This research project proposal examines connecting self-organization use cases in future radio systems. The candidate proposes to apply cooperation and conflict resolution mechanisms to derive methods and algorithms for connections between different self-organization use cases that operate on the same radio network and affect the same parameters. Allowing use cases to interact appropriately can help optimize metrics and resolve conflicts, improving network performance and reducing costs. The interactions between use cases are complex and coordinating their simultaneous operation requires establishing effective mechanisms.
Copyright:
Attribution Non-Commercial (BY-NC)
Available Formats
Download as DOCX, PDF, TXT or read online from Scribd
This research project proposal examines connecting self-organization use cases in future radio systems. The candidate proposes to apply cooperation and conflict resolution mechanisms to derive methods and algorithms for connections between different self-organization use cases that operate on the same radio network and affect the same parameters. Allowing use cases to interact appropriately can help optimize metrics and resolve conflicts, improving network performance and reducing costs. The interactions between use cases are complex and coordinating their simultaneous operation requires establishing effective mechanisms.
This research project proposal examines connecting self-organization use cases in future radio systems. The candidate proposes to apply cooperation and conflict resolution mechanisms to derive methods and algorithms for connections between different self-organization use cases that operate on the same radio network and affect the same parameters. Allowing use cases to interact appropriately can help optimize metrics and resolve conflicts, improving network performance and reducing costs. The interactions between use cases are complex and coordinating their simultaneous operation requires establishing effective mechanisms.
Copyright:
Attribution Non-Commercial (BY-NC)
Available Formats
Download as DOCX, PDF, TXT or read online from Scribd
This research project proposal examines connecting self-organization use cases in future radio systems. The candidate proposes to apply cooperation and conflict resolution mechanisms to derive methods and algorithms for connections between different self-organization use cases that operate on the same radio network and affect the same parameters. Allowing use cases to interact appropriately can help optimize metrics and resolve conflicts, improving network performance and reducing costs. The interactions between use cases are complex and coordinating their simultaneous operation requires establishing effective mechanisms.
Copyright:
Attribution Non-Commercial (BY-NC)
Available Formats
Download as DOCX, PDF, TXT or read online from Scribd
Download as docx, pdf, or txt
You are on page 1of 20
Research Project Proposal
For PhD Thesis
Title: Connecting Self-Organisation Use Cases in Future Radio Systems.
Candidate: Stephen S. Mwanje
Supervisor(s): Prof. Dr. -Ing. Habil. Andreas Mitschele-Thiel
V2.0 Nov 2011
l
ABSTRACT ireless communication networks involve conIiguration, optimization and management oI a large number oI parameters in order to guarantee the network`s degree oI availability, reliability and quality oI the services provided. It is expected that Iuture networks will be SelI-Organizing in all these Iunctional areas as has been proposed Ior LTE. A number oI SelI-Organization Networks (SON) Use Cases (UCs) have been proposed and their methods studied Ior diIIerent radio network Iunctionalities where SelI Organization can be applied. During the parallel operation oI any number oI these SON methods, interrelationships may be realized in the operation oI the UCs either as requirements Ior interaction or as conIlict situations. This is especially so because the UCs operate on the same radio network, adjusting the same set oI network parameters and together aiming to achieve the same global objectives (measure by a set oI metrics). Interactions are realized where one UC triggers another or where multiple UCs cooperate (work together) to optimize a particular metric. ConIlicts on the other hand, result Irom 2 UCs adjusting the same parameters or when a parameter adjusted by one UC aIIects a metric oI another UC. All these connectivity scenarios are bound to exist in any single SON environment. It is then necessary, to establish the required mechanism(s) Ior these connections and Ior the simultaneous operation. e envisage application oI cooperation and conIlict resolution mechanisms to derive methods and algorithms Ior the connections. Such mechanisms may include hierarchical ordering, consensus, game theoretic approaches or methods to decouple UCs or sub SONs based on the UC characteristics like time oI run, Irequency oI execution, or breadth oI eIIect oI parameter changes. This work is part oI the Graduate School oI Mobile Communications` and is related to other work on SelI Organization oI Coverage and Capacity in Iuture Radio access networks being undertaken by Muhammad Naseer ul Islam, Nauman Zia and Elke Roth-Mandutz.
