Houxiang Zhang

Intelligent Lighting Control for Vision-Based Robotic Manipulation

The ability of a robot vision system to capture informative images is greatly affected by the condition of lighting in the scene. This paper reveals the importance of active lighting control for robotic manipulation and proposes novel strategies for good visual interpretation of objects in the workspace. Good illumination means that it helps to get images with large signal-to-noise ratio, wide range of linearity, high image contrast, and true color rendering of the objects natural properties. It should also avoid occurrences of highlight and extreme intensity unbalance. If only passive illumination is used, the robot often gets poor images where no appropriate algorithms can be used to extract useful information. A fuzzy controller is further developed to maintain the lighting level suitable for robotic manipulation and guidance in dynamic environments. As carried out in this paper, with both examples of numerical simulations and practical experiments, it promises satisfactory results with the proposed idea of active lighting control.

Three-dimensional point cloud plane segmentation in both structured and unstructured environments

This paper focuses on three-dimensional (3D) point cloud plane segmentation. Two complementary strategies are proposed for different environments, i.e., a subwindow-based region growing (SBRG) algorithm for structured environments, and a hybrid region growing (HRG) algorithm for unstructured environments. The point cloud is decomposed into subwindows first, using the points neighborhood information when they are scanned by the laser range finder (LRF). Then, the subwindows are classified as planar or nonplanar based on their shape. Afterwards, only planar subwindows are employed in the former algorithm, whereas both kinds of subwindows are used in the latter. In the growing phase, planar subwindows are investigated directly (in both algorithms), while each point in nonplanar subwindows is investigated separately (only in HRG). During region growing, plane parameters are computed incrementally when a subwindow or a point is added to the growing region. This incremental methodology makes the plane segmentation fast. The algorithms have been evaluated using real-world datasets from both structured and unstructured environments. Furthermore, they have been benchmarked against a state-of-the-art point-based region growing (PBRG) algorithm with regard to segmentation speed. According to the results, SBRG is 4 and 9 times faster than PBRG when the subwindow size is set to 3x3 and 4x4 respectively; HRG is 4 times faster than PBRG when the subwindow size is set to 4x4. Open-source code for this paper is available at https://github.com/junhaoxiao/TAMS-Planar-Surface-Based-Perception.git.

Fast plane detection for SLAM from noisy range images in both structured and unstructured environments

This paper focuses on fast plane detection in noisy range images. First, two improvements to the state-of-the-art region growing algorithm are presented to make it faster without losing precision for unstructured environments. One is to add the seed selection procedure based on local shape information to avoid blind growth. The other is to simplify the plane fitting mean square error computation complex. Second, a novel algorithm called grid-based region growing is presented for structured environments. The point cloud is divided into small patches based on neighborhood information when it is viewed as a range image. The small patch is called grid. Then the grids are classified into different categories according to their local appearance, including sparse, planar, spherical and linear. Finally, the planar grids are clustered into big patches by region growing. The plane parameters are incrementally computed whenever a new grid is added. The resulting planes can be used for 3D plane simultaneous localization and mapping (SLAM). Experimental results show promising plane detecting speed for both structured and unstructured environments.

Planar Segment Based Three‐dimensional Point Cloud Registration in Outdoor Environments

We present an odometry-free three-dimensional (3D) point cloud registration strategy for outdoor environments based on area attributed planar patches. The approach is split into three steps. The first step is to segment each point cloud into planar segments, utilizing a cached-octree region growing algorithm, which does not require the 2.5D image-like structure of organized point clouds. The second step is to calculate the area of each segment based on small local faces inspired by the idea of surface integrals. The third step is to find segment correspondences between overlapping point clouds using a search algorithm, and compute the transformation from determined correspondences. The transformation is searched globally so as to maximize a spherical correlation-like metric by enumerating solutions derived from potential segment correspondences. The novelty of this step is that only the area and plane parameters of each segment are employed, and no prior pose estimation from other sensors is required. Four datasets have been used to evaluate the proposed approach, three of which are publicly available and one that stems from our custom-built platform. Based on these datasets, the following evaluations have been done: segmentation speed benchmarking, segment area calculation accuracy and speed benchmarking, processing data acquired by scanners with different fields of view, comparison with the iterative closest point algorithm, robustness with respect to occlusions and partial observations, and registration accuracy compared to ground truth. Experimental results confirm that the approach offers an alternative to state-of-the-art algorithms in plane-rich environments.

3D point cloud registration based on planar surfaces

This paper focuses on fast 3D point cloud registration in cluttered urban environments. There are three main steps in the pipeline: Firstly a fast region growing planar segmentation algorithm is employed to extract the planar surfaces. Then the area of each planar patch is calculated using the image-like structure of organized point cloud. In the last step, the registration is defined as a correlation problem, a novel search algorithm which combines heuristic search with pruning using geometry consistency is utilized to find the global optimal solution in a subset of SO(3) ∪ R3, and the transformation is refined using weighted least squares after finding the solution. Since all possible transformations are traversed, no prior pose estimation from other sensors such as odometry or IMU is needed, makeing it robust and can deal with big rotations.

