8/7/2019 Enrique SN

1/12

A Vertical Handoff Decision

Algorithm in Heterogeneous

Wireless Networks

Enrique Stevens-Navarro

March 9th, 2007

Second UBC-IEEE Workshop onFuture Wireless Systems

2

Presentation Outline

Heterogeneous Wireless Networks

Vertical Handoff

WLAN/Cellular Interconnection

Current research at UBC:

A vertical handoff decision algorithm forheterogeneous wireless networks

8/7/2019 Enrique SN

2/12

3

Heterogeneous Wireless Networks

Different Access Technologies (radio interfaces) and overlapping coverage.

Different Network Architectures and Protocols for transport, routing andmobility management.

Different Service Demands from mobile users (low-data-rate, high-datarate, voice, etc)

Different Operators in the market.

4

Heterogeneous Wireless Networks

8/7/2019 Enrique SN

3/12

5

[1] M. Stem and R. Katz Vertical handoffs in wireless overlay networks, MobileNetworks and Applications, 1998.

Horizontal vs. Vertical Handoff [1]

Horizontal

Handoff

Vertical

Handoff

6

Horizontal HO mainly use received signal strength (RSS) to decide the

handoff

But Vertical HO?

RSS?

Offered bandwidth?

Price?

Power consumption?

Speed?

.

Handoff

Metrics

Handoff Metrics

8/7/2019 Enrique SN

4/12

7

Steps of the Vertical Handoff Process

Step 1: System Discovery

The MT must know which wireless networks are reachable.

Step 2: Handoff Decision

The MT then evaluates the reachable wireless networks to make a decision.

Step 3: Handoff Execution

If the MT decides to perform a VHO, it executes the VHO procedure required

to be associated with the new wireless network.

8

WLAN Cellular Networks Important case ofheterogeneous wireless networks integration.

Two architectures are proposed [2]:

Tightly-coupled inter-working.

Loosely-coupled inter-working.

[2] M. Buddhikot, G. Chandranmenon, S. Han, Y. Lee, S. Miller and L. Salgarelli,Integration of 802.11 and 3G Wireless Data Networks in Infocom 03.

8/7/2019 Enrique SN

5/12

9

WLAN Cellular Networks Integration

Standardization efforts:

Both 3GPP and 3GPP2 are working in the inter-working with WLAN as an

extension of their radio access networks.

IEEE 802.21 Media Independent Handover Group is working toward the

seamless handoffs between:

IEEE 802.XX family.

IEEE 802.XX and 3G Cellular

Between 802.11 ESSs.

10

Research at UBC: Vertical Handoff Decision Algorithm

The problem is formulated as a Markov Decision Process (MDP) .

There is a link reward function associated with the QoS of the connection

and also a signaling cost function associated with the signaling overhead

and processing load of the vertical handoff.

Both functions are introduced to capture the tradeoffs between the network

resources utilized by a connection and the processing load incurred in the

network.

Objective: determine an optimal policy which maximizes the total expected

reward per connection.

where (s) is the total expected reward, given policy is used with initial state s,

and discount factor .

8/7/2019 Enrique SN

6/12

11

An MDP model consists of five elements: decision epochs, states, actions,

rewards and transitions probabilities.

Decision Epochs

Xt denotes the state at decision epoch t.

Ytdenotes the action chosen at decision epoch t.

Ndenotes the time that the connection terminates.

Model Formulation

12

Vertical Handoff Decision Epochs

8/7/2019 Enrique SN

7/12

13

State Space

NetworkUnits of Bandwidth

Units of Delay

Mdenotes the number of collocated networks.

Bm denotes set of available bandwidth form networkm.

Dm denotes set of delay from networkm.

14

Rewards

Link

Reward

Function

Signaling

Cost

Function

Reward Function

8/7/2019 Enrique SN

8/12

15

Actions

Transition Probabilities

Given the current state s=[i,b1,d1, ,bm,dm] and the chosen action be a,

the probability function that the next state s=[j,b1,d1, ,bm,dm] is:

where Mis the number of collocated networks.

16

Optimality Equations

Let (s) denote the maximum expected reward given initial state s

Then, the optimality equations are given by:

The solutions of the optimality equations correspond to the maximum

expected total reward (s) and the optimal policy *(s).

To solve our problem, the value iteration algorithm was used.

8/7/2019 Enrique SN

9/12

8/7/2019 Enrique SN

10/12

19

Numerical Results

Figure 1 (a) Expected total reward and 1 (b) expected number of vertical

handoff versus switching cost Ki,a with =0.975 and =0.25.

20

Numerical Results

Figure 2 (a) Expected total reward and 2 (b) expected number of vertical

handoff versus discount factor with switching cost K1,2= K2,1=1 and

=0.25.

8/7/2019 Enrique SN

11/12

21

Summary of Results

The MDP scheme gives higher expected total reward and lower expected

number of vertical handoffthan SAW and GRA.

The MDP model can be applied to a wide range of conditions.

Link reward function (e.g. real-time, non-real-time).

Signaling cost function (Ki,a, re-routing complexity, roaming)

Tradeoff between network resources utilized by the connection and the

signaling load incurred in the network.

Extend the MDP model and simulation model.

22

Summary:

Heterogeneous Wireless Networks

Vertical Handoff

WLAN/Cellular Inter-working

Research at UBC

8/7/2019 Enrique SN

12/12

23

Contact Information:

Enrique Stevens-Navarro

Office: Kaiser 4090

Email: enriques@ece.ubc.ca

http://www.ece.ubc.ca/~enriques/

Thank

You!