1
1. INTRODUCTION 1.1. Background Cellular systems have had exceptional growth in the last two decades driven initially by demand Ior mobility oI speech users but recently Ior mobile data applications like web 2.0, mobile video on demand, and mobile oIIice. This was matched by development oI better technologies to deliver the desired services. The most recent oI these technologies is LTE, expected to deliver up to 100Mbps in a 20MHz downlink spectrum at speeds oI up to 15Km/hr |1|. ith expected optimal cell sizes much smaller compared to earlier systems (e.g. 5Km Ior LTE |1|), Iuture radio systems will require many more base stations (eNodeBs, eNBs) compared to 2/3G systems. This translates into high Capital Expenditure (Capex), high Operational Expenditure (Opex) and challenges in ensuring optimum radio network perIormance. It has been identiIied that potential Capex and Opex reductions and signiIicant perIormance improvement can be made through SelI-Organisation (SO) oI the radio systems |2| |3|. 1.2. Motivation and Context In justiIying the need Ior SO, the Next Generation Mobile Networks (NGMN) consortium, a cooperation oI mobile network operators worldwide presented the operators` SON requirements to the standardisation bodies |4|. In |5|, in working towards inclusion oI SON Iunctionality in the standardisation oI LTE, 3GPP generated an inIormative list oI UCs which could be considered priority Ior development. The Socrates Project, a European Union (EU) FP7 Project undertaken by oI major research institutions and equipment vendors across Europe, leaped above these and other previous projects like Monotas |6| and GandalI |7|, to identiIy and describe 25 UCs related to the LTE air interIace Ialling in either one oI the three network Iunctional areas oI conIiguration, optimisation and healing |8|, |9|. Socrates also deIined the criteria Ior evaluating SelI organization methods |10| and identiIied the need Ior integrating these methods. 1.3. Statement and Significance of the Problem The UCs identiIied above operate on the same radio network, adjusting the same set oI network parameters and together aiming to achieve the same global targets / beneIits that are measure by a Iixed set oI metrics. During the parallel operation oI any number oI these SON methods, 2
interrelationships may be realized in the operation oI the UCs either as requirements Ior interaction or as conIlict situations. Interactions are realized where one UC triggers another or where multiple UCs cooperate to optimize a particular metric. ConIlicts on the other hand, result Irom 2 UCs adjusting the same parameters or when a parameter adjusted by one UC aIIects a metric oI another UC. Consequently, it is necessary to establish the required mechanism(s) Ior these connections and Ior the simultaneous operation. To demonstrate the signiIicance oI the problem, consider the operation oI two SelI Optimization UCs handovers (SOHO) and Admission Control (SOAC) in two neighboring cells. Consider a User Equipment (UE) at the edge oI cell A moving towards cell B, where Cell A continually optimizes its Handover (HO) parameters to optimise HO perIormance. As the UE nears the edge oI cell A, HO is initiated to cell B, only to Iind that, as a mechanism to maintain particular desired QoS oI new and on-going calls, cell B has adjusted its parameters in order to control admission oI new entrants call setups or HO Ior a set time period t d . The likely result here is that HO to B will not be perIormed and the UE will stay connected to A although in B`s Iootprint. This may eventually lead to a call drop. Alternatively the UE may be pre-emptively admitted into B but later handed over back to cell A resulting into a ping-pong HO. To solve the problem the two use cases would have to cooperate during the selI-optimization process in order to minimize both HO call drops and ping pong HOs. A optimizes HO parameters to determine that it is initiated at d Km Irom eNB B optimizes AC and decided to reject new sessions Ior a speciIic time t d ~4 in order to maintain QoS Ior ongoing sessions UE reaches dKm Irom A A initiates handover to B
B rejects handover since td has not expired as set in previous admission control optimization UE stays on A although physically within B`s coverage
Dragged session breaks due to bad signal or another handover is initiated resulting in ping-pong handover Figure 1: Interaction of Handover and Admission Control across two cells The interactions presented here can be complicated Iurther iI other SO UCs are running. Consider the case where a load balancing UC (SOLB) is running to adjust HO parameters as a means oI balancing load between the two cells. The SOLB may decide to send more traIIic onto 8 A 8 A 10
L0
L1
L2 3
cell A to relieve B while SOHO on A is adjusting the same HO parameters in order to optimise handover perIormance, which may inherently send traIIic to B. These will deIinitely need to be coordinated. Another case is a congestion control UC (SOCC) running to reduce congestion on B alongside the SOLB. Running both concurrently may result into excessive load reduction on B while over loading A. Ior example the SOLB may decide to shrink B`s cell size to send more traIIic on A. concurrently, the SOCC may decide to drop some oI the connected radio bearers in order to reduce load. The results will be a large sudden reduction in B`s load with a disproportionate load transIer to A. The connectivity among the Iour UCs can be summarised by Iigure 2. It is then necessary to coordinate and resolve these interactions among such related UCs. The same is also required among all other interactions as well as these UCs against the rest.
