View
1
Download
0
Category
Preview:
Citation preview
Enjeux et défis scientifiques pour un déploiement durable de l’IoT : état de l’art de solutions d’optimisation énergétique
Journée Scientifique DS4H - Concilier numérique et environnement ?Sophia Antipolis - 2 Décembre 2019
Alain Pegatoquet – Maître de Conférences, LEAT
▪ Internet of Things is a buzzy phrase…
▪ IoT is not easy to define…
▪ Internet of Things definitions
▪ ”A network of items each embedded with sensors which are connected to
the Internet.” IEEE Institute. March 2014
▪ ”Cyber physical systems (CPS) involve connecting smart devices and
systems in diverse sectors like transportation, energy, manufacturing and
healthcare in fundamentally new ways.” NIST
2
What is Internet of Things (IoT)?
▪ IoT Human + Physical Objects (sensors, controllers, actuators, devices,
computing, storages) + Internet
3
Internet of Things Elements [Farhan 2018]
P2P People to People
M2P Machine to People
M2M Machine to Machine
4
Internet of Things Architecture [Farhan 2018]
▪ IoT Sensors/Actuators (Application dependent) + Gateway (Access Point) +
Internet + Services
▪ Wireless Sensors Networks (WSNs) is one of the key underlying enabling
technology for IoT
2
Wearable Devices
Current Trends
▪ More functionality per area
▪ Greater HW/SW requirements
▪ Longer Battery lifetime
Biggest Challenges
▪ Energy storage (capacity)
▪ Energy consumption
▪ Energy harvesting
▪ Smart devices
▪ Miniaturization
[Eggimann 2019]
6
Smart Connected Glasses
Drowsiness Detection Fall Detection Activity Tracking
▪ A typical IoT device
▪ Different application use cases
Constraints :
▪ Computations
▪ Memory
▪ Battery
[Arcaya 2019]
7
IoT Challenges
▪ Business : Silos vs. Horizontal, Application Domains, The revenue
(per device, connectivity, Data), New business model…
▪ Societal : Privacy, Data Ownership, Security, Easiness of Use, Social Cooperation
▪ Technical : complexity, communication, Data, Software platforms, energy
consumption…
[Farhan 2018]
▪ Size of the IoT market worldwide from 2017 to 2025 (in billion U.S. dollars)
8
IoT Market
[Source: Statista 2019]
▪ The global IoT market is expected to grow to $212 billion by the end of 2019.
▪ The technology reached 100 billion dollars in market revenue for the first time in
2017, and forecasts suggest that this figure will grow to around 1.6 trillion by
2025.
▪ IoT: 75 billions of connected objects in 2025…
9
What about the Overall Energy Consumption?
Sources : [Lucero 2016] [Rioual 2019]
Optimize Energy Everywhere !
10
▪ IoT Devices Power Management (Node level)
▪ Optimize Components power consumption (DPM, Power/Clock gating, DFVS, etc.)[Ben Ameur 2018] [Mbarek 2013]
▪ Optimize communications (RF activities)[Rault 2014] [Castagnetti 2014] [Le 2016] [Maitra 2016] [Kim 2015] [Huang 2013] [Sentieys 2013] [Ait Aoudia 2017]
▪ Optimize activities (task scheduling)[Arcaya 2019]
▪ Autonomous IoT Devices
▪ Energy harvesting [Vullers 2010]
▪ Power management[Kansal 2007] [Castagnetti 2012]
▪ Network-level Optimizations
▪ IA-based approach
11
Optimizing Energy Everywhere !
