Trajectory Compression With Distributed Compressive Sensing And Linear Simplification

With the rapid development of hardware,a large number of mobile devices,such as vehicle navigator,mobile phones,sports watches,have been equipped with GPS sensors.These devices produce huge amounts of data,which is a heavy burden for transmission,data processing and calculation.Therefore,it is important to apply compression algorithm into trajectory transmission.Trajectory data include uniformly sampled data and non-uniformly sampled data.In this study,we mainly design new compression algorithms from two aspects: the distributed compressed sensing based compression algorithm for uniformly sampled data and the window approximation algorithm of trajectory based on linear simplification for non-uniformly sampled data.For uniformly sampled trajectories,we can decompose them into longitude,latitude and sampling frequency.Compressive sensing based trajectory compression algorithms regard longitude and latitude as two signals and compress them separately.In order to incorporate the correlation between the longitude and the latitude for better compression performance,a Distributed Compressive Approximation of Trajectory(DCAT)based on Distributed Compressive Sensing(DCS)is proposed.The DCS theory utilizes the correlation of containing common component.In addition,based on the linear correlation,we propose a method for training a correlation matrix which aims to decrease the total sparsity.In summary,we utilize the correlation between longitude and latitude by DCS and correlation matrix.Finally,a series of experiments have been conducted to compare with other state-of-the-art compression algorithms on pedestrian datasets and automobile dataset,our model shows significant improvements in the accuracy(more than 10 percent)compared with SQUISH and SimpleTrack.Especially on automobile dataset,the improvement can be 40 to 50 percent.Linear compression algorithms are the traditional research field of trajectory compression,and they can process the non uniformly sampled trajectories.For the improvement of linear compression algorithm in trajectory data receiver,we propose a Window Accumulation Simplification Heuristic algorithm(WASH)by combining the advantages of TD-TR and SQUISHE(λ).We design the algorithm based on window approximation for the consideration that the received data should be compressed and uploaded timely.Inside the window,we use the modified TD-TR to obtain the optimal solution.But compared to the global optimal solution,the algorithm is always far worse as the standard between windows is not uniform.Taking this into account,we use a priority queue and an adaptive error ceiling to reconcile the criteria of each window,so that the maximum error of each window is maintained at a close range.Adaptive error ceiling is obtained by sliding average method.We validate our algorithm on two automobile datasets and compare with OPW-TR,SQUISH-E(λ)and TD-TR.It can be seen that our algorithm outperforms OPW-TR,SQUISH-E(λ)and close to the global optimal algorithm TD-TR.