Figure 2: Possible Connections among Handover, Admission Control, Congestion Control and Load Balancing SO UCs
CongesLlon ConLrol Admlsslon ConLrol Load 8alanclng Pandover CongesLlon ConLrol Pandover CooperaLlon L8 ln cell A may Lrlgger CC ln cell 8
SC ln Cell A SC ln Cell 8 8ad CoS PC arameLer ConfllcLs 1rlgger CooperaLlon arameLer Ad[usLmenL
2. REVIEW OF THE LITERATURE 2.1. SO Use Cases Introduction Most oI the work on SON in LTE has been on three major Ironts -the operators` perspective oI the beneIit oI SONs by NGNM |4|, the standardization oI SON in LTE by 3GPP |5| and research on the requirements, nature and implementation oI SONs by SOCRATES |8| |11| |12| |13|. The 18 UCs presented by NGNM covered the pre-operational and operational states oI the network distributed into 4 categories oI network planning, deployment, optimization and maintenance |4|. Other related work was undertaken in the Monotas project that studied 'Mobile Network Optimisation Through Advanced Simulation |6| and by the GandalI` project that deIined solutions Ior selI-diagnosis, selI-testing and selI-tuning in multi-system environment |7|. In an eIIort to standardize the development and operation oI SON Iunctionality especially Ior a multivendor environment, 3GPP has described the necessary measurements, procedures and open interIaces to support the operation oI nine (9) SON UCs. Covered within the 3GPP TR 36.902 standard, the 9 are so selected in consideration oI the manual eIIort especially in the early deployment phase that would be required to set up and optimize the network to achieve a stable system setup. For each UC, the standard describes the required Iunctionality, evaluation scenarios and expected results, the solution description and where applicable the O&M requirements Ior radio related Iunctions. Development of SON Methods / Solutions In studying towards a common Iramework Ior development oI required algorithms Ior the UCs in the SOCRATES project, all the 25 UCs were Iully described in a uniIorm structure that deIined all the parameters that relate to the operation oI the individual UCs |5|. Each UC is deIined in terms oI a metric that represents a speciIic network physical state or operational behaviour that ought to be maximized; a trigger event or situation that initiates the execution oI a particular SON UC method; a set oI parameters that can potentially be adjusted as well as a set oI input inIormation sources which will be used in order to determine which and to what values parameters should be adjusted |9|. The project partners also attempted to priorities the UCs (table 1) in order to determine the most important ones that could be developed Iirst |11|.
Table 1: Overview oI Socrates Use Case Prioritization per partner |11| USE CASE Over all Rank SelI-Healing: Cell outage compensation 1 SelI-Optimization: Coverage hole detection 2 SelI-Optimization: Home eNodeB 3 SelI-Optimization: Load Balancing 4 SelI-Healing: Cell outage detection 5 SelI-Optimization: InterIerence Coordination 6 SelI-Optimization: Management oI relays and repeaters 7 SelI-Optimization: Packet Scheduling 8 SelI-Optimization: Handover 9 SelI-Optimization: Admission Control 10 SelI-Optimization: Reduction oI energy consumption 11 SelI-Optimization: Congestion Control 12 SelI- ConIiguration: Intelligently selecting site locations 13 SelI-Healing: Cell outage prediction 14 SelI-ConIiguration: Automatic generation oI deIault parameters 15 SelI-Optimization: Physical Channels 16 SelI-Optimization: RACH Optimization 17 SelI-Optimization: MIMO 18 SelI-Optimization: Link Level Retransmission Scheme 19 SelI-Optimization: Spectrum Sharing 20 SelI-Optimization: Neighbour Cell List 21 SelI-Optimization: Tracking Areas 22 SelI-ConIiguration: Network Authentication 23 SelI-ConIiguration: Hardware/capacity extension 24 SelI-Optimization: TDD UL/DL Switching Point 25 Development status and Results ithin SOCRATES, a Iramework has been studied to deIine the guidelines Ior development oI the SO algorithms and within this Iramework, a high level study oI the interaction and dependencies between use cases was undertaken |4|. Individually, some UCs have been studied to varying depths and with varying results, with the most Iamous being Handover Optimization |14|, |15|, |16| |17|, Admission Control Optimization |18|, |17| and Load balancing |19| |20|. In |14| an algorithm that selects the best hysteresis and time to trigger combinations Ior Handover Optimization is simulated in a speciIic scenario with better results when compared to the same system and scenario but with static settings. In |15|, the authors study the perIormance oI a selI-organising algorithm Ior handover parameters optimisation against two reIerence cases one with no protection against oscillations i.e. where UEs are connected to the best server at all times and another with strong protection. They observe that application oI a SO algorithm improves system perIormance in both DL and UL throughput and yet concurrently mitigates oscillations at levels similar to the strong protection case.