▪ Routing protocols (multi-hop)
12
Network Level Optimizations
▪ Cooperative networks
▪ The nodes work together to improve the communication or the energy consumption of the
overall network (e.g. cluster head) [Liang 2010] [Chen 2016]
▪ Edge Computing…
▪ Process data as close as possible from the nodes to minimize the transmission, reduce the
latency and optimize the bandwidth utilization…
[Messous 2018]
Edge Computing &
IA based approach
13
▪ IA has already been widely used to optimize performance/energy for embedded
systems
▪ All types of learning approaches
▪ Supervised, unsupervised, reinforcement
▪ Bio-inspired
▪ For all kinds of problematic
▪ Communications (MAC, Routing, Self-learning radios, Cooperative networks, Rate control…)
▪ Tasks scheduling
▪ Power management
▪ Energy Harvesting
▪ A new paradigm for embedded architectures
14
IA for Power Management
Neuroscience
▪ Spiking Neural Networks (SNN) motivations
15
Hardware Spiking Neurons for Embedded Artificial Intelligence
➔ Accuracy
➔ Hardware implementation
◆ Powerful computer (CPU/GPU)
◆ Embedded systems (FPGA/ASIC)
Neuroscience
Spiking Neural Networks (SNN)
Machine
Learning
Formal Neural Networks
[Abderrahmane 2020]
▪ Spiking versus Formal Neural Networks
16
Hardware Spiking Neurons for Embedded Artificial Intelligence
Integrate-and-FirePerceptron
SNN neuron:
▪ Addition
▪ No more Multiplication !
▪ Comparison
FNN neuron
▪ Multiplication + Addition
▪ Non-linear activation function
[Abderrahmane 2020]
▪ Results with MNIST classification
17
Hardware Spiking Neurons for Embedded Artificial Intelligence
SNN vs. FNN hardware cost and accuracy
ANN Model FNN SNN SNN Gain (%)
Accuracy (%) 95.73 95.37 -0.38
Total power (mW) 52.22 28.46 45.37
Total area (mm²) 1.888 0.869 53.98
Place and route synthesis on 65nm technology
▪ For almost the same accuracy, SNN is around ~50% more efficient in terms
of power / area
[Abderrahmane 2020]
18
[Abderrahmane 2020] Nassim Abderrahmane, E. Lemaire and Benoit Miramond, Design Space Exploration of Hardware Spiking Neurons for Embedded
Artificial Intelligence, to appear in Neural Networks Journal, Vol. 121, pp. 366-386, 2020.
[Ait Aoudia 2017] Faycal Ait Aoudia, “Energy Harvesting Wireless Sensor Networks Leveraging Wake-up Receivers: Energy Managers and MAC
Protocols”, PhD Thesis, 28th Sept. 2017, Université de Rennes 1, Lannion, France.
[Arcaya 2019] A. Arcaya Jordan, A. Pegatoquet and A. Castagnetti, Smart Connected Glasses for Drowsiness Detection: a System-Level Modeling
Approach, IEEE International Sensor Applications Symposium (SAS), Sophia Antipolis, France, March 11-13, 2019
[Ben Ameur 2018] Amal Ben Ameur, Michel Auguin, François Verdier, Valerio Frascolla, Mobile Terminals System-Level Memory Exploratio for Power
and Performance Optimization, 28th International Symposium on Power and Timing Modeling, Optimization and Simulation
(PATMOS), pp. 23-28, Costa Brava, Spain, July 2-4th, 2018
[Castagnetti 2012] A. Castagnetti, A. Pegatoquet, C. Belleudy and M. Auguin, An Efficient State of Charge Prediction Model for Solar Harvesting WSN
Platforms,19th IEEE International Conference on Systems, Signals and Image Processing (IWSSIP), Vienna, Austria, 11-13 April,
2012.
[Castagnetti 2014] Andrea Castagnetti, Alain Pegatoquet, Trong-Nhan Le and Michel Auguin, A Joint Duty-Cycle and Transmission Power
Management for Energy Harvesting WSN, IEEE Transactions on Industrial Informatics Journal, Special section on “Industrial
Wireless Sensor Networks”, Volume 10, Issue 2, pp. 928-936, May 2014.
[Chen 2016] Hongbin Chen, Xueyan Li et Feng Zhao. A reinforcement learning-based sleep scheduling algorithm for desired area coverage in
solar-powered wireless sensor networks. In IEEE Sensors Journal 16.8 (2016), p. 2763-2774.
