Guaning Chen

Preemptive scheduling of independent jobs on a hypercube

Abstract We consider the problem of scheduling n independent jobs on an m-dimensional hypercube system to minimize the finishing time, where each job Ji is associated with a dimension di and a processing time ti, meaning that Ji requires a di-dimensional subcube for ti units of time. An O(n2) algorithm is presented that decides if all n jobs can be finished by a given deadline T. Using this algorithm, one may obtain a minimum-finishing-time schedule in polynomial time.

Scheduling independent jobs on partitionable hypercubes

We consider the problem of scheduling k independent jobs on an n-dimensional hypercube system to minimize finishing time, where each job Ji is associated with a dimension di and a processing time ti, meaning that Ji requires a di-dimensional subcube for ti units of time. This problem is NP-hard if no preemption is allowed. We propose a simple heuristic called LDLPT (largest dimension longest processing time) and analyze its worst-case performance; the ratio of the heuristic to the optimal finishing time does not exceed 2 − 12n−1.

Energy-constrained bi-objective data muling in underwater wireless sensor networks

For underwater wireless sensor networks (UWSNs), data muling is an effective approach to extending network coverage and lifetime. Sensor data are collected when a mobile data mule travels within the wireless communication range of the sensor. Given the constrained energy available on a data mule and the energy consumption of its communications and movement operations, a data mule may be prevented from visiting every deployed sensor in a tour. We formulate the tour planning of a data mule collecting sensor data in UWSNs as an energy-constrained bi-objective optimization problem termed the Underwater Data Muling Problem (UDMP). UDMP has the two conflicting objectives of minimizing the length of a tour and maximizing the number of sensors contacted, while satisfying the energy constraint of the data mule at all times. We design an approximation algorithm to solve one special case of this NP-hard problem, which computes a set of Pareto-efficient solutions addressing the tradeoff between the two optimization objectives so as to make proper tour planning. Simulation results validate the effectiveness of this algorithm.

Constructing parallel paths between two subcubes

The authors consider a hypercube system that runs more than one job at a time, with each job allocated a subcube. They discuss the problem of migrating (relocating) a job from one subcube to another, assuming a circuit-switching hypercube network. An algorithm is presented for constructing parallel circuits between two subcubes so that the tasks of a job can be migrated simultaneously. It is shown that no matter how fragmented the hypercube is, one can always construct parallel paths between two given subcubes. Furthermore, one can always minimize the maximum length of the constructed circuits. A solution that minimizes the maximum length of the circuits will also minimize the total length. The circuits are mutually edge-disjoint and do not use any edge that has been used by other jobs. The time complexity of the algorithm is O(n/sup 2/m), where n is the dimension of the hypercube system and m is the number of jobs already in the system. >

Scheduling Independent Jobs on Hypercubes

We consider the problem of scheduling k independent jobs on an n-dimensional hypercube system to minimize finishing time, where each job J i is associated with a dimension d i and a processing time t i , meaning that J i requires a d i -dimensional subcube for t i units of time. This problem is NP-complete if no preemption is allowed. We propose a simple heuristic called LDLPT (largest dimension largest processing time) for this problem and analyze its worst-case performance: the ratio of the heuristic finishing time to the optimal does not exceed 2 — 1/2n-1.

An Indoor Collaborative Pedestrian Dead Reckoning System

Indoor localization has become a popular topic in recent years. While self-contained pedestrian dead reckoning (PDR) systems can be conveniently implemented on a smartphone with built-in inertial sensors for indoor localization, the error of the estimated position for a PDR system can accumulate quickly and results in an unacceptable position accuracy. To address this issue, we propose the collaborative pedestrian dead reckoning (CPDR) system. The main idea of the CPDR system is when users are near to each other, we can leverage the proximity information to improve their estimated positions by means of the opportunistic Kalman filter. In addition, the backward correction scheme is used to improve the accuracy of users trajectory. To evaluate the CPDR system, a prototype is implemented on Apples iPhone 5. The experiment results show that the CPDR system achieves a better position accuracy than the raw PDR system.

Energy-efficient mobile targets detection in the presence of mobile sinks

Tracking moving targets has become an increasingly important application for sensor networks. Sensor nodes may sense moving targets far away from the Source, and hence a large amount of energy may be wasted by them to send sensory data to the Source. Designing efficient algorithms and protocols for data dissemination to mobile sinks is an interesting research and engineering issue, especially for large-scale wireless sensor networks (WSNs). Sink mobility brings new challenges to the design of data dissemination. The location updates for each mobile sink need to be continuously propagated through the field to all sensor nodes, so that future data reports can be correctly delivered to the sink. As energy and resources of a sensor node are limited, these algorithms and protocols should meet a high energy efficiency and a high delivery ratio. To deal with this issue, we propose a framework, called Tree Overlay Grid (TOG), for data collection and dissemination. To route queries and deliver data efficiently in our framework, a geometric routing GFB (Greedy Forwarding within Bound) is proposed to create a TTDD-like grid network, and a tree protocol is used to construct local trees around sinks. In addition, two mechanisms are introduced to prolong the network lifetime. The first mechanism tries to save energy by reducing the traffic load; the second one tries to slow down energy consumption by balancing the traffic load. The simulation results show that TOG outperforms the best known data collection solution and some current data collection solutions for WSNs with multiple mobile sinks.

Energy-Efficient Wireless Sensor Networks for Mobile Target Detections

In battlefields, the wireless sensor network can be used for target detection and data collection. However, the mobility of targets as well as the sinks creates challenges for the design and implementation of the system. While there are research proposals capable of handling multiple mobile sinks, these protocols either assume the knowledge of sink location or create too many flooding, leading to a quicker energy consumption. To deal with this issue, we propose a framework for data collection in wireless sensor networks, namely Dynamic and Adaptive Grid (DAG). In our framework, we take the advantages of both grid and tree data structures to route queries and data efficiently. In addition, two mechanisms are introduced to balance the load in the network. The first one helps to distribute the tasks more evenly; and the second one adjust the size of grids to balance the traffic load in each grid. Last, an improved query aggregation is proposed to reduce the query response time and the traffic associated with each query. The simulation results show that DAG outperforms the best known data collection solutions for wireless sensor networks with multiple mobile targets and sinks.

Identifying Relative Vehicle Positions via Electronic and Visual Signals

The availability of relative location of nearby vehicles is critical in providing safety alerts to the drivers and enhancing driving experience. However, most of wireless localization techniques either fail to provide sufficient accuracy to identify the relative vehicle positioning or require expensive hardware to achieve high accuracy. To resolve this issue, in this paper we propose E-V relative vehicle positioning. To effectively pair electronic and visual signals, the E-V matching algorithm is used, which can maximize the probability of correct pairing between an vehicles electronic identity and its visual appearance. To evaluate performance of the E-V relative vehicle positioning, a prototype is built on the Raspberry B+ self-driven car. The conducted experiment results show that E-V relative vehicle positioning system is able to achieve a much better vehicle relative positioning accuracy and the matching result is efficient and stable throughout the experiment.

Simple Floor Plan Construction via Cloudsourcing

Indoor localization has significant growth in recent years. The key requirement to indoor location based-service (LBS) is the availability of floor plans. However, floor plans are difficult to obtain. In this paper, we present a system, Traces-to-Map (T2M), that automatically infers floor plans from pedestrian trajectories collected by smartphones. T2M consists of two parts. First, it takes advantage of collected trajectories to infer a floor plan with wall information. Second, it identifies the region type by considering the standard deviation of moving directions within an area. We validate our method with multiple floor plans. The result shows the T2M system is able to provide a floor plan highly correlated to the real floor plan.