Similarly, a call admission control SON algorithm is simulated and Iound to comply better to a deIined policy, than the static algorithm with Iixed HO Threshold (Th HO ) |18|. Conversely, system level simulation results are presented in |19| based on a LB algorithm that evaluates the load condition in a given cell and its neighbouring cells and estimates the impact oI changing the HO parameters in order to improve the overall network perIormance. It is observed that a gain in terms oI increase oI average number oI satisIied users is possible albeit dependent on the load situation in a cell and the available capacity in neighbouring cells. In another study, an algorithm was presented, which tunes the RACH power control parameters and simulations ran proved that RACH selI-tuning is indeed possible given that UE assisted measurements are available Ior the selI-tuning mechanism |21|. Other Iunctions have also been studied including Packet Scheduling Parameter Optimisation, Load Balancing, SelI-Optimisation oI Home eNodeBs, Cell Outage Management, X-Map Estimation, and Automatic Generation oI Initial Parameters Ior eNodeB Insertion |22| 2.2. The Connectivity challenge UC Connections Individually, the UCs in 2.1 optimise only a narrow Iield. Globally, however, all the UCs must be operated on the same radio network to achieve the desired goals and in many cases, many UCs must be simultaneously operated. This will necessitate interaction among them and may in some cases result into conIlicts. It is then necessary, to establish the required mechanism(s) Ior these connections and Ior the simultaneous operation. Four major connectivity scenarios have been identiIied and can be categorisations under two groups either as Interactive or ConIlict type connections. %rigger interactions: In this case, a UC Iails to solve a trigger situation and should call in another UC to take action. The trigger event oI one UC to another must be managed. Cooperation interactions: In this case, a single UC cannot on its own solve a trigger situation and must exchange inIormation with another UC during execution to concurrently adjust their parameters in order to achieve a certain metric level. Parameter conflicts: To achieve any one metric with a given UC method, a number oI parameters have to be tuned. The same parameters however may be tuned by a diIIerent method
to achieve a diIIerent metric. An example here is the adjustment oI Handover parameters by HO optimisation and Load balancing UCs Observability conflicts: A UC optimises its own parameters but inadvertently aIIects a metric oI another UC through adjustment oI at least one parameter that may not actually be shared between the two UCs. A mechanism is necessary to ensure that actions oI the two UCs are coordinated Ior good overall network perIormance. Other relationships may exit but would not require integration activity during SON development. An example is having two UCs that share inputs. In such case there is no need Ior integration unless the two UCs also exhibit one oI the 4 connections above. Benefits and preliminary results of SON UC integrations In demonstrating the need Ior integration and potential beneIits, some connected UCs have been simulated together. In one such case, Simulation results show that, compared to the case without the selI-optimisation oI HO parameters, the AC parameter optimisation algorithm considerably improves the HO perIormance by reducing the amount oI calls that are dropped prior to or during HO |17|. There is however little negative interaction between the two algorithms |22|. In another study, interaction between a HO SO algorithm and a LB SO algorithm is simulated with a coordinator that controls the two base algorithms. Results show that the coordinator is able to control the two algorithms to reasonable system perIormance and in some cases to better perIormance than any oI two separately by combining the strengths oI the two algorithms |20|. To consider the eIIect oI HO optimisation between macro and Home eNBs, simulation results do not show any perIormance improvement Ior the scenario oI a simpliIied, trend based, macro HO optimisation algorithm |22|. 2.3. Concerns for UC integration Following successIul development oI standalone UCs, the integration must address the major concerns relating to their concurrent operation including the Iollowing: Dependencies among UCs: The Iour observed connections can be realised among any set oI UCs at diIIerent times depending on the network trigger situation. The triggers are any oI the network perIormance
situations including high or low number oI dropped or blocked calls; high, low or imbalanced QoS or traIIic load; high or low cell capacity; a need Ior a new site; or the presence oI a coverage hole or a cell outage. Combined simulation: A number oI simulation tools have been developed in the development oI solutions Ior standalone UCs. It is then necessary to devise means to integrate the simulation environment Ior the integration studies. A generic discussion oI possible alternatives has been presented in |13| where it is noted that the Iinal choice oI approach will depend on a particular grouping oI UCs in terms oI the speciIics oI the combined SON Iunctions; control parameters and key measurables; possible diIIerences in operational time scales oI the integrated SON Iunctions, and the suitability oI the available simulation tools Ior consideration oI integrated SON Iunctions. Architecture: Depending on the extent oI their inIluence in terms oI number oI cells they aIIect, individual UCs have diIIerent preIerred architectural designs which could be centralised, distributed or hybrids oI the other two. hen combined however, the architectural demands may change depending on the other UCs they interconnect to as well as their degree oI interconnectivity. A demonstration oI the degree oI connectivity is given in Iigure 3 which shows the connectivity between UCs and parameters Ior the 15 top most UCs as prioritised by Socrates in |11|.
