1. Improving end-to-end packet
delivery ratio in high traffic
multi-hop ad hoc networks
Syed Rehan Afzal, Majid Nabi, Sander
Stuijk, Twan Basten
Electronic Systems Group (ES), Dept. of EE,
Eindhoven University of Technology

2. 2
Context & Problem Statement
Limited Energy
Limited Tx Range
Unreliable
Discovery (d > Tx)
Connectionless Data
End-to-end QoS
Dynamic Qualities
Packet Delivery
Ratio (PDR)
Bandwidth
Latency
Throughput
Connection with highest
Packet Delivery? Grub

3. • We present xDDR (eXpected Directional Delivery Ratio)
– Based on ETX metric
– Leverages on the reasons discovered during our
experiments accounting to ETX’s shortcomings
– xDDR shows considerable gain in end-to-end PDR
3
xDDR – An Improved Packet Delivery Ratio Estimation Metric
End-to-end Packet Delivery Ratio Comparison

4. 4
Network Architecture
• Type of qualities
– Link-level
– Stream/connection level

5. 5
Proactive Beaconing & Routing Module
GrubNode A √ √ X √ X X √ √ . . . . . . .
Node B √ X X √ √ √ X √ . . . . . . .
Node C √ √ X X √ X √ . . . . . . .
Pick best & Reply
Link-level PDR
Connection Discovery
Data Transmission
Proactive Beaconing
In-band Signaling
Connection 1?
Connection 2?
Connection 3?

6. • Role of MAC Layer & the broadcast problem
– Unicast communication improves with MAC retry Limit
– While broadcast packets are oblivious to MAC Retry Limit
– Do not employ RTS/CTS mechanism
– Simulation Environment (Omnet++ with INETMANET)
6
Motivation - Problems with ETX
Data Delivery Ratio v/s ETX estimate

7. • Role of broadcast discoveries on link ETX
– Under high traffic circumstances excessive broadcast
discovery packets cause some nodes to wrongly under
estimate link quality
7
Motivation - Problems with ETX
Link ETX estimates w.r.t. route discoveries

8. 8
xDDR (eXpected Directional Delivery Ratio) - Outline
• Broadcast only during initialization phase
• Uses unicast as opposed to broadcast to cater RTS/CTS
support as well as MAC retry limit
• Selects a subset of neighboring nodes as proactive beaconing
unicast recipients
𝜆 𝑛 𝑀, 𝑛 𝑂, 𝑡′ =
𝑐𝑜𝑢𝑛𝑡(𝑡′ − ŵ, 𝑡′)
ŵ 𝜀
𝑐𝑜𝑢𝑛𝑡(𝑡′ − ŵ, 𝑡′) = number of proactive beacons received
𝜆 = Directional level delivery ratio
ŵ = WindowSize
𝜀 = Beaconing interval

9. 9
xDDR – Window Selection
• xDDR uses xDDRWindow, double the size of ETXWindow
• Each node attempts to estimate PDR by selecting a window
where there is no overlap with RREQ broadcast discovery
packets
• This window is the xDDRmWindow

10. 10
Experimental Setup & Results
10
End-to-end Packet Delivery Ratio Comparison
xDDR vs ETX Estimation

11. 11
xDDR packet overhead
11
Grub
Ratio of packets transmitted by MAC layer of each node w.r.t. total UDP
packets delivered

12. ― Proposed xDDR, an improved link-level delivery ratio
estimation metric based on ETX
― In high traffic scenarios, selective unicast beaconing yields
more accurate packet delivery ratio estimates (xDDR-I)
― xDDR-I coupled with xDDRWindowSelection avoids quality
estimation during conflicting window slots
― Coupled with quality aware routing results in a higher end-
to-end multi-hop data delivery
― This implies on average 3 times gain over Min. hop count,
1.94 times gain over ETX and 1.7 times gain over HETX
12
Conclusions

13. Thank You
S. Rehan Afzal
s.r.afzal@tue.nl
Electronic Systems Group (ES), Dept. of EE,
Eindhoven University of Technology, The Netherlands