reinforcement learning
play

Reinforcement Learning Shaswot Shresthamali, Masaaki Kondo, Hiroshi - PowerPoint PPT Presentation

May 24, 2023 •378 likes •771 views

Adaptive Power Management for Energy Harvesting Sensor Nodes using Reinforcement Learning Shaswot Shresthamali, Masaaki Kondo, Hiroshi Nakamura The University of Tokyo CONTEXT Energy Harvesting Sensor Nodes Sensor Node (capable of varying

  1. Adaptive Power Management for Energy Harvesting Sensor Nodes using Reinforcement Learning Shaswot Shresthamali, Masaaki Kondo, Hiroshi Nakamura The University of Tokyo

  2. CONTEXT

  3. Energy Harvesting Sensor Nodes Sensor Node (capable of varying the duty cycle) + Battery + http://www.libelium.com/resources/ima ges/content/products/plug- sense/details/solar_powered_photo.png Energy Harvesting Module Theoretically capable of perpetual operation 3 8 March 2017 NAKAMURA LABORATORY

  4. Challenge I Say your battery is at 75% and there is plenty of sunshine Do you ◦ Use the solar power to charge your battery only ◦ Use the solar power to charge your battery and drive the sensor node. If so, then with what proportion? 20000 800 ENERGY HARVESTED 18000 700 16000 600 14000 BATTERY 500 12000 Policy 1 10000 400 Policy 2 8000 300 Policy 3 6000 200 Energy Harvested 4000 100 2000 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 TIME 4 8 March 2017 NAKAMURA LABORATORY

  5. Challenge II Environmental Sensor Networks – P.I. Corke et. al. MOVING SENSORS https://sites.google.com/site/sarmavrudhula/home/research/energy-management-of-wireless-sensor- networks DIFFERENT ENVIRONMENTS DIFFERENT SENSORS http://www.mdpi.com/sensors/sensors-12-02175/article_deploy/html/images/sensors-12- 02175f5-1024.png 5 8 March 2017 NAKAMURA LABORATORY

  6. Challenge III BILLION AND TRILLIONS OF NODES https://bendeetech.files.wordpress.com/2015/12/theinternetofthings2-540x334.jpg?w=350&h=200&crop=1 6 8 March 2017 NAKAMURA LABORATORY

  7. Challenges II and III When dealing with TRILLIONS of sensor nodes, Customizing each node is impractical, impossible ◦ Nodes should OPTIMIZE themselves. ◦ Nodes should ADAPT to their changing environments. ENERGY HARVESTING NODES NEED TO BE ADAPTABLE SELF CALIBRATING 7 8 March 2017 NAKAMURA LABORATORY

  8. What this presentation is about To demonstrate how to overcome the challenges by using Reinforcement Learning (RL) ◦ Brief introduction to Reinforcement Learning ◦ Our approach using RL ◦ How this strategy performs compares to other methods ◦ How this strategy adapts to changing environment 8 8 March 2017 NAKAMURA LABORATORY

  9. OBJECTIVES

  10. Objectives Energy Neutral Operation (ENO) ◦ Energy consumed = Energy harvested Maximize Performance ◦ Maximize Duty Cycle Minimize Battery Downtime ◦ Battery should never drop to zero Minimize Energy Waste ◦ Battery should not overcharge Energy Waste = Energy Harvested –Energy consumed by Node –Energy to charge battery 10 8 March 2017 NAKAMURA LABORATORY

  11. SYSTEM MODEL

  12. System Model Solar Energy (Ideal) Solar Sensor Battery Panel Node Duty Cycle Adaptive Power Battery Reserve Level Manager Energy being harvested 12 8 March 2017 NAKAMURA LABORATORY

  13. REINFORCEMENT LEARNING (RL) A BRIEF INTRODUCTION

  14. What is RL Type of Machine Learning - Learns by interacting with Environment Suited for Sequential Decision Making Tasks Map situations (states) into actions – receive as much reward as possible Based on iterative process of trial and error – similar to how humans learn. ( Search and Memory ) 14 8 March 2017 NAKAMURA LABORATORY

  15. Why Reinforcement Learning By using RL, it is possible o To optimize nodes with raw high level data and minimal human input. o To adapt to changes in the environment parameters. 15 8 March 2017 NAKAMURA LABORATORY

  16. Reinforcement Learning What action should I take to accumulate total maximum OBSERVATIONS: Battery Level reward? Energy Harvested REWARD, New State ACTION: Choose Duty Cycle Environment Agent http://www.canstockphoto.com/go- (Power Manager) green-icons-concept-tree-12796260.html http://wedreamabout.com/product/bb-8-droid- the-coolest-star-wars-toy-ever 16 8 March 2017 NAKAMURA LABORATORY

