Fast implementation of dense stereo vision algorithms on a highly parallel SIMD architecture

In this paper, we present faster than real-time implementation of a class of dense stereo vision algorithms on a low-power massively parallel SIMD architecture, the CSX700. With two cores, each with 96 Processing Elements, this SIMD architecture provides a peak computation power of 96 GFLOPS while consuming only 9 Watts, making it an excellent candidate for embedded computing applications. Exploiting full features of this architecture, we have developed schemes for an efficient parallel implementation with minimum of overhead. For the sum of squared differences (SSD) algorithm and for VGA (640 × 480) images with disparity ranges of 16 and 32, we achieve a performance of 179 and 94 frames per second (fps), respectively. For the HDTV (1,280 × 720) images with disparity ranges of 16 and 32, we achieve a performance of 67 and 35 fps, respectively. We have also implemented more accurate, and hence more computationally expensive variants of the SSD, and for most cases, particularly for VGA images, we have achieved faster than real-time performance. Our results clearly demonstrate that, by developing careful parallelization schemes, the CSX architecture can provide excellent performance and flexibility for various embedded vision applications.     

Design and implementation of a self-guided indoor robot based on a two-tier localization architecture

We consider building an indoor low-cost mobile robot that can be used in home applications. Due to the complicated nature of home environments, it is essential for such a robot to be self-guided in the sense that it is able to determine its current location as well as navigate to locations where it is commanded to. We propose a two-tier architecture to achieve this goal at centimeter-to-meter-level accuracy. The robot can even roam into an area which is new to it. We demonstrate a prototyping system based on an extended iRobot and the results have important implications on intelligent homes.     

Fast Implementation of forward robot kinematics of position withdistributed arithmetic architecture

This paper refers to the fast implementation of the forward kinematic equations of position of robotics manipulators, using a distributed arithmetic-based pipeline architecture. The building blocks of this pipeline architecture are the distributed arithmetic-based circuits that implement the matrix-vector multiplications involved in the calculation of the forward kinematics of position. The matrix-vector multiplications are implemented in the distributed arithmetic technique by using auxiliary binary functions, which are stored in look-up tables. The digit-serial configuration of the proposed implementation is described. The serial and the parallel configurations may result as special extreme cases of the digit-serial configuration     

Fast Operations on Raster Images with SIMD Machine Architectures

Algorithms for fast operations on a raster image data structure are described. The data structure and algorithms have been designed to exploit SIMD parallel architectures. Images of lines, discs and circles can be created efficiently. Images can be translated, scaled, and combined. The data structure is closely related to runlength encodement. The algorithms have been implemented on an ICL DAP on which the operations can be done in real time (on non-trivial images).     

时钟共享多线程处理器SIMD控制器设计与实现

针对图形图像处理器中指令与数据加载以及数据收集的问题,设计和实现了一种时钟共享多线程处理器中的SIMD控制器,完成相关SIMD指令的发送、数据的加载和数据的收集。该控制器以实现高效的数据级并行计算为目标,采用有限状态机实现了前向处理单元、行控制器和列控制器的设计。实验结果表明,所设计的专用硬件电路能够有效提高图形图像处理器处理并行数据的能力。     

Fast parallel Preconditioned Conjugate Gradient algorithms for robot manipulator dynamics simulation

In this paper fast parallel Preconditioned Conjugate Gradient (PCG) algorithms for robot manipulator forward dynamics, or dynamic simulation, problem are presented. By exploiting the inherent structure of the forward dynamics problem, suitable preconditioners are devised to accelerate the iterations. Also, based on the choice of preconditioners, a modified dynamic formulation is used to speedup both serial and parallel computation of each iteration. The implementation of the parallel algorithms on two interconnected processor arrays is discussed and their computation and communication complexities are analyzed. The simulation results for a Puma Arm are presented to illustrate the effectiveness of the proposed preconditioners. With a faster convergence due to preconditioning and a faster computation of iterations due to parallelization, the developed parallel PCG algorithms represent the fastest alternative for parallel computation of the problem withO(n) processors.     

Linearization and optimization of robot dynamics via inertial parameter design

In this article, the concept of linearity number (LN) is introduced to measure the “linearity” of the equations of motion of a serial manipulator. This number is computable in closed-form and is an average quantitative index of the degree of linearity of the robot over a specified region in the joint space. The definition is flexible, allowing the user to create custom-made definitions according to his or her specific needs. Using the concept of LN and the developed computer package CADLOR, one can design the kinematic and/or inertial parameters of the robot such that the robot is completely linear (or as linear as possible) and obtain the corresponding equations of motion in closed-form. Three manipulators are used to illustrate the linearization algorithm. The relation between LN and the performance of a linear controller is demonstrated by a case study in which the first 3 links of a PUMA 560 type robot is considered. © 1996 John Wiley & Sons, Inc.     

