Multiuser Detection of Collided AIS Packets with Accurate Estimates of Doppler Frequencies

Initially, the automatic identification system (AIS) aims at ship-to-ship or ship-to-coast wireless communications. To enhance AIS, Japan Aerospace Exploration Agency (JAXA) has conducted an experiment named space based AIS experiment (SPAISE), which enables satellites to observe AIS signals that are transmitted from ships across the open ocean and affected by Doppler shift. This experiment has shown that conventional and simple detection schemes for the satellites cannot always operate well, because the wide coverage of the satellites causes frequent collisions of several AIS signal packets. To cope with such a problem, a multiuser detection scheme such as parallel interference cancellation (PIC) has been used for the signal detection. However, the bit error rate (BER) performance is degraded when channel parameters such as Doppler frequencies are estimated by the correlation between the received signals and the known training sequence. To alleviate the degradation, this paper proposes to employ a highly accurate estimation scheme based on the quasi-Newton method for Doppler frequencies that affect the BER performance most seriously. Computer simulations under Rician fading channel conditions demonstrate that PIC employing the proposed method can improve the BER performance drastically.