  17. Reinforcement Learning The question is: WHICH ACTION TO TAKE WHEN YOU ARE IN A GIVEN STATE? EXAMPLE: Lots of sunlight | Battery at 60% Do you [1] [2]  drive the sensor node at full strength without recharging?  drive the sensor node at half strength with partial charging? [1] https://spaceplace.nasa.gov/sun-corona/en/ [2] https://handyenergy.ru/ 17 8 March 2017 NAKAMURA LABORATORY

  18. Q- Value Assign every state- action pair → Q -Value, Q(s,a) Q(s,a) means if the agent • Starts from state s • Takes action a • Q(s,a) is the total reward it can expect in the best case scenario Higher the Q-value, better the action for State X that particular state Action 3 Action 1 Action 2 Q(X,3) Q(X,1) Q(X,2) 18 8 March 2017 NAKAMURA LABORATORY

  19. Q-Learning Challenge → Determining the Q -Values for all state- action pairs. Q-table -> contains Q-Values of all possible state- action pairs Accomplished by Q-Learning Algorithm ◦ Q-values are learned by interacting with environment. ◦ Iterative Process ◦ Bootstrapping approach 19 8 March 2017 NAKAMURA LABORATORY

  20. Q-Learning Algorithm Q-Learning Algorithm ◦ Use arbitrary estimates for Q-values ◦ Use these estimates to decide on actions ◦ Update Q-table by using the rewards received ◦ Repeat until Q-value sufficiently converge 20 8 March 2017 NAKAMURA LABORATORY

  21. EXPERIMENTS ON ADAPTIVE POWER MANAGEMENT USING Q-LEARNING

  22. State Space State is defined by : • amount of battery remaining • 200 possible levels • amount of energy harvested • 5 possible levels Total possible states: 200 x 5 = 1000 22 8 March 2017 NAKAMURA LABORATORY

  23. Action Space Action: Choose duty cycle of the sensor node 𝐵 = 𝑏 𝑢 𝑙 ∈ 10%, 20%, 30% … .100% 10% 50 mW 50% 250 mW 100% 500 mW 23 8 March 2017 NAKAMURA LABORATORY

  24. Reward Function The reward depends on: • Distance from energy neutrality at time t k ∆𝑓 𝑜𝑓𝑣𝑢𝑠𝑏𝑚 𝑢 𝑙 = 𝑓 ℎ𝑏𝑠𝑤𝑓𝑡𝑢 𝑢 𝑙 − 𝑓 𝑜𝑝𝑒𝑓 𝑢 𝑙 • Amount of battery remaining 24 8 March 2017 NAKAMURA LABORATORY

  25. RESULTS

  26. Training and Testing Training: Tokyo (2000 to 2009) Testing : Tokyo (2010) ◦ Compare it with other methods. ◦ Adaptation to diurnal and seasonal variations. ◦ Greedy and ε -greedy Implementations 26 8 March 2017 NAKAMURA LABORATORY

  27. Comparison with other methods 100% Higher Efficiency 90% 𝐵𝑑𝑢𝑣𝑏𝑚 𝐸𝑣𝑢𝑧 𝐷𝑧𝑑𝑚𝑓 80% 70% 𝐵𝑑ℎ𝑗𝑓𝑤𝑏𝑐𝑚𝑓 𝑁𝑏𝑦𝑗𝑛𝑣𝑛 𝐸𝑣𝑢𝑧 𝐷𝑧𝑑𝑚𝑓 60% 50% Efficiency(%) 40% Energy Wasted(%) 30% 20% 10% 𝑈𝑝𝑢𝑏𝑚 𝐹𝑜𝑓𝑠𝑕𝑧 𝑋𝑏𝑡𝑢𝑓𝑒 Lower Waste 0% 𝑈𝑝𝑢𝑏𝑚 𝐹𝑜𝑓𝑠𝑕𝑧 𝐼𝑏𝑠𝑤𝑓𝑡𝑢𝑓𝑒 Naïve Kansal Our method using RL 𝐹𝑜𝑓𝑠𝑕𝑧 𝑋𝑏𝑡𝑢𝑓 = 𝐹𝑜𝑓𝑠𝑕𝑧 𝐼𝑏𝑠𝑤𝑓𝑡𝑢𝑓𝑒 Fix duty cycle for Duty Cycle is −𝑂𝑝𝑒𝑓 𝐹𝑜𝑓𝑠𝑕𝑧 present day by proportional to − 𝐷ℎ𝑏𝑠𝑕𝑗𝑜𝑕 𝐹𝑜𝑓𝑠𝑕𝑧 predicting total battery level energy for next day 27 8 March 2017 NAKAMURA LABORATORY