Figure 3: Expected connectivity among top 15 UCs (adapted from |22|) -elghb LlsL Po Pys PC CffseL 111 uL 1x ower upllnk C AnLenna 1llL uL 1x o/88 8eam lorm Sched aram Admlss 1hr PC CpLl m Load bal Pe-8 PC CpL CC A-8 uC aram CCu CPu CCC AC Pe-8 Cov/lnL lnLef Coord ackeL Sched Admlss CLr Lgy Cons Cong CLr
Stability: In a single complex system, there are likely to be many causes oI instability ranging Irom single parameters reaching their maximum values to ping pong eIIects between pair(s) oI UCs. These need to be determined and solutions devised Ior them. Integration of Operator policies It may be the case that human eIIort is still needed iI not to deIine the requirements to be IulIilled by Son Iunctions but also to deIine the compromise Iunctions between competing objectives oI SON methods |13|. Assessment Methods and metrics: It is important on integration oI multiple UCs, to consider what the most appropriate assessment method and metrics Ior such an integrated case would be. In |13| a generic discussion is given on the possible alternatives oI deIining the assessment metrics Ior the integrated UC as well as an example assessment method based on distributed coordination Iunctions 2.4. State of the Art Solutions Functional parameter groups In |11|, |23| an idea was presented to classiIy diIIerent radio parameters into groups - called Iunctional parameters groups created in a way that parameters in any one group contribute to the satisIaction oI the same goal(s). It was expected that parameters in one group would not be coupled with the goals oI other groups, which would allow Ior the identiIication oI UCs that can be developed in parallel and that would not need to be coordinated and / or simulated together. However, it was Iound that a majority oI the parameters Iell under the same Iunctional parameter group which would lead to impracticable results, as the majority oI the algorithms would need to be simulated together. An alternative attempt to reduce the complexity considered the relation between the parameters by introducing a metric interrelation weight (I) such that high Is would suggest that the associated parameter pair be jointly investigated and optimised. Control and Coordination |11|, |23| suggest handling conIlicts by introducing a control plane which compares the measurements against operator deIined thresholds and decides on the activation oI triggers 10
and a coordination plane which processes the parameter changes proposed by concurrent Iunctionalities beIore real network parameter adjustments are executed. The authors speciIy the requirements Ior the control and coordination planes as well as the challenges including identiIication oI possible conIlicts. They also speciIy the Iunctions that may be included in the coordinator to include autognostics, super operator policy Iunctionality, guard Iunctionality against extreme behaviour due to SON, arbitration and SON parameter execution, but leave the details oI the speciIication and implementation oI coordination Iunctions to Iuture work. In |24| an experimental system that realizes SON Iunction coordination based on fexible operator policy-based decision coordination developed in |25| is presented. Coverage and Capacity Optimization (CCO) is used as use case to demonstrate successIul coordination oI multiple independent SON Iunctions Ior which coordination decisions are built into policies on a decision tree. A case study Ior a coordination Iunction has also been implemented in |20| |22| to evaluate the Integration oI Handover Optimisation and Load Balancing. The study implements an alignment Iunction that prevents handover optimisation Irom adjusting the hysteresis oI a cell that has previously been oIten overloaded. The SON coordination combines the beneIits oI both algorithms to signiIicantly improve the call drop ratio and not so much the HO Iailure ratio although with slight increases in number oI unsatisIied customers and HO ping pong ratio. 2.5. Open Issues and Conclusion The state oI the art has presented ideas that have only been partially validated. The degree oI dependence among the SON UCs has also not been Iully described although has been noted to most likely be vendor dependent depending on the vendors` implementation oI the SON methods. The proposed coordination Iunction has only been partially justiIied and needs to be Iully validated including validation oI its proposed Iunctions. This project will contribute to study, development and validation oI the proposed as well as new solutions and ideas to implement SON integration and where possible demonstrate how the UCs can be de-coupled.
