements. However, because of the modeling difficulty and low efficiency of the existing solutions, this task is usually performed offline. In this paper, an online parameter optimization method is developed. Gaussian Process Regression(GPR) is utilized to model the relationship between the process parameters and system performance. The GPR surrogated Bayesian Optimization Algorithm(GPRBOA) is proposed to optimize the process parameters. To reduce the risk of converging to a local minimum, a random variation factor is added to the Lower Confidence Bound(LCB) acquisition function to balance the exploration and exploitation processes. To deal with the computational burden of GPR, a switching criterion is proposed to coordinate the optimization process and production process to reduce the computational complexity. Experiments were performed using a peg-in-hole process. The experimental results verify the effectiveness of the proposed algorithm and demonstrate its efficiency and accuracy compared to Design Of Experiment(DOE) methods. The proposed method is the first attempt of model-driven assembly process parameter optimization and will generate big economic impact.

Dynamic error modeling and compensation in high speed delta robot pick-and-place process

In Delta Robot based high speed pick and place applications, the key for success is to sense, track and predict the workpiece position precisely in real time. However, because the workpiece is continuously moving on the conveyor, the vision and the robot may not share the same workspace, even the hand eye is precisely calibrated, it is difficult to achieve high precision in high speed system. Dynamics of all components have to be involved, modeled and compensated. In this paper, we first formulate the coordinate frame transformations between camera, conveyor and robot, then analyze the error propagation in the system work flow. By analyzing the sequences of the vision system, the camera trigger delay is modeled and compensated to synchronize the detected object position and the corresponding conveyor position. Instead of using the analytical model and calculate the robot motion time delay, we propose to model the robot motion delay as a fixed part plus a linear part and calibrate these parameters online automatically. A self feed high speed pick and place experimental platform is built and the proposed algorithm is implemented. The experimental results show the effectiveness.

Object handling using Autonomous Industrial Mobile Manipulator

Autonomous Industrial Mobile Manipulators (AIMMs) combine the advantages of both mobile robots and industrial manipulators; therefore they have great mobility, flexibility and functionality. However, what the AIMM brings not only bigger operational range, but also lower manipulation accuracy. Great challenges arise in the sensing, calibration and coordination problems. This paper proposes a solution for the coordination problems of an object handling task for AIMM with limited operational range sensors and manipulators. Different from the existing Image Based Visual Servo(IBVS) with single camera-in-hand configuration, a RGB-D camera with camera-to-hand configuration is utilized. The object is detected by matching the color feature on the target object and measured by calibrating and matching the depth image with the RGB image. The hand-eye calibration between the sensor and the manipulator is performed using the specifically chosen object locations which lie in both the range of the two devices. Furthermore, instead of steering the mobile base along a preplanned trajectory, a position based visual servo controller is designed to guide the robot towards the object. The methods are implemented using an AIMM and the experimental results verify the effectiveness of the proposed coordination and control algorithm.

Energy based Nonlinear Control of Underactuated Brachiation Robot

This paper studies a type of dynamically moving robot modeling a long armed apes locomotion which called brachiation robot. Brachiating is the hand over hand swinging locomotion used by various primates, especially the long armed apes, and the manner of moving like a long armed ape is called brachiation. By analyzing the behavior of the long armed ape, which moves from branch to branch swinging its body like a pendulum, the basic principle and brachiating control of the simplified 2-link brachiation robot is studied. In this paper we present the energy based nonlinear control method to control the swing manner of brachiation, and a feed back control law is designed using the Lyapunov stability theory. The target brachiation is an effective locomotion from one bar to another by exchanging kinetic energy with potential energy like a pendulum, which the same as the real long armed ape uses. Finally the experimental results on a 2-link-model brachiation robot show feasibility of the developed controller design scheme.

