Towards Cognitive Autonomous Networks in 5G

Transcription

1 Towards Cognitive Autonomous Networks in 5G Stephen S. Mwanje Nokia Bell Labs, Germany

2 Why is this discussion important? Need for Cognition in Network Automation Multi-RAT/Band Networks higher complexity, CapEx, OpEx Solution = Automation e.g. SON SON is inadequate Static, rule-based mapping of KPIs to actions inflexible to the infinitely many contexts Operator Customer BSS Customer OSS Planning Customer Service fulfilment experience Mgt Data & Analytics Network Engineering Network Management SON - Self-Organizing Networks KPI - Key Performance Indicator Network Management Automation (NMA) maximizes value of automation

3 But what is Cognition? Automation, self-organization and cognition Nature motivates Self-Organization E.g. insect colonies on foraging for food Non-deterministic self-organization -.. foraging is also a learning process. [6]. Behavior emerges from internal cognitive process Cognition (in humans): the collection of mental sub-processes to support acquisition of knowledge and understanding through sensory stimuli, experience and thought Problem solving sensation Learning thought attention memory perception Intelligence Four basic processes as core; Higher processes as combinations of basic and other higher processes

4 How then should cognition be modelled? Cognition: A perception-reasoning pipeline Actions Reflex (e.g. blink) subconscious (e.g. stop) conscious (e.g. run, fight, get a weapon) Sense Conceptualize Perception Contextualize Reasoning Organize Infer R R L L R L R L F F S S F S Memory F S

5 What levels of automation can be achieved in networks? Taxonomy - Two dimensions decision & action Scripted network human-controlled scenario specific execution of scripts Self-Organizing Network automatically interprets events to determine cause-effect relations, selects and executes of actions Autonomous network - acts on its own, but may not be able to reason suitably Preset time/sequence trigger Detect a network event Correlate events/data Diagnose events Contextualize events/data Anticipate standalone events Anticipate correlated events Take one or more action(s) Automated Scripted Self-Organizing Autonomous Cognitive-Autonomous Cognitive network - reasons to formulate decision/actions, although actions may require operator approval before execution Different capabilities are achieved by the different form of automation

6 How do we get to CANs? Transition trajectory towards CANs Manual Automated Cognitive - automated Cognitive - autonomous User User User User Environment Environment Configure interact SON contextualize Derive Set Network Parameters Operator OAM Days- Weeks BSS/OSS NM configure Network OAM Days- Months SON KPIs BSS/OSS Mins configure Network Define contexts months OAM/ History CNM Secs- Months KPIs BSS/OSS configure Mins Network Update/ upgrade OAM/ History contexts BSS/OSS CAN Secs- Months KPIs Mins configure Network Operator decides how and when to execute actions SON Function selects from manually pre-set actions Operator defines contexts; CF learns context-specific actions Cognitive-Autonomous Function learns contexts and their actions

7 How will the CAN system look like? CAN Framework Cognitive Function (CF) may contain: DApps - Learners for specific objectives CME (controller) - set boundaries and decide if recommendation is executed CE - resolve conflicts among DApps recommendations EMA - define and label observed network events and contexts NOM - translate the operator s goals to technical targets for CFs Operator Goals Network Objectives Manager Interfaces: a - Technical targets b - Raw network data c - Network states d - Preferred config ns e - Network config ns f - Config n reports g - Coordination decisions h - DApps config ns a Decision Applications (Dapps) (DApps) Environment Modelling & Abstraction d f, h Configuration Management Engine f (Controller) e NETWORK & ENVIRONMENT Static nature, e.g. Locality, BS location, RAT, Spectrum c g Network data b Coordination Engine KPIs Dynamic nature, e.g. Velocity, User count, Applications, etc

8 on Can we concretize these ideas? Example CF Mobility Robustness Objective: learn optimal behavior per mobility (velocity) state in cell Control - Hysteresis (Hys) & Timeto-Trigger (TTT) KPIs Rates of HO Ping-Pongs (PPR) RLF due to early HO (FER) RLF due to late HO (FLR) Scenario: LTE network of 21 cells Velocity: 3,10,30,60,90,120 kmph HO: handover; RLF: Radio Link Failures Implementation Supervised learning on Velocity, Hys, TTT and rates Nearest Neighbors Regression (KNN) Decision Trees (DT ) Random Forests (RFR) Extremely Randomized Trees (ETR) Evaluation: RE - absolute error relative to expectation of rate E{y} RE = y i y i E{y}

9 What has been observed so far? Learning network response ML algorithms predict behavior in unknown states using response in known states

10 Any lessons from the study? Not simply off the shelf Achieve better outcomes through smart combinations of algorithms

11 Learning methods can automate network management not only to learn the best configurations for specific states but also the underlying network response which can then be used to predict response in unknown states

12 Thank you