11
3. METHODOLOGY 3.1. Research Plan This project will seek to answer how algorithms can be designed, implemented and evaluated Ior interactive use cases in selI organized networks with consideration oI the LTE radio system as an example. SpeciIically, it will study, design and simulate the algorithms Ior a Iinite set oI interacting selI-organization use cases which will be determined in the process. ithin the project, the work will be structured as Iollows: A) An analysis oI use cases, interactions and algorithms which will involve O Study the requirements Ior the optimization oI individual use cases and the results so Iar available oI individual algorithm developments. O Study the nature oI the interactions among the use cases and the Irameworks Ior evaluating the interactions. O Study the characteristics oI individual Use Cases and their implications towards interactivity and conIlict management Ior the Use cases O Study the state oI the art oI proposed solutions as well other possible solutions that have been applied in other systems computational, biological or otherwise. This will include a consideration oI the limitations oI the diIIerent strategies and their beneIits B) Algorithm design and development O Propose new and improvements in solutions Ior observed interactions and scenarios; to possible designs and implementation strategies Ior the interactive algorithms and comparison oI the useIulness oI diIIerent strategies Ior the implementation oI the interactions and / or conIlict resolution strategies in the SON use cases. O Create a Iramework Ior designing algorithms Ior the selected integrated SON use cases, including the guidelines Ior evaluating the designed algorithms. O Design and evaluate the algorithms Ior the use cases where necessary Ior UCs individually and Ior interactive UCs together. C) Demonstration and Simulations 12
O Create a Iramework Ior simulating algorithms individually and together, including the guidelines Ior evaluating the simulations and their results. Critical here will be the shared metrics Ior achieving individual and global objectives oI the multiple use cases. O Simulate algorithms and analyze the results both individual Use Case as well as Ior a set oI interactive Use Cases simulated together. 3.2. Proposed Solution Ideas Hybrid Architecture The best solution lies in implementing distributed algorithms Ior individual UCs. These then have to be coordinated in a cell and eNB to achieve global network objectives. A controller` also evaluates shared metrics aIIected by non-shared parameters to ensure appropriateness oI solution(s) to network Iunctionality. The structure oI the hybrid systems may be such that the individual UC and the single cell multi UC coordination and control Iunctionalities are Iully distributed while the multi eNB coordination is Iully centralised. To counter the eIIects oI centralisation in a multivendor environment, the single vendor multi eNB coordination may also be partially distributed in a way that one oI the interacting eNBs would assume leadership, guiding the others towards an agreed solution.
Figure 4: Hybrid Implementation of Use Case Entities 13
Hierarchical Clustering of UCs A hierarchical ordering oI UCs may be used to avoid ping pong activity between UCs especially in UC-UC trigger scenarios and parameter value conIlicts. A sample rule could be that a lower rank UC cannot change the decision made by a higher rank UC on a conIlict parameter within a speciIied time period. Another rule could be that in case where one UC could trigger another, a lower rank UC may be reIuted Irom execution Iollowing activity Irom a higher rank UC. "uantitative arbitration Arbitration could be undertaken in parameter value conIlicts by evaluating a cost to the entire network oI a parameter taking a value diIIerent Irom the ones proposed by the contending UCs. The Iinal value selected would be the one that minimises this network wide cost. Coarse & fine Optimization Parameters A course vs. Iine adjustable parameter ranking may be undertaken so that some parameters those with eIIect on many UCs, are optimised Ior a larger set oI UCs while others are used to optimise individual UCs. An example is optimising transmit power and Downlink transmit power Ior a set oI 9 UCs to some semi-permanent values and then using the rest oI parameters to optimise individual UCs (Iigures 5). The course parameters like transmit power could be adjusted much less Irequently than the Iine parameters like Ior example in an average ratio oI 1:1000 times.
