RSS-Based UAV-BS 3-D Mobility Management via Policy Gradient Deep Reinforcement Learning

Abstract

We address the mobility management of an autonomous UAV-mounted base station (UAV-BS) that provides communication services to a cluster of users on the ground while the geographical characteristics (e.g., location and boundary) of the cluster, the geographical locations of the users, and the characteristics of the radio environment are unknown. UAV-BS solely exploits the received signal strengths (RSS) from the users and accordingly chooses its (continuous) 3-D speed to constructively navigate, i.e., improving the transmitted data rate. To compensate for the lack of a model, we adopt policy gradient deep reinforcement learning. As our approach does not rely on any particular information about the users as well as the radio environment, it is flexible and respects the privacy concerns.Our experiments indicate that despite the minimum available information the UAV-BS is able to distinguish between high-rise (often non-line-of-sight dominant) and sub-urban (mainly lineof-sight dominant) environments such that in the former (resp. latter) it tends to reduce (resp. increase) its height and stays close (resp. far) to the cluster. We further observe that the choice of the reward function affects the speed and the ability of the agent to adhere to the problem constraints without affecting the delivered data rate.

Seamless Replacement of UAV-BSs Providing Connectivity to the IoT

“…The mobility management of a UAV-BS providing connectivity services to a cluster of ground users is investigated in [7]. The paper considered the cases where the geographical characteristics of the cluster and the radio environment are unknown.…”

“…Finally, condition (17) incorporates the definition of Y t u . The above problem formulation is not linear, which is due to the objective function (7) that expresses the product of variables (Z l,l ′ ,t vs (1−Y t u ) and Z l,l ′ ,t vt Y t u ). This can be linearized by defining new variables Q l,l ′ ,t u and P l,l ′ ,t u , which will be imposed to equal Z l,l ′ ,t vs (1 − Y t u ) and Z l,l ′ ,t vt Y t u , respectively, by the following set constraints:…”

Future Prospects and Challenges of On-Demand Mobility Management Solutions

“…Similarly, authors in [16] cover the on-demand HO management for various sessions and slices. Further, a policybased MM technique on a per-flow basis has been presented [17]. Furthermore, the SDN and NFV approaches have been applied to the network, CN (correspondence node), and MN (mobile node) [14] [16].…”