Abstract: In existing MWSN(mobile wireless sensor network)， the problem of hot spot allocation was not solved well， and the utilization rate of network was low. It is possible to optimize the network structure by predicting the trajectory of mobile nodes. A trajectory prediction algorithm named MTPA combined with acceleration was proposed. Firstly， modeling the motion state of the node was established. Secondly， a step motion state probability transfer matrix was done. Finally， Markov chain based multi-step probabilistic transfer matrix algorithm was presented. In order to verify the performance of the algorithm， experiments were carried out on the STM32F407 platform， the experimental results show that comparing with traditional uniform prediction algorithms and frequency statistics algorithms， the prediction accuracy of MTPA has certain advantages， and relevant research results can be used in MWSN.

Key words: movement trend, Markov chain, probability matrix, trajectory prediction, embedded