  28. ADAPTATION TO SEASONAL CHANGES 28 8 March 2017 NAKAMURA LABORATORY

  29. Performance in Summer High Duty Cycle even during the night 29 8 March 2017 NAKAMURA LABORATORY

  30. Performance in Winter Lower duty cycle during the night Duty Cycle (%) 100% Harvested Energy (%) 80% 60% 40% 20% 0% 1360 1400 1440 1480 EPOCH Battery (%) 30 8 March 2017 NAKAMURA LABORATORY

  31. ADAPTATION TO CHANGE IN LOCATION 31 8 March 2017 NAKAMURA LABORATORY

  32. Implementation: ε -greedy approach Perfect Q-convergence takes too long. Instead, use ε -greedy approach with non-converged Q-table. ε -greedy approach: ◦ Take the best action by default. ◦ Take a random action with probability ε. ◦ Increasing ε → Exploration ◦ Decreasing ε → Exploitation 32 8 March 2017 NAKAMURA LABORATORY

  33. Adaptation to change in climate • Wakkanai (very little sunshine) • Compare between • a greedy approach (Offline) and • an  -greedy approach (Online). • Training: 2000-2009 Tokyo • Testing: 2010 Wakkanai 33 8 March 2017 NAKAMURA LABORATORY

  34. Adaptation to change in location Total number of times the 65 battery was completely 64 Average Duty Cycle (%) exhausted 63 Greedy (Non adaptive) 62 61 (8) 60 59 (14) (14) 58  -greedy (adaptive) 57 With  -greedy implementation, the agent adapts to the 56 environment and minimizes instances of battery exhaustion. 55 2010 2011 2012 2013 2014 2015 Wakkanai Offline Wakkanai Online 34 8 March 2017 NAKAMURA LABORATORY

  35. With and Without Forecast Information 35 8 March 2017 NAKAMURA LABORATORY

  36. CONCLUSION

  37. CONCLUSION • Proposed system is able to meet objectives of • Energy neutrality • Maximizing performance • Exceeds the performance of other schemes • Capable of adaptation • Inclusion of weather forecast results in smarter operation 37 8 March 2017 NAKAMURA LABORATORY

  38. THANK YOU FOR LISTENING ANY COMMENTS OR QUESTIONS ARE WELCOME

Recommend

RL Overview of topics About Reinforcement Learning The Reinforcement Learning Problem

RL Overview of topics About Reinforcement Learning The Reinforcement Learning Problem

Introduction to Reinforcement Learning RL Overview of topics About Reinforcement Learning The Reinforcement Learning Problem Inside an RL agent Temporal difference learning Many faces of Reinforcement Learning What is

551 views • 35 slides
Reinforcement Learning Reinforcement Learning Reinforcement Learning in a nutshell g Imagine

Reinforcement Learning Reinforcement Learning Reinforcement Learning in a nutshell g Imagine

Reinforcement Learning Reinforcement Learning Reinforcement Learning in a nutshell g Imagine playing a new game whose rules you dont know; after a hundred or so moves your don t know; after a hundred or so moves, your opponent announces, You

511 views • 30 slides
Reinforcement Learning You can think of supervised learning as the teacher providing answers

Reinforcement Learning You can think of supervised learning as the teacher providing answers

Reinforcement Learning You can think of supervised learning as the teacher providing answers (the class labels) Reinforcement learning: In reinforcement learning, the agent learns Markov Decision Processes based on a punishment/reward

89 views • 6 slides
Examples of Reinforcement Learning Robocup Soccer Teams Stone & Veloso, Reidmiller et al.

Examples of Reinforcement Learning Robocup Soccer Teams Stone & Veloso, Reidmiller et al.