[Eggimann 2019] Manuel Eggimann, Stefan Mach, Michele Magno and Luca Benini, A RISC-V Based Open Hardware Platform for Always-On
Wearable Smart Sensing, IEEE 8th International Workshop on Advances in Sensors and Interfaces (IWASI), 2019
[Farhan 2018] Laith Farhan et. al, A Concise Review on Internet of Things (IoT) - Problems, Challenges and Opportunities. 11th International
Symposium on Communication Systems, Networks, and Digital Signal Processing (CSNDSP 2018), Budapest, Hungary, 2018
[Huang 2013] P. Huang, L. Xiao, S. Soltani, M. W. Mutka, and N. Xi, The Evolution of MAC Protocols in Wireless Sensor Networks: A Survey, IEEE
Communications Surveys Tutorials, vol. 15, no. 1, pp. 101-120, January 2013.
[Kansal 2007] A. Kansal et al., Power management in energy harvesting sensor networks, ACM Transactions on Embedded Computing Systems (TECS), vol.6,
2007
References
19
[Kim 2015] T. Kim, I. H. Kim, Y. Sun, et Z. Jin, « Physical Layer and Medium Access Control Design in Energy Efficient Sensor Networks: An
Overview », IEEE Transactions on Industrial Informatics, vol. 11, no 1, p. 2-15, févr. 2015.
[Le 2016] T-N. Le, A. Pegatoquet, Trinh Le Huy, Leonardo Lizzi and Fabien Ferrero, Improving Energy Efficiency of Mobile WSN Using
Reconfigurable Directional Antennas, IEEE Communications Letters, Vol. 20, Issue 6, pp. 1243-1246, April 2016.
[Liang 2010] Xuedong Liang, Min Chen, Yang Xiao et al. MRL-CC : a novel cooperative communication protocol for QoS provisioning in wireless
sensor networks. In International Journal of Sensor Networks 8(2):98-108 · August 2010.
[Lucero 2016] Sam Lucero et al. IoT Platforms : Enabling the Internet of Things. IHS market, White paper, 2016.
[Maitra 2016] Tanmoy Maitra and Sarbani Roy, A comparative study on popular MAC protocols for mixed Wireless Sensor Networks: From
implementation viewpoint, Computer Science Review, Elsevier, Volume 22, Pages 107-134, November 2016
[Mbarek 2013] O. Mbarek, A. Pegatoquet and M. Auguin, Power Domain Management Interface: Flexible Protocol Interface for Transaction-Level
Power Domain Management, IET Computers & Digital Techniques Journal, Volume 7, Issue 4, pp. 155-166, July 2013.
[Messous 2018] S. Messous, N. Liouane, A. Pegatoquet and M. Auguin, An improved energy-efficient routing protocol based on Minimum Spanning
Tree for Wireless Sensor Network, International Conference on Sensors, Systems, Signals and advanced technologies (SSS 2018),
Hammamet, Tunisia, May 10-12, 2018
[Rault 2014] Tifenn Rault, Abdelmadjid Bouabdallah and Yacine Challal, Energy efficiency in wireless sensor networks: A top-down survey,
Computer Networks, Volume 67, Pages 104-122, 4 July 2014
[Rioual 2019] Yohann Rioual, RL-based Energy Management for Autonomous Cyber Physical Systems, PhD thesis, University of Bretagne Sud,
Dec 2019
[Sentieys 2013] Olivier Sentieys, Protocol-Level Power Optimization of Wireless Sensor Networks, EcoFac2012, May 21-25 2012, La Colle-sur-Loup,
France
[Statista 2019] Online - https://www.statista.com/statistics/976313/global-iot-market-size/
[Vullers 2010] R. J. M. Vullers, R. v. Schaijk, H. J. Visser, J. Penders, and C. V. Hoof, Energy Harvesting for Autonomous Wireless Sensor Networks,
IEEE Solid-State Circuits Magazine, vol. 2, no. 2, pp. 2938, Spring 2010.
References
Journée Scientifique DS4H - Concilier numérique et environnement ?Sophia Antipolis - 2 Décembre 2019
Thank you for your attention !
Recommended