Future satellites equipped with flexible payloads will allow their resources to be allocated in response to the temporal variations of the scenario, such as dynamic traffic demands and undesired interference events. The satellite resources to be configured are the beam pattern, the transmit power and the frequency allocation. At the time of writing, this configuration is expected to be performed by an operator using a graphical tool. That is, an operator will provide the Earth coordinates in order to modify the coverage of the satellite. Regarding the frequency and power allocation, the same idea is followed. Bearing this in mind, in the presence of an event, the operator shall compute manually the best resulting payload configuration
The idea behind the conceived ML model is to assist flexible payloads and eliminate the human intervention in the payload re-configuration. This will reduce the operational expenditures, reduce the time-to-react to system events resulting in an increase of the customer QoS. As a general statement, the resulting ML model is not meant to substitute the operator. Instead, it shall be able to provide a relevant payload configuration able to assist the operator’s job in producing a new payload configuration.
Flexible Payloads Optimization assisted via deep learning
The return link of a multibeam satellite system in the presence of undesired interference is considered here. The satellite generates a geographical footprint in order to attend a total of Nu fixed satellite terminals. Each user terminal generates a traffic request of Ji bits/s for i = 1, …, Nu. Let us denote with hi the gain experienced by the signal transmitted by the i-th user terminal and received by the ground station. This gain accounts for the receiving and transmitting antennas gains and the propagation impairments. For the return link, the satellite has available a set of Nch channels. Additionally, we consider that each user terminal can only be served by a subset of frequency subcarriers (i.e. two different user terminals cannot use the same subcarrier).
The downlink (space to Earth) power transmission of the return link is assumed to be fixed and the same for all subcarriers. The return link transmission takes place in presence of an external interference at a certain geographical location. This interference is assumed to transmit at a fixed power of Pint and its equivalent channel is represented by hi. The interference location is assumed known and the system designer can modify the satellite footprint in order to reject the received interference by a factor µ. This factor is assumed to be -6, -10, -15 and -18 dB.
The assessment of the model for flexible payloads depicted in Figure shows that the genetic algorithm model provides good solutions, while the GA assisted via a DNN is able to obtain better results by using initial solutions from the DNN rather than pure random.
Actually, the results show that, in the event of an intentional interference, a commercial software equipped with the proposed model would able to provide a new bandwidth allocation able to satisfy the user demands in a short time period. Thus, these results also allow us to foresee that the algorithms presented will be of use in satellite missions where other payload models are available.
Related Resources
- SATAI – Machine Learning and Artificial Intelligence for Satellite Communications ARTES FPE 1A.104
- M. Á. Vázquez, P. Henarejos, J. Carlos Gil, I. Pappalardo, A. I. Pérez-Neira, Artificial Intelligence for SatCom Operations , in Proceedings of International Conference on Space Operations (SpaceOps), 18-22 May 2020, Cape Twon (South Africa).
- P. Henarejos, M. Á. Vázquez, A. I. Pérez-Neira, Deep Learning For Experimental Hybrid Terrestrial and Satellite Interference Management , In Proceedings of 20th IEEE International Workshop on Signal Processing Advances in Wireless Communications (SPAWC 2019), 2-5 July 2019, Cannes (France).