Object tracking is a key component of self-driving systems and has important meanings to alleviate traffic accidents. Therefore, it is meaningful to design a high performance and real-time tracker for improving the stability and safety of self-driving systems. In this paper, an effective and efficient feature fusion tracker, which dynamically fuses gradient and color features to model the appearance of the target object, is designed with the correlation filters framework for fast tracking. To be specific, two complementary correlation filters for gradient (e.g. HOG) and color (e.g. ColorNames) features are maintained during tracking, and the proposed feature fusion method adaptively adjusts the weights of them to deal with large appearance changes of the target object in challenging tracking scenes. The weights are decided by the consistency of the final tracking result and the predicted results obtained by two correlation filters. Moreover, a failure detection scheme is designed to alleviate the model drift issue caused by undesirable model updates to improve the tracking accuracy. If a tracking result is identified as a failed case, re-detection operations are performed to accurately localize the target object. The experimental results prove that the proposed tracker can achieve competitive tracking performance and a satisfactory tracking speed of 25.3 FPS in comparison with several state-of-the-art trackers on challenging tracking benchmarks.

In this paper, a novel spatial-aware tracker (SAT), which utilizes the Siamese network and multiple correlation filters, is proposed to deal with fast motion and model drift problem in visual tracking. Specifically, the Siamese network is first used by an adaptive spatial search strategy to detect the target object in larger search areas. An extended search patch is generated if the target position obtained by the Siamese network is far away from the previous target position. Then, multiple correlation filters perform detection operations on both the extended search patch and the original search patch. With the proposed spatial selection scheme, SAT can accurately track the target object in challenging tracking scenes. By taking advantage of the Siamese network and multiple correlation filters, the proposed SAT tracker can effectively deal with fast motion and model drift problems to achieve better tracking performance. Extensive experimental results demonstrate that the proposed SAT tracker performs superiorly against several state-of-the-art trackers on OTB-2015 tracking benchmark.

The authors proposed in this paper an improved geometric formulation of POE (product of exponential) based kinematic calibration of serial robots. We use both joint offset-free formulation and adjoint transformation errors of joint screws, and apply it to the calibration of an elbow manipulator. Our formulation explains why the original POE calibration always fails with the existence of joint offset errors; the adjoint formulation of joint screw errors eliminates joint screw constraints that was imposed in the original iterated least square calibration algorithm. The second contribution of this paper is the proposal of a modified POE formulation which adopts point measurement data instead of frame measurement data of the end-effector, which can be more realistic and convenient for practical implementation. Simulation results show that the proposed method is plausible and effective. An experiment is under preparation to verify the effectiveness of the proposed calibration method on an elbow manipulator built by Googol Technology.

This paper investigates the synthesis of a spatial stiffness matrix using simple line springs. A new algorithm is developed, which enables the selection of constituent springs based on their positions and directions. The constraining space of the line springs is then investigated. It is shown that an isotropic stiffness matrix, in general, can be split into the sum of two rank-3 stiffness matrices. The three line springs of the first matrix can be selected to pass through any arbitrary points in space, while the three line springs of the second stiffness matrix lie on a quadric surface, which is usually a hyperboloid of one sheet.

Dextrous manipulation is a fundamental problem in the study of multifingered robotic hands. Given a robotic hand and an object to be manipulated by the hand in an environment filled with obstacles, the main objectives of dextrous manipulation are to have the hand grasp the object and transfer it from a start configuration to a goal configuration without collision. To fulfill such a task in general, we will need: (a) a manipulation planner to generate a feasible path for the hand; and (b) a controller to implement the planned path. In this overview paper, we define the manipulation planning problem and present a unified control system architecture for multifingered manipulation (CoSAM/sup 2/). By incorporating the various kinematic and static relationships of a multifingered robotic hand system with proper sensory data inputs at different stages, CoSAM/sup 2/ achieves the various objectives of dextrous manipulation. Theoretical background of the control system design along with real-time experimental results are described.

Investigates the stiffness and kinematics of an object grasped by multiple elastic fingers. The derivation is based on the work of Svinin (1995), but derived with the approach of minimization of energy and using Montana's (1989) equations of contact model. The result gives a closed form expression of object stiffness in terms of contact finger stiffness, contact locations and contact point curvatures.

Dextrous manipulation is an important issue in the study of multifingered robotic hands. Determining the contact velocities in dextrous manipulation with rolling contact is the key problem. We propose a coordinated manipulation scheme in which the contact velocities are determined by tactile sensor feedback. We address coordinated motion generation which maintains or improves the grasp quality in dextrous manipulation. We implement several experiments using the HKUST three-fingered robotic hand. The experimental results illustrate the effectiveness of the proposed scheme.

Capacitive tactile sensors are constructed and installed to the fingers of the HKUST hands for measurement of position, force and direction of principle curvature of contact point. The hardware and software for signal processing are designed such that the contact information is sent to the motion control computer in real time. Experiments in rolling and sliding contact motions are then performed for testing the functionality of the tactile sensing system in motion control. The measurement of contact velocities obtained from the sensor is also compared with that calculated from the theoretical contact equations. This paper describes the tactile sensing system and the experimental result in contact motion control.

In this paper, we propose a unified control system architecture for multifingered manipulation (CoSAM/sup 2/). CoSAM/sup 2/ achieves simultaneously three objectives of multifingered manipulation: (a) Motion trajectory (velocity/force) tracking of a grasped object; (b) Improving the grasp configuration in the course of object manipulation; and (c) Optimizing grasping forces to enforce contact constraint and compensate for external object wrenches. CoSAM/sup 2/ is organized in a modular and hierarchic structure so that each module implements a specified function using inputs from its predecessors and a minimum number of sensory data signals. CoSAM/sup 2/ is also flexible in accommodating addition of new modules. Here, we give the details for the coordinated motion generation module and the grasping force generation module.