Non-horizontal ricochetal brachiation motion planning and control for two-link Bio-primate robot

Ricochetal Brachiation is a sophisticated locomotion for Bio-primate robot. It requires precise cooperation of three different types of locomotion. The tight coupling of locomotion and underactuated characteristics is the main difficulties for realizing ricochetal brachiation. Moreover, for the non-horizontal ricochetal brachiation, which refers brachiating among supporting bars with different heights, another difficulty orients from the asymmetric property of the flight trajectory. To solve the locomotion coupling problem, by building and analyzing the segmented dynamic and kinematic models and considering the constraint and switching condition, this paper proposes a flexible, comprehensive and adaptive ricochetal brachiation motion planning algorithm to obtain the pre-flying posture and post-flying posture. Also, the horizontal and vertical differences between adjacent supporting bars are modeled and included, which makes the algorithm applicable for general configurations. To eliminate the difficulty brought by under-actuated characteristics, virtual constraint based trajectory planning and tracking control method is adopted to ensure that the system can arrive in the pre-flying posture accurately. Finally, a ricochetal brachiation simulation model is built and the results show the effectiveness of the proposed trajectory planning and control strategy.

Efficient hand eye calibration method for a delta robot pick-and-place system

Delta Robot is a kind of parallel robot manipulator specifically designed for the pick-and-place task. Most of such system requires vision to detect the randomly placed workpieces. Therefore, hand eye calibration is a necessary process when installing new product line or modifying the existing robotic configuration. Because the vision and robot do not share the same workspace, and the workpiece move continuously on the conveyor, it is difficult to perform the hand eye calibration efficiently and directly. In this paper, we propose a systematic framework to efficiently calibrate the vision, conveyor and the robot. Beyond two commonly used chess board based calibration board, no external equipments are needed. A virtual coordinate frame which shares the same origin with the vision coordinate frame is imposed on the conveyor. Based on this assumption, two independent experiments can be performed to gather data sets which can be used to optimize the coordinate transformation matrices. Linear Mean Square(LMS) algorithm is used to find the optimal parameters. A self feed high speed pick and place experimental platform is built and the proposed algorithm is implemented. The experimental results show the effectiveness and efficiency. All the system can be calibrated within 10 minutes.

Learning from demonstration enabled robotic small part assembly

The small parts assembly process is a complex but necessary task in the modern manufacturing industry. It is not easy to use industrial robots to afford this task because different control modes and coordination algorithms have to be designed and updated periodically to adapt the rapidly changing products. It is desired to improve the conventional robot programming technology. The concept of Learning from Demonstration(LfD) is adopted and a LfD framework for small parts assembly problem is proposed in this paper. We categorize the robot control modes into Position Control(PC), Force Control(FC) and Hybrid Control(HC) and formulate the small parts assembly skill learning problem into control mode identification and coordination problem. By recording the human teachers demonstration and learning the control modes and mode switching conditions from the records, the robot is able to imiate the small parts assembly skill autonomously. According to this framework, a LfD method is specifically designed for the semi-structured phone cover assembly platform. The experimental results verify the effectiveness of the proposed algorithm. This method can be implemented in different small part assembly applications, releasing the human workers from the dull and repetitive jobs and improving the productively.

General swing-up methodology for the vertical three-link underactuated manipulator

The underactuated mechanical systems (UMS) are a special class of nonlinear systems with fewer inputs than their degree-of-freedoms (DOF). Currently the research works of such systems mainly focus on the equilibrium point stabilization and periodic trajectory tracking problem. The problem of steering a UMS to arbitrary points in its state space is hardly studied. The existing methods for the swing-up control problem of UMS mainly deal with two DOF UMS and are difficult to be applied in high order DOF UMS. Here in this paper, the problem of swing the UMS to arbitary point is studied. We proposed a virtual constraint based algorithm to generate periodic trajectories that pass through desired point and utilized the Lyapunov based control algorithm to track the generated trajectory. The proposed methods can be applied to a class of UMS and a UMS can be steered from an initial position to the desired position by tracking the generated periodic trajectories. The proposed methods are implemented using a RRR underactuated manipulator and the simulation results demonstrated their performance and verified their effectiveness.