Centroidal dynamics of a humanoid robot

The center of mass (CoM) of a humanoid robot occupies a special place in its dynamics. As the location of its effective total mass, and consequently, the point of resultant action of gravity, the CoM is also the point where the robot’s aggregate linear momentum and angular momentum are naturally defined. The overarching purpose of this paper is to refocus our attention to centroidal dynamics: the dynamics of a humanoid robot projected at its CoM. In this paper we specifically study the properties, structure and computation schemes for the centroidal momentum matrix (CMM), which projects the generalized velocities of a humanoid robot to its spatial centroidal momentum. Through a transformation diagram we graphically show the relationship between this matrix and the well-known joint-space inertia matrix. We also introduce the new concept of “average spatial velocity” of the humanoid that encompasses both linear and angular components and results in a novel decomposition of the kinetic energy. Further, we develop a very efficient $O(N)$ O ( N ) algorithm, expressed in a compact form using spatial notation, for computing the CMM, centroidal momentum, centroidal inertia, and average spatial velocity. Finally, as a practical use of centroidal dynamics we show that a momentum-based balance controller that directly employs the CMM can significantly reduce unnecessary trunk bending during balance maintenance against external disturbance.     

A multi-agent architecture with cooperative fuzzy control for a mobile robot

The challenges of robotics have led the researchers to develop control architectures composed of distributed, independent and asynchronous behaviors. One way to approach decentralization is through cooperative control, since it allows the development of complex behavior based on several controllers combined to achieve the desired result. Robots, however, require high-level cognitive capacities, and multi-agent architectures provide the appropriate level of abstraction to define them. This article describes a multi-agent architecture combined with cooperative control developed within the agent. The experiments were carried out on an ActivMedia Pioneer 2DX mobile robot.     

Fast synthesis of the Fredkin gate via quantum Zeno dynamics

We propose a scheme for fast synthesizing the Fredkin gate with rf SQUID qubits. This scheme utilizes the quantum Zeno dynamics induced by continuous couplings and the non-identical couplings between SQUIDs and superconducting cavity. The effects of decoherence on the performance for the gate are analyzed in virtue of master equation and non-unitary evolution with full Hamiltonian. The strictly numerical simulation shows that the fidelity of this Fredkin gate is relatively high corresponding to current typical experimental parameters. Furthermore, an equivalent physical model is also constructed in an array of coupled cavities.     

Study of a distributed control architecture for a quadruped robot

Looking at legged robots, it is sometimes very important to take into account some of the practical aspects (when focusing on theoretical ones) in order to implement control-command levels.In this way, we have treated the problem of the realization of dynamic or quasi-dynamic gaits with a quadruped robot using a new approach from which we have derived an efficient control/command scheme. This is based on a simple consideration which lies in the fact that the Dynamic Model (DM) can be decomposed into two main parts. From our point of view, we consider a part devoted to the command of the legs which could be called a Leg Inverse Dynamic Model (LIDM). We consider a second part dealing with the global characteristics of the platform. At this level, one can control the system. It will be called the LPIM (Leg to Platform Interaction Model).This goal is reached assuming a dichotomy in a distributed architecture and by the way we present it. Further justification of our method will be given in several stages throughout the paper. We paid great attention to time-saving considerations with respect to communication protocols and data exchange at the same level and between the three levels we derived from our basic investigations.     

Animating human locomotion with inverse dynamics

Because the major force components (the internal muscular forces and torques) are not known a priori over time, you cannot use forward dynamics to predict how the human body will walk. The alternative to the apparently intractable problem of specifying the joint torque patterns in advance is to use inverse dynamics to analyze the torques and forces required for the given motion. Such an analysis can show, for example, that the motion induces excessive torque, that the system is out of balance at a certain point, or that the step length is too great. We present a method of using an inverse dynamics computation to dynamically balance the resulting walking motion and to maintain the joint torques within a moderate range imposed by human strength limits. This method corrects or predicts a motion as indicated by the inverse dynamics analysis. Dynamic correctness is a sufficient condition for realistic motion of nonliving objects. In animating a self-actuated system, however, visual realism is another important, separate criterion for determining the success of a technique. Dynamic correctness is not a sufficient condition for this visual realism. An animation of dynamically balanced walking that is also comfortable in the sense of avoiding strength violations can still look quite different from normal human walking. A visually realistic and dynamically sound animation of human locomotion is obtained using an effective combination of kinematic and dynamic techniques     

A tree architecture with hierarchical data processing on a sensor-rich hexapod robot

This paper presents the hardware and software architecture of Golem, a hexapod robot designed as a flexible, scalable, general purpose development and experimenting tool targeted to academia, industry, and defense environments. The system is technologically innovative in its architecture, performance, size and integration, and is industrially promising in its filling the gap between low-performance commercial solutions and costly application-specific proprietary solutions.     

