We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

eess

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Electrical Engineering and Systems Science > Signal Processing

Title: Deep Reinforcement Learning Aided Monte Carlo Tree Search for MIMO Detection

Abstract: This paper proposes a novel multiple-input multiple-output (MIMO) symbol detector that incorporates a deep reinforcement learning (DRL) agent into the Monte Carlo tree search (MCTS) detection algorithm. We first describe how the MCTS algorithm, used in many decision-making problems, is applied to the MIMO detection problem. Then, we introduce a self-designed deep reinforcement learning agent, consisting of a policy value network and a state value network, which is trained to detect MIMO symbols. The outputs of the trained networks are adopted into a modified MCTS detection algorithm to provide useful node statistics and facilitate enhanced tree search process. The resulted scheme, termed the DRL-MCTS detector, demonstrates significant improvements over the original MCTS detection algorithm and exhibits favorable performance compared to other existing linear and DNN-based detection methods under varying channel conditions.
Subjects: Signal Processing (eess.SP); Information Theory (cs.IT); Machine Learning (cs.LG)
Cite as: arXiv:2102.00178 [eess.SP]
  (or arXiv:2102.00178v1 [eess.SP] for this version)

Submission history

From: Ronald Chang [view email]
[v1] Sat, 30 Jan 2021 07:29:04 GMT (761kb,D)

Link back to: arXiv, form interface, contact.