Research Theme

Some of the research themes that will be enabled by the development and deployment of TurboRAN include:

Proactive Mobility Management:
Our overarching goal will be to leverage TurboRAN to trigger a paradigm shift by fundamentally transforming mobility management from being a reactive to a proactive process by developing an Advanced Mobility Management and Utilization Framework (AM-MUF). This project will build on our recent theoretical analysis that indicates that if characterized and predicted, user mobility can be exploited to boost network’s performance. For example, by approximately predicting a user’s next cell e.g. from past cell transitions, handover overhead can be reduced by over 40%. Building on these intriguing theoretical insights, in this project we will use TurboRAN infrastructure to conduct experimental research to answer the following questions:
  • 1) Can mobility prediction be utilized for Proactive (Predictive) Handover in real network?
  • 2) What amount of gain is possible in a real network from Proactive Handover?
  • 3) What are the new spatio-temporal handover margins within which a handover must be conducted in a real UDMN with different user speeds?
  • 4) What changes in RAN and EPC are needed to execute the handovers within the reduced spatio-temporal handover margins?
  • 5) Can mobility prediction be utilized for proactive radio resource reservation in next cell?
  • 6) What amount of gain is possible from proactive radio resource reservation in real network?
  • 7) How is the resource efficiency affected by the prediction accuracy?
  • 8) How does cell cluster level prediction, as opposed to cell level prediction, affect the system performance?
  • 9) What is the tradeoff between reserving resources in ‘N’ potential next cells, and one next cell?
  • 10) Can the mobility prediction be exploited to aid in intra-frequency small cells discoveries?
  • 11) Can cell individual offset adaptation in UDMN be leveraged to improve mobility management and radio resource efficiency?
  • 12) Can introduction of a new parameter, user individual offset in UDMN, while taking into account user mobility intelligence, be used for load balancing and improving quality of experience?
  • Additionally, building on PI’s recent theoretical work on cell discovery optimization, TurboRAN will also be used to develop, validate and optimize novel mmWave cell discovery algorithms.
    Proactive SON Coordination Framework:
    TurboRAN will provide the necessary platform to advance research on self-coordination, thereby removing the bottleneck in SON adaptation. In addition to enabling advancement and validation of recently proposed self-coordination and SON conflict avoidance solutions by PIs and other researchers, TurboRAN, thanks to its end-to-end programmability and real RAN interface, will also enable a new approach towards self- coordination for future P-SON functions. In state of the art SON functions the degree of this dependency is known only qualitatively or not known at all. This leads to parametric, objective or logical conflicts when multiple SON functions operate in a network. TurboRAN can be exploited to identify relationships between different performance indicators and optimization parameters, through an extensive experimentation, data collection and machine learning framework as proposed in detail in PI’s recent vision paper on next generation P-SON.

    Splitting the Control and Data Plane in RAN for Higher Resource Efficiency in UDMN:
    The TurboRAN testbed will be used to not only advance and validate the theoretical CDSA research already initiated by the PI, but we will also run experiments designed to determine answers to following questions with the goal of enabling the CDSA implementation in real networks by 2022:
  • 1) Is it practically viable to split control and data planes as theoretically proposed in various CDSA variants in recent years,
  • 2) Will the frame structures proposed in PI's work for CDSA deliver the projected gains in real implementation?
  • 3) What changes in EPC are required to accommodate CDSA split in RAN?
  • 4) What is the optimal split of measurements and control signaling between small cells while taking into account practical aspects and limitations of hardware and a real network? 5) How do mmWave-based data cells perform when compared to HF-based data cells?
  • 6) Does the optimal split point change with different KPIs namely, capacity, user experience, and energy efficiency?
  • 7) Does the optimal split change with user and traffic distributions in space and time?
  • 8) If yes, is it practically possible to design CDSA architecture that adapts the split of control and user functionalities between DBS and CBS dynamically?
  • 9) If yes, such dynamic splitting must be done centrally or it can be done in distributed fashion?
  • 10) How much capacity can be gained practically with CDSA when compared to conventional cellular networks as well as what are the tradeoffs?
  • 11) How much capacity can be gained with CDSA compared to conventional HetNets, and what are the tradeoffs?
  • 12) How the DBS switching on/off latency effects the energy gains and the QoS?
  • Transforming Database aided Multi-Layer Ultra-Dense Deployments (D-MUD), from concept to a reality:
    D-MUD is a recently proposed architecture for UDMN where certain local data consisting of received signal strength and quality reports, user mobility patterns, spatiotemporal interference footprints, load conditions, Signal to interference and Noise Ratio (SINR) map, and others is stored at small and/or macro cells. The intelligence gathered form this data can then be used to optimize cell selection even when the small cells are asleep and are not transmitting any pilot signals, or to perform load balancing and energy savings by invoking appropriate P-SON functions. D-MUD can be implemented with CDSA or without CDSA. In CDSA, CBS can use dual connectivity and back-haul connection to perform the P-SON functions. With D-MUD, P-SON functions can be executed without relying on additional connectivity between CBS, DBS and UE. While preliminary theoretical analysis of D-MUD conducted by PI and other researchers indicates significant performance gains for D-MUD compared to conventional HetNet architecture, a number of research challenges remain to be addressed before D-MUD can be leveraged to enable the UDMN. These challenges include investigation into the effects of:
  • 1) imperfections in measurements and database;
  • 2) signaling delay;
  • 3) DBS switching on and off delay; and
  • 4) various mobility types.
  • Investigations of all these issues require experimentation based on TurboRAN in addition to theoretical and simulation based analysis already initiated by PIs. Using a testbed is necessary to advance research on D-MUD, primarily because the database of real time measurements must be built, processed, and accessed in real time to obtain meaningful insights into actual D-MUD performance. The TurboRAN with an integrated Hadoop-cluster, will enable such research. Furthermore, our recent theoretical and simulation based analysis of D-MUD shows that D-MUD performance heavily depends on the accuracy of measurements such as user and AP locations that in turn determine accuracy of other measurements such as SINR maps in the database. Research questions that must be experimentally investigated using TurboRAN to transform CDSA and D-MUD from an idea into reality, include:
  • 1) What measurements should be gathered (e.g., RSRP, RSRQ, SNR, SINR and CQI)?
  • 2) What should be the measurement rate for various measurements?
  • 3) What are the optimal volumes for each measurement maps, and how often the maps should be updated in a real network?
  • 4) Do the optimal volumes change with traffic patterns and if yes, how?
  • 5) Does the optimal measurement change for mmWave data cells when compared to HF data cells?
  • 6) How accurately the theoretically calculated ASE in PI’s recent study for D-MUD, with given UE and AP location error, predicts ASE achievable in practical D-MUD?
  • 7) How do inaccuracies in channel estimation affect D-MUD performance?
  • Other Research Areas:
  • 1) User-centric radio resource management (RRM)
  • 2) Advanced interference management schemes
  • 3) Advanced small cells/multi-tier heterogeneous networks
  • 4) Context aware RAN
  • 5) Proactive Self Organizing Networks (P-SON)
  • 6) Operation on Unlicensed band
  • 7) RAN sharing
  • 8) Network function virtualization
  • 9) Movable network components
  • 10) Advance multiple access schemes
  • 11) Device-to-device communication
  • 12) Very low power consumption operation modes (Internet of Things)
  • 13) System level performance evaluation of mmWave cells
  • 14) System level performance evaluation with new waveforms
  • Search