Marine Robotics

1
Marine Robotics
Ari Requicha Carl Oberg Gaurav Sukhatme Beth
Stauffer David Caron Amit Dhariwal Deborah
Estrin Eric Shieh Chongwu Zhou Bin Zhang
2
Fundamental biological questions (driving forces
behind the work)

• What are the spatial/temporal distributions in
  nature of microorganisms of interest to ecosystem
  or human health?
• What factors (environmental or otherwise) explain
  the distributions of these microorganisms?
• How can we most effectively and efficiently
  address these issues?
• Not simply a question of more data (although
  that also helps). CENS approach will provide the
  ability to characterize the chemical, physical,
  biological community on temporal and spatial
  scales that are pertinent to the organisms.

3
Typical Scales of Ocean Sensing/Sampling
3 meter discus buoy (Autonomous Aerosol Sampler
Sholkovitz et al., WHOI)
CTD rosette (Ross Sea, 1999)
4
What are the problems with these approaches?

Scale of Measurements Six orders of magnitude
between scales of observation and scales of
microbial interactions (we do meters - need to
do micrometers) Spatiotemporal Coverage One
ocean, one instrument approach (workable sea
state, instrument cost) Instrument Capability
(biosensor limitation) Chlorophyll (phytoplan
kton only not specific)


5
Examining small-to-microscale distributions of
aquatic microorganisms. What factor act to
establish these patterns? What do the patterns
mean? How can we investigate these features?
East Sound, WA (Donaghay) (http//www.gso.uri.edu/
criticalscales/about/index.html)
6
Validation

• Create a thermocline in the tank
• Place sensors so as to find the thermocline
  accurately and with a small part of the total
  network
• Acquire water samples at strategically-located
  points
• Determine microorganism content (offline)
• Analyze the data to correlate algal behavior with
  thermocline.

7
Thermocline Localization

• The search space is 1D, and is divided into
  regions
• Each node explores one region with binary search
• Each node tries to persuade others that the
  thermocline is within its search region
• A process of data aggregation is enacted on the
  route from each node to the edge node (the one at
  the top of the tank)

8
The Experimental Setup

• A tethered system of small robots with radios
  (limited range)
• A small number of mobile underwater robots
• Initially focus on temperature measurements
• Collect water samples for offline analysis

9
Improving Energy Efficiency with a Data Mule

• Motivation
• After first several steps, most nodes become
  inactive
• However, many of them have to be awake to forward
  the messages from the active nodes to root
• Solution
• Create shortcuts from active nodes to root with a
  robotic submarine
• Assumption
• Submarine can be recharged

10
The Goal
Pattern-Triggered Data Collection
Biosensors Correlatiing small-scale spatial
distributions with chemical and physical structure
Sensor network Defining small-scale physical
structure
11
Improving regional scale characterization of
temporal/spatial distributions
How to characterize and sample features of
biological interest in aquatic ecosystems?
What we do now. (large uniform grid,
occupied sporadically)
12
What we want to do. Aggregation of sampling
nodes at feature(s) of interest based on
sensing-directed movement
13
Marine Microorganism Monitoring Subsystems
14
Research Implications

• High spatial density (cm-mm), small sensors
  (cm-nm) of small size and limited capability.
• Sensor-coordinated mobility and actuation,
  -deploy sensors and sample collectors where
  and when they are needed.
• -data processing inside the network, e.g., to
  direct sensing and/or sampling.
• Rapid microorganism identification in-situ ? New
  sensing techniques

15
Adaptive Sampling

• How to aggregate sensors where they are needed
• Within CENS there are several parallel strategies
  being pursued
• Statistically rigorous adaptive sampling (NIMS)
• Event-aware (triggered) sampling (NIMS)
• Bacterial motion-inspired swarming

16
Bacterial Motion

• Bacteria move by interspersing propulsion with
  random turns (tumbles)
• Taxis is achieved by varying the duration of
  propulsion
• This is in effect, a biased random walk

17
Summary/Conclusions

• Lab-based experimental application of
  sensor-actuated microbial sampling.
• Movement from lab-based ground-truthing into the
  real world.
• Small-scale coordinated sensor nodes (mobile
  stationary working together)
• Multiple mobility modes
• True biosensor development