1. KTH ROYAL INSTITUTE
OF TECHNOLOGY
Grant-Free Radio Access for IoT
Communications
Amin Azari
RS Lab, ICT School, KTH
2. Part I: Introduction
Part II) Coexistence analysis
Interference analysis
Coverage analysis
Deployment and operation strategies
Analytical coverage, reliability, cost modeling
Optimized operation points
Part IV) Protocol design
Elements of the protocol
Performance evaluation and results
Part V: Future works
Outline
2 / 41
3. Amin Azari, born 1988, Iran
Education:
BS (2010), MS (2013), Tehran & Rostock University (Iran & Germany).
Lic. (2016) from ICT school, KTH:
Battery lifetime-aware cellular network design
Results: 1 thesis, 1 patent, 3 journals, 7 conf. publications
Towards PhD (from 2017):
Grant-free access for IoT communications
Distributed optimization, machine learning, stochastic geometry
Visit Aalborg University (3 months in spring 2017, Danish grant)
I.1) About me
3 / 41Part I) Introduction
4. I.2) Motivation
Characteristics of IoT
massive in number of connections
small payload size
require long battery lifetime
Legacy cellular networks: grant-based radio access
extensive signaling
sending several bits requires several bytes overhead
– inefficiency for battery-limited devices
– causes congestion in network for control
– for URLLC: reliability depends on 3 transactions that cannot be
coded jointly, i.e. access reservation, response, data transfer.
4 / 41Part I) Introduction
5. I.2) Motivation
no need for synchronization
– reduce delay & overhead & energy consumption
– reduce cost of device, i.e. low-cost oscillator
no need for access reservation
– reduce delay & overhead & energy consumption
but the above benefits are not coming for free…
– When/how much can we gain from grant-free access?
Recent interests:
– [3GPP, “Overall solutions for UL grant free transmission”, 2017.]
– SigFox, LoRaWAN, etc.
5 / 41Part I) Introduction
Grant-free access
6. I.3) Prior Works
Performance Evaluation:
– 2D time-frequency interference modelling using stochastic
geometry for performance evaluation LPWANs,
[Z. Li et al., 2016]
Protocol design:
– Enhanced contention resolution ALOHA
[Clazzer et al, 2013]
– Asynchronous contention resolution diversity
[De Gaudenzi et al, 2014]
6 / 41Part I) Introduction
7. Part I) Introduction
Part II) Coexistence analysis
Interference analysis
Coverage analysis
Part III) Deployment and operation strategies
Analytical coverage, reliability, cost modeling
Optimized operation points
Part IV) Protocol design
Elements of the protocol
Performance evaluation and results
Part V) Future works
Outline
7 / 41
8. II.1) system model
Devices: 𝐾 types:
– Heterogeneous, e.g. SigFox, LoRa.
– Different in:
• Pattern of time/freq. usage,
• Transmit power,
• Rates of packet arrival at nodes, etc.
Dense IoT device deployment, PCP.
– (𝜆 𝑘, 𝜐 𝑘, 𝑓(𝑥)) density of PP, avg. DP per PP, dist. of DP
Fading: Nakagami-m. Pathloss:1/(𝑎 + 𝑏 𝑧
𝛿
)
Application: ISM-band solutions, Cellular-band solutions
8 / 41Part II) Coexistence analysis
Type-1 node
Type-2 node Type-2 AP
Type-1 AP
𝐾 = 2, ISM
9. II.2) Research questions
Probability of uplink coverage at distance 𝑑 from
the AP?
Reliability of communications at a random point of
network for
– a given deployment of APs (joint and independent
reception), and
– a given number of transmissions per packet?
9 / 41Part II) Coexistence analysis
10. II.3) Interference analysis
Finding Laplace functional (LF) of the interference,
because:
LF of interference in cellular networks mature,
while for short packet communications there is no
prior work.
10 / 41Part II) Coexistence analysis
LF interference LF noise
11. II.3) Interference analysis
Problems in using stochastic geometry (prior art):
Partial overlapping in time and frequency
Different transmit powers, transmission times, packet
generation rates, BW of signals, total BW for
communications.
11 / 41Part II) Coexistence analysis
12. II.3) Interference analysis
By defining time and frequency activity factors, we
have derived LF of interference:
In PPP reduces to:
12 / 41Part II) Coexistence analysis
×
ℒ 𝐼Ψ 𝑧 𝑠 =
𝑖=1
𝐾
exp(− 𝜐𝑖 𝐸 ℎ 𝛿 Γ 1 − 𝛿 𝑠 𝑃𝑖
𝛿
)
Own cluster
All other clusters
13. II.3) Coverage analysis
Lowerbound for type 𝑖 at distance ||𝑧||:
Closed-form exact expressions for PPP.
Insights: impact of 𝑊, 𝑃𝑡, 𝜆 𝑎on coverage.
13 / 41Part II) Coexistence analysis
Where for 𝑔 𝑧 = 𝑧
−𝛿
.
14. II.3) Coverage analysis
Now, for any point in the network,
– given distance vector d from neighbour APs
– given number of transmissions per packet
We are able to derive the outage probability.
In PPP deployment of APs:
– CDF of distance to 𝑙-th AP:
then, closed-form expression for outage for a random
device in the network is derived.
14 / 41Part II) Coexistence analysis
18. Part I) Introduction
Part II) Coexistence analysis
Interference analysis
Coverage analysis
Part III) Deployment and operation strategies
Analytical coverage, reliability, cost modeling
Optimized operation points
Part IV) Protocol design
Elements of the protocol
Performance evaluation and results
Part V) Future works
Outline
18 / 41
19. III.1) system model
Similar to part II:
– Dense IoT device deployment, PCP.
