JOURNAL OF COMPUTER AND ROBOTICS
Year:2010 | Volume:3 | Issue:1|Papers Count:8

Reconfigurable instruction set processors allow customization for an application domain by extending the core instruction set architecture.Extracting appropriate custom instructions is an important phase for implementing an application on a reconfigurable instruction set processor. A custom instruction (CI) is usually extracted from critical portions of applications and implemented on a reconfigurable functional unit. In this paper, our proposed RFU architecture for a reconfigurable instruction set processor is introduced. As the main contribution of this work, an integrated framework of temporal partitioning and mapping is introduced that partitions and maps CIs on the RFU. Temporal partitioning iterates and modifies partitions incrementally to generate CIs. The proposed framework improves the timing performance particularly for the applications comprising a considerable amount of CIs that could not be implemented on the RFU due to architectural limitations. Furthermore, exploiting similarity detection and merging as two complementary techniques for the integrated framework brings about reduction in the configuration memory size.

Despite the extensive deployment of multi-core architectures in the past few years, the design and optimization of each single processing core is still a fresh field in computing. On the other hand, having a design procedure (used to solve the problems related to the design of a single processing core) makes it possible to apply the proposed solutions to specific-purpose processing cores. The instruction fetch, which is one of the parts of the architectural design, is considered to have the greatest effect on the performance. RISC processors, which have architecture with a high capability for parallelism, need a high instruction width in order to reach an appropriate performance. Accurate branch prediction and low cache miss rate are two effective factors in the operation of the fetching unit. In this paper, we have designed and analyzed the fetching unit for a 4-way (4-issue) superscalar processing core. We have applied the cost per performance design style and quantitative approach to propose this fetch unit. Moreover, timing constrains are specially analyzed for instruction cache to enable the proposed fetch unit to be in a super pipeline system. In order to solve the timing problem, we have applied the division method to the branch prediction tables and the wave pipelining technique to the instruction cache.

Quorum-based power saving (QPS) protocols allow “asynchronous” wireless hosts, operating in a multi-hop ad-hoc network, to tune to the low power mode conceived in IEEE 802.11 MAC standard. QPS schemes guarantee that the wake-up schedule for every two neighboring hosts would ultimately overlap within a bounded latency so as to be able to accomplish their reciprocal “unicast” communications. A major drawback in quorum-based rendezvous schemes, however, lies in the absence of an efficient mechanism for enabling the simultaneous reactivation of all PS neighbors to receive “broadcast” messages. In this paper, a novel asynchronous wake-up scheduling mechanism is proposed, which specifically tackles the broadcast transmission problem in QPS systems. We introduce a special control packet at the MAC layer through which a sending host notifies its neighbors of forthcoming broadcast traffic, allowing the receivers to estimate the approximate re-activation time for ensuring the successful delivery of the messages. We will investigate, analytically, the optimum frequency with which to emit notifications so that the energy overhead induced is minimized in both single-hop broadcasting as well as network-wide flooding scenarios. Evaluation results derived from our simulation experiments reveal that the proposed mechanism can effectively improve the performance of an asynchronous QPS system in terms of both throughput as well as energy saving ratio; for instance, when operating with a wake-up ratio of 16%, network throughput will be enhanced by at least 60% in comparison with the existing schemes.

A wireless sensor network consists of many inexpensive sensor nodes that can be used to confidently extract data from the environment. Nodes are organized into clusters and in each cluster all non-cluster nodes transmit their data only to the cluster-head. The cluster-head transmits all received data to the base station. Because of energy limitation in sensor nodes and energy reduction in each data transmission, appropriate cluster-head election can significantly reduce energy consumption and enhance the life time of the network. In the proposed algorithm, a modified fuzzy logic approach is presented in order to improve the cluster-head election based on four descriptors energy, concentration, centrality, and distance to base station. Cluster-head is elected by the base station in each round by calculating the chance each node has to elect as a cluster-head by considering descriptors. Network life time is evaluated based on first node dies metric, so energy depletion of one node causes the network to die. Simulation shows that the proposed algorithm can effectively increase the network life time. Sensor network is also simulated when sensor nodes move with random velocity in random direction in each round. Simulation shows that network life time is increased by considering this assumption in the proposed algorithm and can develop a better performance.

