English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

In this paper, we consider jointly optimizing cell load balance and networkthroughput via a reinforcement learning (RL) approach, where inter-cellhandover (i.e., user association assignment) and massive MIMO antenna tiltingare configured as the RL policy to learn. Our rationale behind using RL is tocircumvent the challenges of analytically modeling user mobility and networkdynamics. To accomplish this joint optimization, we integrate vector rewardsinto the RL value network and conduct RL action via a separate policy network.We name this method as Pareto deterministic policy gradients (PDPG). It is anactor-critic, model-free and deterministic policy algorithm which can handlethe coupling objectives with the following two merits: 1) It solves theoptimization via leveraging the degree of freedom of vector reward as opposedto choosing handcrafted scalar-reward; 2) Cross-validation over multiplepolicies can be significantly reduced. Accordingly, the RL enabled networkbehaves in a self-organized way: It learns out the underlying user mobilitythrough measurement history to proactively operate handover and antenna tiltwithout environment assumptions. Our numerical evaluation demonstrates that theintroduced RL method outperforms scalar-reward based approaches. Meanwhile, tobe self-contained, an ideal static optimization based brute-force search solveris included as a benchmark. The comparison shows that the RL approach performsas well as this ideal strategy, though the former one is constrained withlimited environment observations and lower action frequency, whereas the latterones have full access to the user mobility. The convergence of our introducedapproach is also tested under different user mobility environment based on ourmeasurement data from a real scenario.

Pareto Deterministic Policy Gradients and Its Application in 5G Massive MIMO Networks