– Type of devices:
• Heterogeneous, e.g. SigFox, LoRa, Alarms, etc.
• Different in:
– Pattern of time/frequency resource usage,
– Transmit power,
– Rates of packet arrival at nodes, etc.
– Fading: Nakagami-m.
19 / 41Part III) Deployment and operation
Type-1 node Type-2 node BS
𝐾 = 2, Cellular
20. III.2) Research questions
Deployment phase:
– Regarding impact of 𝑊 and 𝜆 𝑎 on coverage:
*optimized amount of investment in densification and
spectrum leasing for a given reliability for IoT.
Operation phase:
– Regarding impact of transmit power and
subchannel/code selection:
*optimized online policy for transmission for a given
reliability and AP deployment density.
20 / 41Part III) Deployment and operation
21. III.3) KPI modeling
Cost of the access network:
Reliability of communications:
– Investigated in part II, e.g. for grid AP deployment:
Expected battery lifetime of devices:
– 𝐿 𝑖 =
[Energy Storage: 𝐸0]
[Energy Consumed Per Reporting]
[Reporting Period: 𝑇𝑖]
21 / 41Part III) Deployment and operation
22. III.4) Analysis (ongoing)
Investigation of the following problems:
– Deployment:
– Operation:
22 / 41Part III) Deployment and operation
24. Part I) Introduction
Part II) Coexistence analysis
Interference analysis
Coverage analysis
Part III) Deployment and operation strategies
Analytical coverage, reliability, cost modeling
Optimized operation points
Part IV) Protocol design
Elements of the protocol
Performance evaluation and results
Part V) Future works
Outline
24 / 41
25. IV.1) system model
Dense IoT device deployment, PPP
Packet arrival at nodes: Poisson
signal BW
Available spectrum
≪ 1
– Utra-NarrowBand (UNB) system
25 / 41Part IV) Protocol design
26. IV.3) Key idea
In UNB the signal bandwidth is smaller than the
precision of the carrier frequency
random frequency deviation used as an
identification code as in:
[Fyhn, Jacobsen, Popovski, Scaglione, Larsen, 2011]
frequency
Intended carrier frequency
Max
drift
Max
drift
𝑉𝐹𝑖
Δ𝑓𝑖
𝜏𝑖
time
26 / 41Part IV) Protocol design
27. IV.4) Transmission structure
virtual frame
– sends one replica of packet immediately;
– forms a virtual frame, selects N slots to send replicas.
– N: design parameter
– use of SIC
𝑀 slots =𝑀𝑇𝑝 seconds
Reference time
carrier frequency:
𝑓𝑖 = 𝑓 + Δ𝑓𝑖
selected N slots for
main packet/replicas
transmissions
27 / 41Part IV) Protocol design
28. IV.5) Receiver design
Problem of partial interference due to
asynchronicity
Frequency
Intended carrier frequency
Max
drift
Max
drift
𝑉𝐹𝑖
Δ𝑓𝑖
𝜏𝑖
Time
28 / 41Part IV) Protocol design
29. IV.5) Receiver design
i. Use of Zadoff-Chu as preamble in transmitters:
i. length of preamble: design parameter.
ii. tradeoff: overhead vs. decodeability
ii. window the received signal
i. time length: design parameter.
ii. offers tradeoff: delay and decodeability
iii. decode with intended frequency
iv. use periodogram and search for peaks
i. peaks drift from carrier frequencies in the receive
signal, i.e. represent a signature of some devices.
29 / 41Part IV) Protocol design
32. IV.6) Analysis
TiSy: devices are slot synchronized.
FrSy: the CFOs of the devices can take equally-spaced
discrete values, i.e. the devices are sub-channel
synchronized, where the channels are spaced each 200 Hz.
reference: granted radio access with
10 random access resources each 2 seconds.
offered load per channel
32 / 41Part IV) Protocol design
37. IV.7: concluding remarks
In low to medium traffic load regime,
grant-free access can:
– achieve low energy consumption,
– Decrease the experienced delay.
There is a switchover traffic load beyond which
granted-access outperform grant-free access.
Under very low load, reliability increased significantly
37 / 41Part IV) Protocol design
38. Part I) Introduction
Part II) Coexistence analysis
Interference analysis
Coverage analysis
Part III) Deployment and operation strategies
Analytical coverage, reliability, cost modeling
Optimized operation points
Part IV) Protocol design
Elements of the protocol
Performance evaluation and results
Part V) Future works
Outline
38 / 41
39. V: Future works
Use of ML for
– physical layer authentication!
– Better contention resolution!
– Less interference!
39 / 41Part V) Future works
40. V: Future works
Publications:
– Grant-Free Radio Access for Short-Packet Communications over 5G
Networks, A Azari, P Popovski, G Miao, C Stefanovic, IEEE GC 2017
– Optimized Deployment and Operation Strategies for Grant-free Radio
Access IoT Networks, Amin Azari, M Masoudi, C Cavdar, IEEE wireless
communications letters (to be submitted, 2017)
– Grant-free, Grant-based, or Hybrid? Mode Switching MAC for Cellular IoT
Networks, Amin Azari, IEEE Transactions on Wireless Communications (to
be submitted, 2018)
– Grant-free Radio Access IoT Networks: Scalability Analysis in Coexistence
Scenarios, M Masoudi, A Azari, EA Yavuz, C Cavdar, IEEE ICC 2018
40 / 41Part V) Future works