Figure 5: Separation of parameters into Course and fine optimisation parameters. Suboptimal solutions with reduced parameter sets Simple non coordination based solutions or at least those requiring minimal coordination Iunctions can be designed Ior sub optimal solutions by reducing input and / or adjustable uL 1x ower upllnk C AnLenna 1llL uL 1x o/88 8eam lorm Sched aram CC uC aram CCu CPu CCC AC Pe-8 Cov/lnL lnLeference Coord ackeL Sched Lgy Cons 1
parameter sets Ior the individual UCs. As an example iI interIerence coordination is undertaken without adjusting antenna tilts, Iigure 5 can still be decoupled along the dotted line. Characteristics based Decoupling It is expected also that the anticipated connections between UCs can also be broken by consideration oI the individual characteristics oI the UCs. For example even though AGP and interIerence coordination both adjust the antenna tilt, the Iact that the tow UCs run at diIIerent network stages (i.e. conIiguration Ior AGP and optimisation Ior interIerence coordination) implies that they are actually not connected and do not need integration. Other considerations Other strategies will also be considered especially Ior speciIic problem cases. For example applicable approaches could be game theory in conIlicts situations and consensus in both conIlict and cooperation scenarios. Coordination and integration Iunctions could be designed based on available methods in these approaches Ior quantitative and / or algorithmic solutions. 3.3. Purpose of the Study The major purposes oI this project are to understand the interactions among a set oI UCs and to develop algorithms that can be used to implement their integration in a way that they are able to work in a coordinated manner during the selI-optimization process. The project will however, contribute towards validating other research work by applying the results obtained in that research in the design and simulation oI the algorithms. Finally, the project will contribute new knowledge on the subject in both methods oI designing and implementing the solutions as well as in the new results culminating Irom this research. 3.4. Risks The major risk to the project is the ability to manage the complexity combined with little literature on the subject. OI critical importance also will be the knowledge oI algorithms and data structures which may be required in order to design, implement and evaluate eIIicient algorithms. Other Ioreseeable risks could be challenges with stability and robustness oI the algorithms although these can be managed. 1
4. CONTRIBUTIONS This project will be undertaken in the international graduate school oI mobile communication at TU Ilmenau within the context oI research towards autonomous adaptation oI systems to dynamically changing environments. The core oI the work will be within the Modeling and Evaluation` working group oI the school but is expected to use the results and algorithms oI the SelI Organized Decision Making` working group and will test the results through the Demonstrator` working group. The work on interaction oI selI-organization use cases in LTE is strongly related to other topics oI the Graduate School especially on SON. SpeciIically, it interIaces with SelI-Organization oI Coverage and Capacity in Future Radio Access Networks (RAN) by Muhammad Naseer, and SelI-Organized Mobility Load Balancing Ior Future Radio Access Networks by Nauman Zia. 1
5. WORK SCHEDULE 10/1/11 12/31/11 3/31/12 /30/12 /2/12 12/2/12 3/30/13 /2/13 /2/13 12/2/13 3/2/1 /2/1 /2/1 reparaLlon Any Courses? Self organlsaLlon course 2 SLaLe of Lhe ArL 8ackground SLudy SC- CperaLlonal descrlpLlon SC- demo CongesLlon CLr SC- uC ConnecLlons rellmlnary roblem descrlpLlon uemonsLraLe ConnecLlons ueLalled roblem lormulaLlon SLudy of roposed MeLhods Analyse llmlLaLlons of Lhe meLhods uemonsLraLe Clobal Challenge SoluLlons uevelop and demonsLraLe ldea 1 SlmulaLe soln and documenL resulLs uevelop and demonsLraLe ldea 2 SlmulaLe soln and documenL resulLs uevelop and demonsLraLe ldea 3 SlmulaLe soln and documenL resulLs CompleLlon AcLlvlLles 1hesls WrlLlng uesserLaLlon uefence Ma[or uuraLlon AcLlvlLy uuraLlon Mll esLone 1