One-step implementation of a Toffoli gate of separated superconducting qubits via quantum Zeno dynamics

Based on the quantum Zeno dynamics, a scheme is presented to implement a Toffoli gate of three separated superconducting qubits (SQs) by one step. Three separated SQs are connected by two resonators. The scheme is insensitive to the resonator decay because the Zeno subspace does not include the state of the resonators being excited. Numerical simulations indicate that the scheme is robust to the fluctuation of the parameters and the Toffoli gate can be implemented with high fidelity.     

Momentum control with hierarchical inverse dynamics on a torque-controlled humanoid

Hierarchical inverse dynamics based on cascades of quadratic programs have been proposed for the control of legged robots. They have important benefits but to the best of our knowledge have never been implemented on a torque controlled humanoid where model inaccuracies, sensor noise and real-time computation requirements can be problematic. Using a reformulation of existing algorithms, we propose a simplification of the problem that allows to achieve real-time control. Momentum-based control is integrated in the task hierarchy and a LQR design approach is used to compute the desired associated closed-loop behavior and improve performance. Extensive experiments on various balancing and tracking tasks show very robust performance in the face of unknown disturbances, even when the humanoid is standing on one foot. Our results demonstrate that hierarchical inverse dynamics together with momentum control can be efficiently used for feedback control under real robot conditions.     

An implementation of backpropagation learning on GF11, a large SIMD parallel computer

Current connectionist simulations require huge computational resources. We describe a neural network simulator for the IBM GF11, an experimental SIMD machine with 566 processors and a peak arithmetic performance of 11 Gigaflops. We present our parallel implementation of the backpropagation learning algorithm, techniques for increasing efficiency performance measurements on the NETTALK text-to-speech benchmark, and a performance model for the simulator. Our simulator currently runs the backpropagation learning algorithm at 900 million connections per second, where each “connection per second” includes both a forward and a backward pass. This figure was obtained on the machine when only 356 processors were working; with all 566 processors operational, our simulator will run at over one billion connections per second. We conclude that the GF11 is well-suited to neural network simulation, and we analyze our use of the machine to determine which features are the most important for high performance.     

Evaluating robustness in a two layer simulated robot architecture

Many two layer robot architectures have been proposed and implemented. While justification for the design can be well argued, how does one know it is really a good idea? In this paper, one describes a two layer architecture (reinforcement learning in the bottom layer and POMDP planning at the top) for a simulated robot and summarize a set of three experiments in which one evaluated the design. To address the many difficulties of evaluating robot architectures, one advocates an experimental approach in which design criteria are elucidated and then form the basis for the evaluation experiments. In our case, one tests the implementation for its reliability and generalization (our design criteria) by comparing our architecture to one in which a key component is substituted; in these experiments, one demonstrates significant performance gains on the design criteria for our architecture.     

WOLF-Linux的五层结构设计与实现

构建基于Linux的嵌入式系统,对可靠性、安全性、功耗等具有较高的要求,因此需要一种模型来清晰完整的描述系统,指导系统的设计与实现.综合嵌入式系统各种要素,结合结构模型和层次思想,提出和分析了嵌入式系统五层结构模型,并将其应用于嵌入式Linux系统--WOLF-Linux的开发中.五层结构模型较好地指导了系统的完成,验证了该模型对于嵌入式系统设计与实现的有效性.     

Practical inverse kinematics of a kinematically redundant robot using a neural network

In solving inverse kinematics problems, traditional methods such as RMRC (resolved motion rate control) and the IKM (inverse kinematic method) are mostly complicated and time-consuming. Using a neural network, however, a practical algorithm for obtaining accurate joint angles in a much shorter time is possible. The neural network approach assumes a transfer function between inputs and outputs and trains the network to satisfy the representative input-output pairs in the least squares sense. First, a test of the appropriateness of the neural network method is performed for the case of a planar two degrees of freedom (DOF) robot. Then the neural network method is employed to find three joint angles of a planar 3-DOF robot maximizing local manipulability. In this algorithm, the proximal redundant joint angle is determined from a neural network and then the remaining joint angles are determined from analytical functions. The results from this method compare favourably with those from the other two traditional methods.