Junhyuk Kim

MDP based dynamic base station management for power conservation in self-organizing networks

This paper proposes a Markov decision process (MDP) based base station management scheme that dynamically and collectively manages activation of a group of base stations depending on the time-varying traffic demand for power conservation in self-organization networks. Our MDP model is unique in a sense that it accurately captures the dynamics of handover traffic among neighboring cells, and it formulates infeasible actions as constraints in a constrained optimization problem. Simulation results confirm that the proposed scheme can significantly reduce power consumption: 55% of daily power savings, upto 73% of power savings during low traffic periods, and the minimum 23% of power savings even during high traffic periods. Our MDP algorithms find desired optimal policies to deactivate unnecessary base stations without sacrificing network performances.

Energy efficient basestation operation with traffic-specific energy consumption

Basestation on-off algorithms for mobile-network energy saving beyond 4 G can be simplified due to absence of interference in multi-user MIMO systems. We discuss an on-off algorithm based on specific energy and its energy savings performance.

Traffic off-balancing algorithm for energy efficient networks

Physical layer of high-end network system uses multiple interface arrays. Under the load-balancing perspective, light load can be distributed to multiple interfaces. However, it can cause energy inefficiency in terms of the number of poor utilization interfaces. To tackle this energy inefficiency, traffic off-balancing algorithm for traffic adaptive interface sleep/awake is investigated. As a reference model, 40G/100G Ethernet is investigated. We report that suggested algorithm can achieve energy efficiency while satisfying traffic transmission requirement.

Highly scalable fair contention resolution scheme based on idle time

We propose a new contention resolution scheme designed for large scale autonomous wireless systems and fully distributed D2D (device-to-device) communications networks. The proposed contention resolution scheme is designed for scalability, efficiency, and fairness. Unlike traditional contention resolution schemes, the proposed contention resolution scheme uses idle time of channel as a measure of the channel condition. Idle time of channel is not only a much more reliable and versatile measure of channel condition than the traditional way of collision detection using ACK, it also makes it possible to use the proposed contention resolution scheme in situations where ACK is not expected such as with broadcast and multicast messages or in fully distributed synchronization procedures. The simulation study confirms that the proposed scheme exhibits high degree of scalability, efficiency, and fairness.

Modular IPM strategy for energy conservation in densely deployed networks

We consider a 5G small cell model where the cells are deployed in a way to cover users in the neighbor cells in order to introduce cell active-sleep control, so-called inter-cell power management (IPM). Using a Markov decision process (MDP), one can obtain an optimal policy for IPM, but with limited scalability. This paper proposes a modular IPM strategy which is a suboptimal policy to group only a few neighbor cells. The modular IPM strategy can mitigate the computational complexity of MDP while it achieves a near-optimal solution to save the energy usage in small cells. It is verified to be a feasible solution through the memory usage estimation and simulation study. The modular IPM strategy with fine-step quantized states can save even more power than the non-modular strategy with coarse-step quantized states. Consequently, the proposed modular IPM strategy becomes much favored in the base station management for power saving in a large small-cell cluster.

Adaptive resource provisioning using traffic forecasting for energy efficient networks

A traffic forecast model for a multilane sleep/wake link is investigated where energy consumption can be significantly reduced. A Box-Jenkins ARIMA forecast model is investigated and applied for traffic forecast. The relations among forecasting confidential interval, energy saving performance, buffer size requirement for lossless network are investigated.

Highly Adaptive and scalable random access based on idle-time

New idle-time based random access scheme, Adaptive p-persistent, is proposed for super dense wireless networks. By utilizing idle-time information, Adaptive p-persistent scheme is designed to be not only scalable to support massive number of devices but also adaptive to dynamically changing traffic load; each device updates its persistent level p every time it receives or overhears a new packet based on reference inter-arrival time and persistent level p of overhearing neighbors. We present the simulation results comparing Adaptive p-persistent scheme to the existing random access schemes, BEB, EIED and i-EIED. The simulation study confirms that both idle-time based random access schemes, Adaptive p-persistent and i-EIED, are highly scalable. Furthermore, it verifies that the newly proposed Adaptive p-persistent scheme is exceptionally adaptive. This adaptability of Adaptive p-persistent scheme comes from not only its instantaneous updating rule but also its prompt application to the current transmission.

Efficient Content Replacement in Wireless Content Delivery Network with Cooperative Caching

Wireless content delivery networks (WCDNs) have received attention as a promising solution to reduce the network congestion caused by rapidly growing demands for mobile content. The amount of reduced congestion is intuitively proportional to the hit ratio in a WCDN. Cooperation among cache servers is strongly required to maximize the hit ratio in a WCDN where each cache server is equipped with a small-size cache storage space. In this paper, we address a content replacement problem that deals with how to manage contents in a limited cache storage space in a reactive manner to cope with a dynamic content demand over time. As a new challenge, we apply reinforcement learning, which is Q-learning, to the content replacement problem in a WCDN with coooperative caching. We model the content replacement problem as a Markov Decision Process (MDP) and finally propose an efficient content replacement strategy to maximize the hit ratio based on a multi-agent Q-learning scheme. Simulation results exhibit that the proposed strategy contributes to achieving better content delivery performance in delay due to a higher hit ratio, compared to typical existing schemes of least recently used (LRU) and least frequently used (LFU).

Energy efficient user grouping algorithm for multi-user MIMO systems

A multi-user MIMO (MU-MIMO) system configures orthogonal MIMO channels for active mobile users by use of multiple antennas of the base station (BS). Because different grouping of users can introduce different channel properties in MU-MIMO, user groupinging can minimize BS transmit power in order to satisfy an essential requirement for BS energy savings. In this paper, we discuss how such user groupinging can vary transmit power consumption in a hybrid spatial and time diversity access (SDMA/TDMA) downstream. Then, we proposed an energy efficient user grouping algorithm for applications in MU-MIMO systems which finds a solution close to optimal by swapping users between different groups and comparing the transmit powers. The proposed greedy-swapping algorithm achieves 37% transmit power saving, which is close to the optimal for the exhaustive search, compared to the traditional capacity optimal scheduling algorithm PU2RC with 6% of the computational time for the optimal exhaustive search.