An effective heave compensation and anti-sway control approach for offshore hydraulic crane operations

Offshore hydraulic cranes are difficult to operate safely, accurately and efficiently due to their heavy structure, large inertia, non-intuitive control interface and load sway issues that result from external disturbances. This paper presents an effective heave compensation and anti-sway control approach for offshore crane operations, which is based on robotic arm kinematics and energy dissipation principles. Unlike common operator-based joint-by-joint control procedures, this automated method is more flexible, allowing for more intuitive crane operations and more accurate positioning of the hoisted load. In particular, a unique feature of this approach is that the two control functions of heave compensation and anti-sway are transparently combined and simulated in an integrated modelling environment. The system architecture integrates the control model for crane operations, the hydraulic system model for hydraulics characteristic analysis, the 3D model of the crane to be controlled, the vessel and the environment for visualisation. The propos e d control algorithm and simulation model can be extended to any type of crane model regardless of its configuration or degree of freedom (DOF) without influencing the effectiveness of the method. The hydraulic model is built by using Bond Graph elements and integrated with the control model in the 20-sim simulation environment. The crane operation can be simulated and controlled by the operator using a 3-axis joystick, which provides a transparent user interface. Related simulations were carried out to validate the efficiency and flexibility of the system architecture. As a case study, a 3-joints knuckle boom crane was implemented and tested. The simulation results prove the presented control algorithm for heave compensation and anti-sway to be a valid and efficient solution.

A universal control architecture for maritime cranes and robots using genetic algorithms as a possible mapping approach

This paper introduces a flexible and general control system architecture that allows for modelling, simulation and control of different models of maritime cranes and, more generally, robotic arms by using the same universal input device regardless of their differences in size, kinematic structure, degrees of freedom, body morphology, constraints and affordances. The manipulators that are to be controlled can be added to the system simply by defining the corresponding Denavit-Hartenberg table and their joint limits. The models can be simulated in a 3D visualisation environment, which provides the user with an intuitive visual feedback. The presented architecture represents the base for the research of a flexible mapping procedure between a universal input device and the manipulators to be controlled. As a case study, our first attempt of implementing such a mapping algorithm is also presented. This method is bio-inspired and it is based on the use of Genetic Algorithms (GA). Using this approach, the system is able to automatically learn the inverse kinematic properties of different models. Related simulations were carried out to validate the efficiency of proposed architecture and mapping method.

Analysis and Design of Asymmetric Oscillation for Caterpillar-Like Locomotion

Caterpillar crawling is distinct from that of other limbless animals. It is simple but efficient. This paper presents a novel mechanism to duplicate the movement to a modular caterpillar-like robot. First, how caterpillars move in nature is investigated and analyzed systematically. Two key locomotive properties are abstracted from the body shape of caterpillars during crawling. Then, based on a morphological mapping, a hypothesis of asymmetric oscillation with a ratio of two is proposed, followed by a thorough analysis of the kinematics of the caterpillar-like robot. The asymmetric oscillating mechanism is proved capable of generating stable caterpillar-like locomotion. Next, taking advantage of the two locomotive properties and the hypothesis, a new Central Pattern Generator (CPG) model is designed as the controller of the robot. The model can not only generate the signal as expected, but also provide explicit control parameters for online modulation. Finally, simulation and on-site experiments are carried out. The results confirm that the proposed method is effective for caterpillar-like locomotion.

Flexible Modeling And Simulation Architecture For Haptic Control Of Maritime Cranes And Robotic Arm.

This paper introduces a modular prototyping system architecture that allows for the modeling, simulation and control of different maritime cranes or robotic arms with different kinematic structures and degrees of freedom using the Bond Graph Method. The resulting models are simulated in a virtual environment and controlled using the same input haptic device, which also provides the user with a valuable force feedback. The arm joint angles can be calculated at runtime according to the specific model of the robot to be controlled. The idea is to develop a library of crane beams, joints and actuator models that can be used as modules for simulating different cranes. The base module of this architecture is the crane beam model. Using different joint modules to connect several such models, different crane prototypes can be easily built. The library also includes a simplified model of a vessel to which the crane models can be connected in order to get a complete model. Related simulations were carried out using the so-called 20-sim simulator to validate efficiency and flexibility of the proposed architecture. In particular, a two-beam crane model connected to a simplified vessel model was implemented. To control the arm, an omega.7 from Force Dimension was used as an input haptic device.

Development of adaptive locomotion of a caterpillar-like robot based on a sensory feedback CPG model

This paper presents a novel control mechanism for generating adaptive locomotion of a caterpillar-like robot in complex terrain. Inspired by biological findings in studies of the locomotion of the lamprey, we employ sensory feedback integration for online modulation of the control parameters of a new proposed central pattern generator (CPG). This closed-loop control scheme consists of the following stages: First, touch sensor information is processed and transformed into module states. Then, reactive strategies that determine the mapping between module states and sensory inputs are generated according to an analysis of the module states. Finally, by means of a genetic algorithm, adaptive locomotion is achieved by optimising the amount and speed of sensory input that is fed back to the CPG model. Incorporating the closed-loop controller in a caterpillar-like robot, both simulation and real on-site experiments are carried out. The results confirm the effectiveness of the control system, based on which the robot flexibly adapts to, and manages to crawl across the complex terrain.