REFERENCES |1| http://www.3gpp.org/LTE |2| J.L. VAN DEN BERG ET AL, 'SelI-Organisation in Future Mobile Communication Networks, Deutsche Telekom AG, version 1.0, December 2008 |3| L.C. SCHMELZ, J.L. VAN DEN BERG, R. LITJENS,A., M. AMIRIJOO, O. LINNELL, C. BLONDIA, T. KRNER, N. SCULLY, J. OSZMIANSKI, SelI-conIiguration, - optimisation and -healing in wireless networks 2008RF |4| Next Generation Mobile Networks, 'Use Cases related to SelI Organising Network, Overall Description, NGNM, May 2007. http://www.ngmn.org/technology.html |5| 3GPP TR 36.902 V0.0.1, 'Evolved Universal Terrestrial Radio Access Network (E- UTRAN); SelI-conIiguration and selI-optimizing network use cases and solutions |6| Monotas, 'Mobile Network Optimisation Through Advanced Simulation, http://www.macltd.com/monotas/index.php |7| GandalI, 'Monitoring and selI-tuning oI RRM parameters in a multi-system network, http://www.celtic-initiative.org/Projects/Celtic-projects/Call2/GANDALF/gandalI-deIault.asp |8| SOCRATES Deliverable D2.1: Use Cases Ior SelI-Organising Networks, EU STREP SOCRATES (INFSO-ICT-216284), Version 1.0, March 2008 |9| SOCRATES Deliverable D2.2: Requirements Ior SelI-Organising Networks, EU STREP SOCRATES (INFSO-ICT-216284), Version 1.0, June 2008 |10| SOCRATES Deliverable D2.3: Assessment criteria Ior SelI-Organising Networks, EU STREP SOCRATES (INFSO-ICT-216284), Version 1.0, June 2008. |11| SOCRATES Deliverable D2.4: 'Framework Ior the development oI selI-organisation methods, EU STREP SOCRATES (INFSO-ICT-216284), September 2008. |12| SOCRATES Deliverable D2.5: 'Review oI use cases and Iramework, EU STREP SOCRATES (INFSO-ICT-216284), March 2009. |13| SOCRATES Deliverable D2.6: 'Review oI use cases and Iramework II, EU STREP SOCRATES (INFSO-ICT-216284), December 2009. |14| Thomas Jansen, Irina Balany, Ingrid Moermanz, Thomas Kurner, 'Handover parameter optimization in LTE selI-organizing networks, COST2100 TD(10)10068 Athens, Greece, 2010/Febr/03-05 1
|15| Jose Alonso-Rubio, SelI-Optimization Ior Handover Oscillation Control in LTE, IEEE/IFIP Network Operations and Management Symposium, Osaka, Japan, 2010. |16| I. Balan, T. Jansen, B. Sas, I. Moerman & T. Krner, 'Enhanced weighted perIormance based handover optimization in LTE, Proceedings oI FNMS, arsaw, Poland, 2011 |17| B. Sas, K. Spaey, I. Baran, K. Zetterberg and R. Litjens, 'SelI-optimisation oI admission control and handover, Proceedings oI ISON, 15-18 May 2011, Budapest, Hungary |18| K. Spaey, B. Sas, C. Blondia, 'SelI-Optimising Call Admission Control Ior LTE Downlink, Joint COST 2100 / SOCRATES workshop, February 5, 2010 |19| A. Lobinger, S. SteIanski, T. Jansen & I. Balan, Load balancing in downlink LTE selI- optimizing networks, VTC2010-Spring, Taipei, Taiwan, May 16-19, 2010. |20| Lobinger, A.; SteIanski, S.; Jansen, T.; Balan, I.,Coordinating Handover Parameter Optimization and Load Balancing in LTE SelI-Optimizing Networks, IEEE 73rd Vehicular Technology ConIerence (VTC-Spring), Budapest, Hungary |21| M. Amirijoo, P. Frenger, F. Gunnarsson, J. Moe, K. Zetterberg, 'Towards RACH SelI Tuning in LTE IEEE Vehicular Technology ConIerence, Spring, 2009. |22| T. Jansen, M. Amirijoo, U. Trke, L. Jorguseski, K. Zetterberg, R. Nascimento, L. C. Schmelz, J. Turk, I. Balan, 'Embedding Multiple SelI-Organisation Functionalities in Future Radio Access Networks, 69th Vehicular Technology ConIerence, VTC2009- Spring, Barcelona, Spain, 2009 |23| SOCRATES Deliverable D5.9: 'Final Report on SelI-Organisation and its Implications in ireless Access Networks, EU STREP SOCRATES (INFSO-ICT-216284), Dec2010 |24| Tobias Bandh, Henning Sanneck, Raphael Romeikat, 'An experimental system Ior SON coordination, ISON IEEE 73 rd Vehicular Technology ConIerence, Spring 2011 |25| T. Bandh, R. Romeikat, H. Sanneck, H. Tang, 'Policy-based coordination and management oI SelI-Organizing-Network (SON) Functions, IFIP / IEEE Symposium on Integrated Management, Dublin, Ireland, May 2011.