One of the challenging issues in managing the existing large digital image libraries and databases is Content Based Image Retrieval (CBIR). The accuracy of image retrieval methods in CBIR is subject to effective extraction of image features such as color, texture, and shape. In this paper, we propose a new image retrieval method using contourlet transform coefficients to index texture of the images. We employ the properties of contourlet coefficients to model the distribution of coefficients in each sub-band using the normal distribution function. The assigned normal distribution functions are used effectively at the next stage to extract the texture feature vector. Simulation results indicate that the proposed method outperforms other conventional texture image retrieval methods such as, Gabor filter and wavelet transform. Moreover, this method shows a noticeable higher performance compared to another contourlet based CBIR method.

Traditional leveraging statistical methods for analyzing today’s large volumes of spatial data have high computational burdens. To eliminate the deficiency, relatively modern data mining techniques have been recently applied in different spatial analysis tasks with the purpose of autonomous knowledge extraction from high-volume spatial data. Fortunately, geospatial data is considered a proper subject for leveraging data mining techniques. The main purpose of this paper is presenting a hybrid geospatial data clustering mechanism in order to achieve a high performance hotspot analysis method. The method basically works on 2 or 3-dimensional geographic coordinates of different natural and unnatural phenomena. It uses the systematic cooperation of two popular clustering algorithms: the AGlomerative NEStive, as a hierarchical clustering method and k-means, as a partitional clustering method. It is claimed that the hybrid method will inherit the low time complexity of the κ-means algorithm and also relative independency from user’s knowledge of the AGNES algorithm.Thus, the proposed method is expected to be faster than AGNES algorithm and also more accurate thanκ-means algorithm. Finally, the method was evaluated against two popular clustering measurement criteria. The first clustering evaluation criterion is adapted from Fisher’s separability criterion, and the second one is the popular minimum total distance measure. Results of evaluation reveal that the proposed hybrid method results in an acceptable performance. It has a desirable time complexity and also enjoys a higher cluster quality than its parents (AGNES and k-means). Real-time processing of hotspots requires an efficient approach with low time complexity. So, the problem of time complexity has been taken into account in designing the proposed approach.

Piezoelectric actuators are widely used in micro manipulation applications. However hysteresis nonlinearity limits accuracy of these actuators. This paper presents a novel approach utilizing a piezoelectric nano-stage as slave manipulator of a teleoperation system. The Prandtl-Ishlinskii (PI) model is used to model actuator hysteresis in feed forward scheme to cancel out this nonlinearity. A passive coordination control which uses the new outputs to state synchronize the master and slave robots in free motion is extended to achieve position coordination in contact tasks. The proposed approach uses force feedback using a passivity of the systems and Lyapunov stability methods; the asymptotic stability of force reflecting teleoperation with communication delay and position/force scaling is proven.Performance of the proposed controllers is verified through experiments.

Computation of the second order delay in RC-tree based circuits is important during the design process of modern VLSI systems with respect to having tree structure circuits. Calculation of the second and higher order moments is possible in tree based networks. Because of the closed form solution, computation speed and the ease of using the performance optimization in VLSI design methods such as floor planning, placement and routing, the Elmore delay metric is widely implemented for past generation circuits. However, physical and logical synthesis optimizations require fast and accurate analysis techniques of the RC networks. Elmore first proposed matching circuit moments to a probability density function (PDF), which led to the widespread implementation of it in many networks. But the accuracy of Elmore metric is sometimes unacceptable for the RC interconnect problems in today’s CMOS technologies. The main idea behind our approach is based on the moment matching technique with the power-lognormal distribution and proposing the closed form formula for the delay evaluation of the RC-tree networks. The primary advantages of our approach over the past proposed metrics are the ease of implementation, reduction of the complexity and proposing an efficiency formula without referring to lookup tables. Simulation results confirmed that our method illustrates a good degree of accuracy and the relative average of errors is less than 20%.