1 / 15

CyberSEES : Type 2: Towards Sustainable Aquatic Ecosystems: A New Adaptive Sampling and Data-Enabled Monitoring and Mode

CyberSEES : Type 2: Towards Sustainable Aquatic Ecosystems: A New Adaptive Sampling and Data-Enabled Monitoring and Modeling Framework. Xiaobo Tan (PI) , Electrical & Computer Engineering, MSU Elena Litchman , Kellogg Biological Station/Zoology, MSU

holland
Download Presentation

CyberSEES : Type 2: Towards Sustainable Aquatic Ecosystems: A New Adaptive Sampling and Data-Enabled Monitoring and Mode

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. CyberSEES: Type 2: Towards Sustainable Aquatic Ecosystems: A New Adaptive Sampling and Data-Enabled Monitoring and Modeling Framework Xiaobo Tan (PI), Electrical & Computer Engineering, MSU Elena Litchman, Kellogg Biological Station/Zoology, MSU ManthaPhanikumar, Civil & Environmental Engineering, MSU Hayder Radha, Electrical & Computer Engineering, MSU Guoliang Xing, Computer Science & Engineering, MSU

  2. Water Sustainability: A Grand Challenge • Water sustainability: vital to sustaining life, health, and economic development; a global challenge • Water resources and aquatic ecosystems are under increasing stresses Harmful algal bloom Japan Tsunami Gulf oil spill Source for all images: National Geographic • Monitoring and understanding water environments enables informed decision, holding the key to sustainability

  3. Major Technical Challenge • Spatiotemporal heterogeneity in massive water bodies • Some current practices in water sampling Credit: GLERL Credit: WHOI Cannot capture spatial dynamics Cannot capture temporal dynamics Autonomous Underwater Vehicles (AUVs) and underwater gliders Source: UW/APL Source: AUVAC.org Source: www.awi.de Costly; poor portability; limited to ocean-like environments

  4. Robotic Fish • Generate motion through fin and/or body movement: Light weight, low cost, highly maneuverable Univ. of Washington MIT (Triantafyllou) Univ. of Essex, UK • MSU robotic fish: Sharp focus on environmental sensing • Constant actuation required for movement: limited operational time

  5. Breakthrough: Gliding Robotic Fish • Hybrid of (miniature) underwater glider [energy-efficient] and robotic fish [highly maneuverable]; • Expect to achieve energetic autonomy by harvesting solar power: indefinitely long continuous field operation Sampling harmful algal blooms in lake (video) Gliding in pool (video) Sampling oil in river (video) • Potential game changer in underwater sensing: long-duration, low-cost, portable, versatile environments, amenable to networked “school”

  6. Samples & model parameters Samples Gliding robotic fish sensor network Ecosystem monitoring, prediction & action Mechanistic modeling & simulation Data-field reconstruction Data fields A New Framework for Monitoring and Modeling

  7. Harmful Algal Blooms (HABs) as Case Study • World-wide proliferation: A major challenge threatening sustainability. • Impacting ecosystems, human health, and economic development • High spatiotemporal heterogeneity, complex biophysical interactions: ideal proof of concept Source: TX Park and Wildlife Source: National Geographic Gull Lake, Kellogg Biological Station

  8. Partnership Tan (ECE) Underwater robotics Litchman (KBS) Aquatic ecology Xing (CSE) Sensor networks Radha (ECE) Compressed sensing Phanikumar (CEE) Biophysical modeling

  9. Research Thrusts Focus: Addressing critical, fundamental problems at the interfaces Information-driven adaptive sampling, to maximize information gain & minimize energy expenditure: information theory, gliding maneuvers Communication strategy and in-network processing, for robust networking and sub-model parameter estimation:rendezvous High-resolution reconstruction of data fields under limited data samples (<1%), for monitoring, model identification, and path planning: tensor-based signal sparsification Data and mobile network-facilitated mechanistic modeling, for understanding of biophysical mechanisms and for effective forecast: sub-models, patch dynamics Cyber and data-enabled HAB studies, for quantitative understanding and precise forecast: spatiotemporal data, accurate model

  10. Backup Slides: More details on technical thrusts

  11. 1. Information-driven Adaptive Sampling • Objective: Path planning and control schemes for gliding robotic fish that maximize information gain and minimize energy expenditure • Challenges: Complex, nonlinear robot dynamics; significant environmental disturbances (for example, waves and currents) • Approaches [Tan, Radha, Xing]: • Information-theoretic tools, along with reconstructed data fields, will be used to determine the next sampling region • Paths for individual robots are planned with energy-efficient gliding-based rectilinear and spiral motions • Hybrid open/closed-loop control for tracking the planned path: closed-loop control activated only when deviation from path is big

  12. 2. Communication Strategy and In-Network Processing • Objective: A robust mobile network for sampling, communication, coordination, and in-network sub-model estimation • Challenges: Difficulty in underwater communication; uncertainty in robot mobility control; limited onboard processing power • Approaches [Xing, Tan, Phanikumar]: • Rendezvous-based, delay-tolerant communication: robots only “meet” and “talk” in rendezvous regions, designated based on reconstructed data fields and model prediction • Collaborative in-network sub-model parameter estimation: schemes for discretizing PDE sub-models; architecture for fusing information (local samples or parameter estimates) from individual robots

  13. 3. Sparsification-based Reconstruction of High-Resolution Data Fields • Objective: High-resolution data fields for monitoring, parameter estimation of mechanistic model, and path planning for robots • Challenge: Despite unprecedented spatiotemporal coverage by the robotic sensor network, sampled data will be < 1% of full data fields corresponding to the relevant spatiotemporal scale • Approach [Radha, Phanikumar, Tan]: • “Sparsification”: recovering all entries of original signal based on (much) fewer measurements • Generalization of compressed sensing (one dimensional, 1D) and low-rank matrix completion (2D) to low-rank tensor completion (3D/4D) • Use of “virtual samples “ generated by mechanistic models to reduce the required amount of sampled data Original with limited sampling Reconstructed

  14. 4. Data and Mobile Network-facilitated Mechanistic Modeling • Objective: Accurate mechanistic model capturing coupled physical-biological processes for fundamental understanding and effective forecast • Challenge: Over 100 parameters that can vary with location and time • Approach [Phanikumar, Litchman, Xing]: • “Sub-models”: Much simpler models sharing some parameters of the full model (e.g., thermal diffusion coefficients, light extinction coefficients) , which are estimated directly by the robotic sensor network • “Patch dynamics”: Using spatiotemporally localized data samples and (reconstructed) data fields to identify model parameters with high confidence and extract fine details Patch dynamics

  15. 5. Cyber and Data-enabled Studies of HABs • Objective: Quantitative understanding and precise forecast of HAB development • Challenges: Lack of spatiotemporal data; complex coupling between a number of factors (e.g., temperature, solar radiation, nutrients, water stratification) and HAB dynamics • Approaches [Litchman and all other PIs]: • Characterization of spatiotemporal heterogeneity of HABs with sampled data, reconstructed data fields, and modeling • In-depth understanding of mechanisms for HAB development based on accurate biophysical model with quantitative thresholds, and thus significantly improving HAB forecast and intervention

More Related