inspecting large irrigation networks in the indus basin
play

Inspecting Large Irrigation Networks in the Indus Basin: Challenges - PowerPoint PPT Presentation

Inspecting Large Irrigation Networks in the Indus Basin: Challenges and Prospects Abubakr Muhammad, PhD Assistant Professor of Electrical Engineering Director, Laboratory for Cyber Physical Networks and Systems LUMS School of Science &


  1. Inspecting Large Irrigation Networks in the Indus Basin: Challenges and Prospects Abubakr Muhammad, PhD Assistant Professor of Electrical Engineering Director, Laboratory for Cyber Physical Networks and Systems LUMS School of Science & Engineering Lahore, Pakistan IEEE RAS Technical Committee on Agricultural Robotics. Webinar. Feb 27, 2015

  2. Acknowledgements • Joint work with – PhS & MS Students : Syed M. Abbas, Hamza Anwar,Talha Manzoor, Mudassir Khan – Collaborators : Prof Karsten Berns, RRLab, Univ of Kaiserslautern, Germany • Funding – LUMS Faculty Initiative Fund (2014) – DAAD Grant: Robotic Profiling of Waterways. RoPWat (2014-15) • General Support – Punjab Irrigation Department (PID) – International Water Management Institute (IWMI)

  3. Outline • Motivation and context • Smart water grids philosophy • Siltation of canals and rivers • Traditional canal cleaning process ( یئافص لھب ) • Proposed solution • Towards performance limits • Conclusions and outlook

  4. LUMS Overview • Pakistan’s 2 nd Private University – Founded in 1985 – Non-profit organization – 100 acre campus • 2300 Student Body – Approx. 1800 undergraduates & 500 graduate students – 35% women – 60% resident on-campus • 120 faculty members – PhD’s from Stanford, MIT, Yale, Oxford, Cambridge, Imperial 2/27/2015

  5. Cyber Physical Networks & Systems (CYPHYNETS) Lab Est. 2008 • Director PhD Georgia Tech, Postdocs (Penn, McGill) • 2 Jr. Faculty PhD Warick (control), PhD Siegen (robotics) • 3 PhD students • 6 MS students • 4 Full-time Research Assistants • 1 Lab Engineer • 2 Lab Technicians Areas • Water Networks & Hydro-informatics • Agricultural Automation & Robotics • Cyber-Physical Systems http://cyphynets.lums.edu.pk

  6. Graduate Training / Outreach New Graduate Courses at LUMS • EE-561, Digital Control Systems • EE-662, Parameter & State Estimation • EE-562, Robot Motion Planning • CMPE-633c, Geometric Mechanics & Control • CMPE-633b, Robot Dynamics & Control • EE-565, Mobile Robots Field & Assistive Robotics Conferences / Workshops Organized • International Workshops on Intelligent Water Grids (IWG), 2013 – Symposium on Pakistan’s Water Futures – Workshop on Sensing and and Control for Water Networks – Workshop on Hydro-informatics – Mini-course on System Identification of Irrigation Channels • Workshop Series on Field and Assistive Robotics (2011-2015) – 1 st , 3 rd , 5 th , 7 th , 9 th in Lahore, Pakistan – 2 nd , 4 th , 6 th , 8 th in Dagstuhl / Kaiserslautern, Germany NMO Secretariat, IIASA-Pakistan collaboration (2012-) IEEE CSS Society Pakistan Chapter (2011-)

  7. Agricultural Productivity Loss Main reason is water!

  8. Motivation / Concerns Vulnerability sources Source. UNEP South Asia report, 2008 Annual canal diversions and sea escapage Flow reduction due to climate change 8

  9. Managing the World’s Largest Irrigation Network 90,000 Km of watercourses 3 reservoirs, 23 barrages 45 canal commands 36 million acre irrigated area System Efficiency: extremely poor!

  10. Wireless connectivity A Networked Smart Water Grid Embedded controller Flow Measurements Gate control

  11. A Networked Smart Water Grid Cyber Physical Systems / Internet of Things perspectives • Physical elements: rivers, watercourses, barrages, weirs, gates, pumps • Cyber elements : sensors, controllers, comm., services

  12. Smart Water Metering: E. Sadqiya Hakra Br. Canal Command LUMS-IWMI-Punjab Irrigation Dept. Collaboration (2012-14) Project Site: 17 Distributaries in Bahawalnagar. System Architecture (above). Field installations (below). Laboratory for Cyber Physical Networks & Systems Dept. of Electrical Engineering, LUMS

  13. Identification & Control of Irrigation Channels (2011-14) Abstraction dy ( ) t 3 3      i 1 c h 2 ( t ) c h 2 ( ). t   i in , i i 1, out i 1 dt Laboratory for Cyber Physical Networks & Systems Dept. of Electrical Engineering, LUMS

  14. Autonomous Land Vehicles for Demining & Agriculture ALVeDA & MDRD (2010-2013) Collaboration: RRLab, TU Kaiserslautern Funding: DAAD, LUMS, National Instruments Objective: Push performance limits with low- Field Experiments: Channel mapping in Lahore (left). cost vision sensors and simple mechatronics. Scanning a minefield in Beirut (right). Robot Vision: Terrain Classification, RGB-D & Monocular SLAM, Visual Servoing, Soil Estimation in a Bucket Excavator.

