1 Full Duplex MAC
Existing devices communicate in half-duplex since self-interference.
However, with advancing self-interference cancellation (SIC) technologies, full-duplex radio has become feasible.
Modeling Full-Duplex MAC Design
Full-duplex link can be converted to combination of two types, symmetric link and asymmetric link. Transmitting station and receiving station is same in the symmetric full-duplex link but different in the asymmetric link. Two types have different characteristics when stations communicate and we research the efficient MAC protocol to reflect the characteristics.
2 Massive MIMO Massive Multiple-Input Multiple-Output
MIMO is a method for multiplying the capacity of a radio link.
The latest wireless technologies such as 5G and Wi-Fi 6 supports massive MIMO which uses dozens of antennas.
Channel Estimation with AI
As the number of antennas increases, it's difficult to estimate the channel response of all links in real-time. Since classical methods are difficult to apply due to their high complexity, we are researching deep-learning (DL)-based algorithms to tackle this issue.
- Instead of estimating the channel response of all subcarriers, the channel response of several representative subcarriers (e.g., pilots) is estimated.
- Then infer the response of remaining subcarriers through the DL model.
3 Reinforcement Learning for SON
It has become telco's burden to manually configure parameters of each base station in HetNet.
To tackle this issue, we are studying self-optimization of self-organizing network (SON).
Handover & Mobile Load Balancing Optimization
Handover (HO) and mobile load balancing (MLB) influence each other because they have interrelationship about moving users to other base stations, so they must be optimized in joint form. But it is difficult to get closed-form solution because of a complex network topology. Therefore, we are trying to solve this problem by the deep reinforcement learning which can handle complicated states.