Haifeng Zheng

Short-term traffic flow prediction with Conv-LSTM

The accurate short-term traffic flow prediction can provide timely and accurate traffic condition information which can help one to make travel decision and mitigate the traffic jam. Deep learning (DL) provides a new paradigm for the analysis of big data generated by the urban daily traffic. In this paper, we propose a novel end-to-end deep learning architecture which consists of two modules. We combine convolution and LSTM to form a Conv-LSTM module which can extract the spatial-temporal information of the traffic flow information. Furthermore, a Bi-directional LSTM module is also adopted to analyze historical traffic flow data of the prediction point to get the traffic flow periodicity feature. The experimental results on the real dataset show that the proposed approach can achieve a better prediction accuracy compared with the existing approaches.

Short-term traffic flow prediction with optimized Multi-kernel Support Vector Machine

Accurate prediction of the traffic state can help to solve the problem of urban traffic congestion, providing guiding advices for peoples travel and traffic regulation. In this paper, we propose a novel short-term traffic flow prediction algorithm, which is based on Multi-kernel Support Vector Machine (MSVM) and Adaptive Particle Swarm Optimization (APSO). Firstly, we explore both the nonlinear and randomness characteristic of traffic flow, and hybridize Gaussian kernel and polynomial kernel to constitute the MSVM. Secondly, we optimize the parameters of MSVM with a novel APSO algorithm by considering both the historical and real-time traffic data. We evaluate our algorithm by doing thorough experiment on a large real dataset. The results show that our algorithm can do a timely and adaptive prediction even in the rush hour when the traffic conditions change rapidly. At the same time, the prediction results are more accurate compared to four baseline methods.

Spatial-Temporal Data Collection with Compressive Sensing in Mobile Sensor Networks

Compressive sensing (CS) provides an energy-efficient paradigm for data gathering in wireless sensor networks (WSNs). However, the existing work on spatial-temporal data gathering using compressive sensing only considers either multi-hop relaying based or multiple random walks based approaches. In this paper, we exploit the mobility pattern for spatial-temporal data collection and propose a novel mobile data gathering scheme by employing the Metropolis-Hastings algorithm with delayed acceptance, an improved random walk algorithm for a mobile collector to collect data from a sensing field. The proposed scheme exploits Kronecker compressive sensing (KCS) for spatial-temporal correlation of sensory data by allowing the mobile collector to gather temporal compressive measurements from a small subset of randomly selected nodes along a random routing path. More importantly, from the theoretical perspective we prove that the equivalent sensing matrix constructed from the proposed scheme for spatial-temporal compressible signal can satisfy the property of KCS models. The simulation results demonstrate that the proposed scheme can not only significantly reduce communication cost but also improve recovery accuracy for mobile data gathering compared to the other existing schemes. In particular, we also show that the proposed scheme is robust in unreliable wireless environment under various packet losses. All this indicates that the proposed scheme can be an efficient alternative for data gathering application in WSNs.

Design and Analysis of In-Network Computation Protocols With Compressive Sensing in Wireless Sensor Networks

Compressive sensing (CS) provides a new paradigm for correlated data processing and transmission over wireless sensor networks (WSNs). In this paper, we take a new look to investigate the performance of CS for data gathering from the perspective of in-network computation. We formulate the problem of computing random projections for CS in WSNs as in-network function computation in random geometric networks. We focus on the design and performance analysis of the protocols that are efficient in terms of computation complexity. We first propose an efficient tree-based computation protocol with an optimal refresh rate and characterize the scaling laws of energy and latency requirements, which show that it can reduce energy consumption and latency compared with the traditional approach. Then, we present a more efficient block computation protocol by considering the correlations of temporal measurements in function computation to improve the performance in terms of refresh rate and energy consumption. We also devise a gossip-based scheme to improve the robustness of function computation, which is able to distribute computation results to all the nodes throughout the network. We show that the proposed protocol can improve upon the existing gossip-based schemes in terms of energy consumption. Finally, simulation results are also presented to demonstrate the effectiveness of the proposed protocols.

Distributed cell selection in heterogeneous wireless networks

Abstract A critical issue in many wireless networks is how to establish the best possible quality connections between users to base stations. This is in particular challenging when users are randomly located over a geographical region, and each covered by a number of heterogeneous base stations. To this end, we design a smart and efficient cell selection mechanism to improve the user-base station connection in heterogeneous wireless networks. We formulate the cell selection problem as an asymmetric congestion game with consideration of users’ heterogeneity in their locations and their data rates to various cells. We show the existence of pure Nash equilibria (PNE) and propose a concurrent distributed learning algorithm to converge to them. In the algorithm, we allow users to perform random error-tolerant updates synchronously, and guarantee them to reach one or multiple PNE with the largest utilities. In addition, we do a systematical investigation on the implementation of the algorithm in practical networks. Simulation results show that the algorithm can achieve satisfactory performance with acceptable convergence rate.

Traffic estimation in road networks via compressive sensing

The recent advances of compressive sensing (CS) have witnessed a great potential of traffic condition estimation in road networks. In this paper, we propose a traffic estimation approach that applies compressive sensing technique to achieve a city-scale traffic estimation with only a small number of vehicle probes. In particular, we construct a new type of random matrix for CS which can significantly reduce the number of vehicle probes for traffic estimation. Furthermore, we also propose a novel representation matrix to better exploit the correlations of road network to improve the accuracy of traffic estimation. We analyze the incoherence between random measurement matrix and sparsity representation basis. Finally, we validate the effectiveness of the proposed approach through extensive simulations by real-world dataset.

A compressive and adaptive sampling approach in crowdsensing networks

High data quality and low sensing cost are two primary goals in large-scale mobile crowdsensing applications. The oversampling and the undersampling are common problems which always result in a high cost or low data quality that can not satisfy the system requirement. To address this problem, taking into account low-rank latent structure, we propose a compressive and adaptive data sampling scheme (CAS) which exploits adaptivity to identify locations which are highly informative for learning the low-dimensional space of the data matrix. In contrast to existing random sampling methods, it involves a three-pass sampling procedure that firstly assigns a fraction of samples to estimate general information, then samples those more informative locations for exact recovery and finally estimates the values of the unsensing locations. Evaluations on synthetic datasets and real datasets for air quality monitoring show the effectiveness of CAS. The experimental results demonstrate that the proposed scheme is able to not only significantly improve the sensing data quality but also reduce the computation complexity comparing with the state-of-the-art matrix completion methods.