Reinforcement Learning Slides by Rich Sutton Mods by Dan Lizotte Refer to Reinforcement Learning: An Introduction by Sutton and Barto Alpaydin Chapter 16 Up until now we have been Supervised Learning Classifying, mostly

891 views • 31 slides
CS885 Reinforcement Learning Module 2: June 6, 2020 Maximum Entropy Reinforcement Learning

CS885 Reinforcement Learning Module 2: June 6, 2020 Maximum Entropy Reinforcement Learning

CS885 Reinforcement Learning Module 2: June 6, 2020 Maximum Entropy Reinforcement Learning Haarnoja, Tang et al. (2017) Reinforcement Learning with Deep Energy Based Policies, ICML . Haarnoja, Zhou et al. (2018) Soft Actor-Critic: Off-Policy

682 views • 24 slides
Reinforcement Learning Reinforcement Learning Now that you know a little about Optimal Control

Reinforcement Learning Reinforcement Learning Now that you know a little about Optimal Control

Reinforcement Learning Reinforcement Learning Now that you know a little about Optimal Control Theory, you actually have some knowledge in RL. RL shares the overall goal with OCT: solving for a control policy such that the cumulative cost

610 views • 22 slides
Reinforcement Learning CS 4100: Artificial Intelligence Reinforcement Learning II Still

Reinforcement Learning CS 4100: Artificial Intelligence Reinforcement Learning II Still

Reinforcement Learning CS 4100: Artificial Intelligence Reinforcement Learning II Still assume a Marko kov decision process (MDP): A se set o of st states s s S A se set o of a actions ( s (per st state) A A mo model

96 views • 5 slides
R i f R i f Reinforcement Learning III Reinforcement Learning III t L t L i i III III Dec

R i f R i f Reinforcement Learning III Reinforcement Learning III t L t L i i III III Dec

R i f R i f Reinforcement Learning III Reinforcement Learning III t L t L i i III III Dec 03 2008 1 Large State Spaces h When a problem has a large state space we can not longer represent the U or Q functions as explicit tables explicit

436 views • 18 slides
Reinforcement Learning UMaine COS 470/570 Introduction to AI Why reinforcement learning?

Reinforcement Learning UMaine COS 470/570 Introduction to AI Why reinforcement learning?

Reinforcement Learning<br/><br/> 4/25/19, 8*06 PM Reinforcement Learning<br/><br/> 4/25/19, 8*06 PM Reinforcement Learning UMaine COS 470/570 Introduction to AI Why reinforcement learning? Spring 2019 Created:

371 views • 15 slides
7. Motor Control and Reinforcement Learning Outline A. Action Selection and Reinforcement B.

7. Motor Control and Reinforcement Learning Outline A. Action Selection and Reinforcement B.

7. Motor Control and Reinforcement Learning Outline A. Action Selection and Reinforcement B. Temporal Difference Reinforcement Learning C. PVLV Model D. Cerebellum and Error-driven Learning 2/23/18 COSC 494/594 CCN 2 Sensory-Motor Loop

790 views • 56 slides
Deep Reinforcement Learning [Mastering the Game of Go with Deep Reinforcement Learning and Tree

Deep Reinforcement Learning [Mastering the Game of Go with Deep Reinforcement Learning and Tree

Deep Reinforcement Learning [Mastering the Game of Go with Deep Reinforcement Learning and Tree Search, Nature 2016] CS 486/686 University of Waterloo Lecture 21: July 12, 2017 Outline AlphaGo Supervised Learning of Policy Networks

540 views • 15 slides
Reinforcement Learning CS 4100: Artificial Intelligence Reinforcement Learning Ja Jan-Wi

Reinforcement Learning CS 4100: Artificial Intelligence Reinforcement Learning Ja Jan-Wi

Reinforcement Learning CS 4100: Artificial Intelligence Reinforcement Learning Ja Jan-Wi Willem van de Meent Northeastern University [These slides were created by Dan Klein and Pieter Abbeel for CS188 Intro to AI at UC Berkeley. All CS188

145 views • 10 slides
Reinforcement Learning AIMA Chapters: 21.1, 21.2, 21.3. Sutton and Barto, Reinforcement Learning:

Reinforcement Learning AIMA Chapters: 21.1, 21.2, 21.3. Sutton and Barto, Reinforcement Learning:

Reinforcement Learning Reinforcement Learning AIMA Chapters: 21.1, 21.2, 21.3. Sutton and Barto, Reinforcement Learning: an Introduction, 2nd Edition: Chapters 6 (6.1 6.5) Outline Reinforcement Learning Reinforcement Learning: the

587 views • 27 slides
Module 11 Introduction to Reinforcement Learning CS 886 Sequential Decision Making and

Module 11 Introduction to Reinforcement Learning CS 886 Sequential Decision Making and

Module 11 Introduction to Reinforcement Learning CS 886 Sequential Decision Making and Reinforcement Learning University of Waterloo Machine Learning Supervised Learning Teacher tells learner what to remember Reinforcement Learning

401 views • 29 slides
1 Pure Reinforcement Learning vs. Reinforcement Learning Monte-Carlo Planning No knowledge

