12
Jules Jaffe Marine Physical Lab Scripps Institution of Oceanography U.C. San Diego Underwater Robot Swarms and the Challenges for UW Communication

Underwater Robot Swarms and the Challenges for UW

  • Upload
    others

  • View
    3

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Underwater Robot Swarms and the Challenges for UW

Jules JaffeMarine Physical Lab

Scripps Institution of OceanographyU.C. San Diego

Underwater Robot Swarms and the

Challenges for UW Communication

Page 2: Underwater Robot Swarms and the Challenges for UW

Our Solution: A Swarm of Underwater Robots

The Age-Old Problem in Sampling the Ocean:

• The Ocean is a dynamic environment in both space and time.. How can we sample it?

• Walter Munk: “The 20th century will go down as the century in which we under-sampled the ocean”

Page 3: Underwater Robot Swarms and the Challenges for UW

Autonomous Underwater Explorer:AUE

1.5 literTemperature, pressure, acoustic sensors2 modes of buoyancy controlDeployments up to week

Page 4: Underwater Robot Swarms and the Challenges for UW

Deployment Oct. 2, 2013:16 AUEs

5 GPS moored surface pingers

Page 5: Underwater Robot Swarms and the Challenges for UW

Aue Tracks

Page 6: Underwater Robot Swarms and the Challenges for UW

AUE data

Trough

Crest

TemperatureAnomaly

Page 7: Underwater Robot Swarms and the Challenges for UW

A few curves to keep in mind

Hardware:Computer processing and mechanical components

Software:Cost of computer and mechanical hardware

Environment:

Absorption and Spreading

Page 8: Underwater Robot Swarms and the Challenges for UW

Selection of Habitat

Benthic Habitats Pelagic

Page 9: Underwater Robot Swarms and the Challenges for UW

Examples of the need for UW comms. and the importance for adaptive sampling

• Navigation and concurrent estimation of sampling for a variety of studies– Bottom mapping (esp. structure from motion)– Acoustic localization (for longer wavelengths)– Spatial sampling for toxic algal blooms (for tracking and

prediction) to follow gradients (spatial structure is important!!)

• Knowledge of sensor recordings for swarm sensing– Optical (microscopes, laser LIDAR systems)– Acoustical (passive systems)– Chemical (toxic chemical sensors)

Jaffe Lab Activities

Page 10: Underwater Robot Swarms and the Challenges for UW

The lure of in situ classification

• Advanced data product:

Neural net architecture can be run on in situ system.

Resultant data can then be simply genera and easily stored or shared via satellite

Page 11: Underwater Robot Swarms and the Challenges for UW

We now have 800 million imagesnext step….

Convolutional Neural Net Architecture

92% success rate in classifying certain species: Eric Orenstein (Jaffe Lab)

Page 12: Underwater Robot Swarms and the Challenges for UW

Human/Robot Human/Robot Underwater Communication via AR/VR

• Expert on sub or ship to diver or robot on site for inspection/repair

• Robot to human communication for remote inspection

Jaffe Lab new 2-Pi Cam 18 Raspberry Pis : 10k x 6k image