  15. What is Silt? • Particles of earth, slightly larger than clay and slightly smaller than sand. • It is composed of quartz and feldspar. • Occurs as soil, as suspended sediment in a surface water body, or as soil deposited at the bottom of a waterway.

  16. Silt in Waterways • Slow moving water deposits silts on the canal bed. • Reduces channel carrying capacity. • Outlets draw more water than their allotted share due to raised water levels.

  17. Silt Removal in Punjab یئافص لھب • Punjab Irrigation Department first started large-scale de-silting of canals during 1990s. • Since then, PIDA (Punjab Irrigation and Drainage Authority) conducts this campaign annually to clean its canals of silt and other garbage.

  18. Inspection of Canal Water Beds • Bed levels observed every 1000 ft. • Silt depth of more than 6 inches is marked for removal.

  19. Silt Quantity Estimation benchmark B.S F.S change point 1000 ft B.S silt depth silted water bed I.S

  20. Silt Quantity Estimation • X-section area of silt top width – bed width + top width × silt depth 2 • top width = 𝑔(silt depth) – depending on canal geometry bed width silt depth • X-section area calculated after every 1000 ft

  21. Silt Quantity Estimation Estimate of silt clearance benchmark B.S • Water bed divided into patches of 1000 ft length F.S – Area calculated at both ends of a single patch (A 1 and A 2 ). change point 1000 ft • B.S Silt volume (cft) – A 1 + A 2 × 1000 2 silted water bed I.S

  22. Various Methods of Silt Removal

  23. How Can We Automate This Process? • Extent of the canal system (40,000km+) and the tight time-lines (< 3 weeks) • Makes it feasible to consider an automated solution as a scalable and economic alternative to manual operation. • Cleaning automation is too ambitious. • Perhaps, we can start by profiling / inspection only?

  24. Challenges • While cleaning, the original shape of the cross-section and bed-slope must be restored. • What is the “true profile”? • Selection of map granularity to measure the deviation from the true profile. • A way to deploy and recover the profiling system. • Profiler must not obstruct the canal operation. • Profiler must have the capability to negotiate narrow passages and soft muddy beds. • Solution should be fast and easily scalable • Minimal specialist training.

  25. Proposed solution • An autonomous aerial inspection system

  26. Robotic Profiling for Clearing Watercourses (RoPWat) • Development of semi-autonomous robotic system for profiling watercourses • 3D perception system to be deployed on commercial vehicles • UAV for monitoring state of the canal • Simulation of the cleaning process • Simulation of vehicle for cleaning • Sponsored by LUMS Faculty Initiative Fund (FIF) and later by DAAD.

  27. Proposed Solution

  28. CMU Riverine Mapping Project Riverine reconnaissance with a low-Flying intelligent UAS. An online state estimation system A self supervised vision based river detector. A scrolling incremental distance transform algorithm. A novel scanning LADAR configuration & analysis of measurement data. Scherer et al. River mapping from a flying robot: state estimation, river detection, and obstacle mapping, Autonomous Robots, Vol. 32, No. 5, May, 2012.

  29. What is Achievable? • Analysis on 1D version of the aerial inspection problem

  30. Guassian Process (GP) Regression Effect of variation in sampling interval

  31. Guassian Process (GP) Regression Effect of sensor noise variation

  32. Incorporating Localization error • GP regression with noisy remains exactly the same • except that the covariance function is instead of where

  33. Incorporating Localization error

  34. Verifying Results on a Test Rig

  35. What about Estimated Silt Volume? • Area under the (noisy) curve, with mean and variance Error bounds:

  36. Analysis • Localization error matters “more” than sensor precision • Translates to strict requirements on positioning / elevation sensors and good algorithms • The point of the analysis is not to construct new algorithms but to find what is achievable. • This really challenges the state of the art in – analytical methods – systems engineering • Framework is generic: analysis carries over to 3D case.

  37. Progress 2014 : Team building, analysis, framework 2015 : System building, testing of algorithms 2016 : Field Trials

  38. Some Basic Questions … • Why Automation in developing countries like Pakistan? – Devolution of governance – Ensuring rights – Conflict resolution Entitlements Participation • Major challenges – Natural resources – Food and Agriculture – Critical infra-structures Accountability – Security – Healthcare

Recommend


More recommend