1 Pure Reinforcement Learning vs. Reinforcement Learning Monte-Carlo Planning No knowledge

CSE 573: Artificial Intelligence Todo Reinforcement Learning Add simulations from 473 Dan Weld Add UCB bound (cut bolzman & constant epsilon Add snazzy videos (pendulum, zico kolter See http://www See http://www-

462 views • 11 slides
Safe Reinforcement Learning Philip S. Thomas Stanford CS234: Reinforcement Learning, Guest

Safe Reinforcement Learning Philip S. Thomas Stanford CS234: Reinforcement Learning, Guest

Safe Reinforcement Learning Philip S. Thomas Stanford CS234: Reinforcement Learning, Guest Lecture May 24, 2017 Lecture overview What makes a reinforcement learning algorithm safe ? Notation Creating a safe reinforcement learning

1.42k views • 88 slides
Reinforcement Learning Framework Reinforcement Learning Rewards, Returns Lectures 4 and 5

Reinforcement Learning Framework Reinforcement Learning Rewards, Returns Lectures 4 and 5

1 Reinforcement Learning Framework Reinforcement Learning Rewards, Returns Lectures 4 and 5 Environment Dynamics Gillian Hayes Components of a Problem 18th January 2007 Values and Action Values, V and Q Optimal Policies

109 views • 7 slides
Introduction to Reinforcement Learning Bayesian Methods in Reinforcement Learning ICML 2007

Introduction to Reinforcement Learning Bayesian Methods in Reinforcement Learning ICML 2007

Introduction to Reinforcement Learning Bayesian Methods in Reinforcement Learning ICML 2007 sequential decision making under uncertainty ? How Can I ... ? Move around in the physical world (e.g. driving, navigation) Play and win a game

519 views • 21 slides
Neural Combinatorial Optimization With Reinforcement Learning CS885 Reinforcement Learning Paper

Neural Combinatorial Optimization With Reinforcement Learning CS885 Reinforcement Learning Paper

Neural Combinatorial Optimization With Reinforcement Learning CS885 Reinforcement Learning Paper by Bello, I., Pham, H., Le, Q. V., Norouzi, M., & Bengio, S. (2016) Presented by Yan Shi Outline 1. Introduction 2. Background 3. Algorithms

966 views • 27 slides
Todays Outline Reinforcement Learning Dan Weld Reinforcement Learning Q-value

Todays Outline Reinforcement Learning Dan Weld Reinforcement Learning Q-value

5/7/2012 CSE 473: Artificial Intelligence Todays Outline Reinforcement Learning Dan Weld Reinforcement Learning Q-value iteration Q-learning Exploration / exploitation Linear function approximation Linear function

159 views • 12 slides
CSC2621 Topics in Robotics Reinforcement Learning in Robotics Week 11: Hierarchical Reinforcement

CSC2621 Topics in Robotics Reinforcement Learning in Robotics Week 11: Hierarchical Reinforcement

CSC2621 Topics in Robotics Reinforcement Learning in Robotics Week 11: Hierarchical Reinforcement Learning Animesh Garg Between MDPs and semi-MDPs: A framework for temporal abstraction in reinforcement learning Richard S. Sutton , Doina

2.18k views • 29 slides
Reinforcement Learning (RL) CE-717: Machine Learning Sharif University of Technology M.

Reinforcement Learning (RL) CE-717: Machine Learning Sharif University of Technology M.

Reinforcement Learning (RL) CE-717: Machine Learning Sharif University of Technology M. Soleymani Fall 2018 Most slides have been taken from Klein and Abdeel, CS188, UC Berkeley. Reinforcement Learning (RL) Learning as a result of

1.45k views • 133 slides
1 Deep Reinforcement Learning Qianqian Li, Nayeon Koong, Langtian He What is deep reinforcement

1 Deep Reinforcement Learning Qianqian Li, Nayeon Koong, Langtian He What is deep reinforcement

1 Deep Reinforcement Learning Qianqian Li, Nayeon Koong, Langtian He What is deep reinforcement learning? Agent/Actor + Action + Environment + State + Reward How does reinforcement learning work?

792 views • 31 slides
Machine Learning for NLP Reinforcement learning Aurlie Herbelot 2019 Centre for Mind/Brain

Machine Learning for NLP Reinforcement learning Aurlie Herbelot 2019 Centre for Mind/Brain

Machine Learning for NLP Reinforcement learning Aurlie Herbelot 2019 Centre for Mind/Brain Sciences University of Trento 1 Introduction 2 Reinforcement learning: intuition Reinforcement learning: like learning to ride a bicycle.

1.18k views • 58 slides

More recommend