No abstract

Cooperative task offloading emerges a well‐received paradigm for mobile applications that are sensitive to computational power, while dynamic and real‐time characteristics of vehicular networks makes it challenging to guarantee the low delay requirements of vehicular computation offloading. Existing researches cannot satisfy the real‐time computation requests due to the sparse deployment of infrastructure constructions and constrained computing resources of edge servers. Motivated by these, we consider the idea of distributed vehicle‐to‐vehicle task offloading, which makes vehicles act as cooperative nodes to execute tasks. In this paper, we utilize parallel computing of multi‐vehicle cooperation, to provide low‐delay computation services without exceeding the energy constraint. Furthermore, a cooperative vehicles assisted task offloading strategy based on double deep Q‐network is proposed to obtain the optimal task offloading ratio after selecting cooperative vehicles. Simulation results indicate that our proposed strategy can effectively decrease the total system delay. For example, compared with the local execution strategy, the total system delay of the proposed strategy can be reduced by 69.4% on average.

It is a pioneering work to use a Markov chain model to study the pedestrian escape route without visibility. In this paper, based on the Markov chain probability transition matrix, the algorithms with random numbers and the spatial-grid, an escape route in a limited invisible space is obtained. Six pace states (standing, crawling, walking, leaping, jogging, and running) are applied to describe the characteristics of pedestrian behaviors. Besides, eight main direction changes are used to describe the transition characteristic of a pedestrian. At the same time, this paper analyzes the escape route from two views, i.e., pedestrian pace states and directions. The research results show that the Markov chain model is more realistic as a means of studying pedestrian escape routes.

No abstract