Abstract: For large-scale radio frequency identification ( RFID) indoor positioning system, the positioning scale is relatively large, with less labeled data and more unlabeled data, and it is easily affected by multipath and white noise. An RFID positioning algorithm based on semi-supervised actor-critic co-training (SACC) was proposed to solve this problem. In this research, the positioning is regarded as Markov decision-making process. Firstly, the actor-critic was combined with random actions and selects the unlabeled best received signal arrival intensity (RSSI) data by co-training of the semi-supervised. Secondly, the actor and the critic were updated by employing Kronecker-factored approximation calculate (K-FAC) natural gradient. Finally, the target position was obtained by co-locating with labeled RSSI data and the selected unlabeled RSSI data. The proposed method reduced the cost of indoor positioning significantly by decreasing the number of labeled data. Meanwhile, with the increase of the positioning targets, the actor could quickly select unlabeled RSSI data and updates the location model. Experiment shows that, compared with other RFID indoor positioning algorithms, such as twin delayed deep deterministic policy gradient (TD3), deep deterministic policy gradient (DDPG), and actor-critic using Kronecker-factored trust region ( ACKTR), the proposed method decreased the average positioning error respectively by 50.226%, 41.916%, and 25.004%. Meanwhile, the positioning stability was improved by 23.430%, 28.518%, and 38.631%.

Key words: RFID, RSSI, semi-supervised actor-critic, Kronecker-Factored, co-training