UMTS Radio Network Planning Andreas Eisenblätter Thorsten Koch (ZIB) Alexander Martin (TU Darmstadt)
Overview UMTS Network Planning Optimisation model Integrated planning Computational results Conclusions Cooperation: Operators: Vendor: R&D: EU-Project MOMENTUM KPN, E-Plus, Vodafone Portugal Siemens Mobile Atesio, TU Darmstadt, TU Lisbon, ZIB
UMTS Network Planning Scenario Downtown Berlin Network 16 potential sites 3 antennas per site Digital Building Model Berlin (2002), E-Plus Mobilfunk GmbH & Co. KG, Germany
UMTS Network Planning Scenario Downtown Berlin Network 16 potential sites 3 antennas per site Demand/ Traffic voice telephony video telephony file download streaming multimedia
UMTS Network Planning Scenario Downtown Berlin Network 16 potential sites 3 antennas per site Demand/ Traffic voice telephony video telephony file download streaming multimedia
UMTS Network Planning Scenario Downtown Berlin Network 16 potential sites 3 antennas per site Demand/ Traffic voice telephony video telephony file download streaming multimedia
Network Planning Decisions Decisions sectorization antenna height antenna tilt / azimuth antenna type carrier RRM parameters pilot power Which sites to use?
UMTS Universal Mobile Telecommunication Network W-CDMA Multi-service video telephony user
UMTS Universal Mobile Telecommunication Network W-CDMA Multi-service CIR-target Self interference video telephony user C R I interference
UMTS Universal Mobile Telecommunication Network W-CDMA Multi-service CIR-target Self interference Network quality video telephony user C other cell interference R I interference
Modeling: Sets and Parameters C I other cell int. interference R
Modeling: Variables C I other cell int. interference R
Modeling: Coverage Constraints C I other cell int. interference R
Modeling: Uplink Constraints C I other cell int. interference R
Modeling: Downlink Constraints I C I other cell int. interference R
Modeling: Downlink Constraints II C I other cell int. interference R ^
Modeling: Linearized Downlink CIR-Constraints C I other cell int. interference R
MIP Model Scope & Structure Sites site & equipment costs configuration traffic snapshot Traffic sites installations pilot powers mobile assignment UL power DL power multiple profiles multi-service stochastic input active users spatial distribution Serving mobiles uplink (UL) dedicated channels (CIR) downlink (DL) dedicated channels (CIR) pilot channel (E c /I 0 -based)
MIP Model Scope & Structure Sites site & equipment costs configuration traffic snapshot sites installations pilot powers assignment UL power DL power traffic snapshot assignment UL power DL power Traffic multiple profiles multi-service stochastic input active users spatial distribution Serving mobiles uplink (UL) dedicated channels (CIR) downlink (DL) dedicated channels (CIR) pilot channel (E c /I 0 -based CIR)
Integrated Optimization snapshot generator installation processor attenuation multisnapshot optimiser (MIP) configuration rating generator fitter no OK? yes installation mapping static / dynamic simulations - external assessment
Solving the MIP snapshot generator installation generator processor fitter multisnapshot optimiser (MIP) size O(I x M) using ZIMPL to generate MIP (http://www.zib.de/koch/zimpl) solving MIP using CPLEX with tuned settings explicit generation MIR cuts (simple algebraic structure) numerical challenge: dynamic range of input constraint scaling & reformulation using few snapshots at a time careful pre-selection of initial installations
First Computational Results Scenario Downtown Berlin Network 16 potential sites 3 antennas per site Demand/ Traffic voice telephony video telephony file download streaming multimedia MIP reduced: 20857 rows, 5670 columns, 79476 nze CPLEX root LP: 8.21 sec. heuristics, few BB nodes
First Computational Results Scenario Downtown Berlin Network 16 potential sites 3 antennas per site Demand/ Traffic voice telephony video telephony file download streaming multimedia MIP reduced: 20857 rows, 5670 columns, 79476 nze Path loss predictions by E-Plus Mobilfunk GmbH & Co. KG, Germany CPLEX root LP 8.21 sec. heuristics, few BB nodes
Conclusions Conclusions Sectorisation Locations Antenna type Height Tilt Carrier Pilot Power Fairly accurate MIP for UMTS Radio Network Planning Large realistic data sets, huge effort to collect public benchmarks First computational results on small, realistic scenarios Lacking theoretical underpinning Getting to the practitioners (soon) http://momentum.zib.de Proc. 6 th Informs Telcom. Conference