Modeling Uncertainty in Wireless Communication Fabian Kuhn Algorithms

Modeling Uncertainty in Wireless Communication Fabian Kuhn Algorithms and Complexity Group University of Freiburg Germany Fabian Kuhn WRAWN 2013

Modeling Wireless Communication Fabian Kuhn WRAWN 2013 2

Graph-Based Models • Fabian Kuhn WRAWN 2013 3

SINR Models • Fabian Kuhn WRAWN 2013 4

Wireless Models • Many existing algorithms make heavy use of this! Fabian Kuhn WRAWN 2013 5

Simulating Collision Detection • Fabian Kuhn WRAWN 2013 6

Wireless Communication • Fabian Kuhn WRAWN 2013 7

Dynamic Behavior Environmental conditions change over time • External interference changes over time • Geometry, environmental factors can change Network itself might be dynamic • Devices can be mobile (mobile ad hoc networks) © James Mc. Lurkin, Rice U. Fabian Kuhn WRAWN 2013 8

Uncertainty Communication inherently unreliable / unpredictable • Impossible to explicitly model every relevant factor • Hard to exactly measure behavior of comm. channels • Hard to exactly coordinate transmissions of different nodes • Communication channels change over time – Possibly frequently and significantly • Network might even be dynamic Fabian Kuhn WRAWN 2013 9

Modeling Uncertainty Two basic approaches 1. Add random “noise” to the model – Depending on distance, trans. powers, etc. , a message is received with a certain probability 2. Add non-deterministic behavior to the model – • Fabian Kuhn In some cases, an adversary decides whether a message is received Leads to particularly robust algorithms WRAWN 2013 10

Example: Dual Graph Model • Fabian Kuhn WRAWN 2013 11

Single-Message Broadcast • [Kuhn, Lynch, Newport, Oshman, Richa; PODC ‘ 10] Fabian Kuhn WRAWN 2013 12

Single-Message Broadcast • [Kuhn, Lynch, Newport, Oshman, Richa; PODC ‘ 10] Fabian Kuhn WRAWN 2013 13

Weaker Adversaries • Fabian Kuhn WRAWN 2013 14

Modeling Uncertainty • Fabian Kuhn WRAWN 2013 15

Dynamic Networks • Fabian Kuhn WRAWN 2013 16

Some General Questions How much non-determinism can we tolerate to still use techniques of classic models? • Or at least to still do things (almost) as efficiently? What are the inherent costs of dealing with nondeterministic behavior? • Which things are not possible or much more expensive? – Depending on the amount of non-determinism allowed Go beyond broadcast / inf. dissemination? • How should we define local structures in a dynamic graph? Fabian Kuhn WRAWN 2013 17

Summary • Wireless communication behaves unreliably • Needs to be reflected in the models • Leads to particularly robust algorithms • And a better general understanding of the problems – E. g. , what’s the limit of each technique? Fabian Kuhn WRAWN 2013 18

Thanks! ? s n o i t s e ? u s Q t n e m m o C Fabian Kuhn WRAWN 2013 19