Header image  

YIP: Generic Environment Models (GEMs) for Agile Marine Autonomy
Dr. Fumin Zhang
School of Electrical and Computer Engineering
Technology Square Research Bldg., Room 430A
85 Fifth Street NW
Atlanta, Georgia 30332-0250

Award Number: N00014--10—1—0712



This project builds a roadmap to achieve agile marine autonomy that endows unmanned marine systems with ability to take fast responses to environmental changes.


We have developed generic environmental modeling in two applications using different types of AUVs. We have used a YSI Ecomapper for bathymetry survey, where the GEM is built to represent the bathymetry data. We have also used underwater gliders for ocean sampling, where the GEM is built to represent ocean flow.

Grand Isle Bathymetry Survey

Figure 1. Bathymetry survey using an YSI Ecomapper for a retention pond in Grand Isle State Park, LA


The EcoMapper (Figure 1, upper right) is an autonomous underwater vehicle purchased from YSI Inc. It is operated through “Windows remote desktop” via Wi-fi. GPS is available for localization on water surface and DVL (Doppler Velocity Log) is employed for localization underwater relative to the bottom. Our Ecomapper is also equipped with Conductivity and Temperature Sensors, Depth Sensor (measure depth from surface), Depth-Sounding sonar (measures height from bottom) and Three-axis digital compass. If operating in the autonomous mode, the EcoMapper follows a predefined course, either on surface or below surface and records all sensor measurements into log files. As part of an NSF funded project “Autonomous Control and Sensing Algorithms for Surveying the Impacts of Oil Spills on Coastal Environments” led by the PI, the EcoMapper was deployed to survey the tidal lagoon located at the Grand Isle State Park  (Figure 1, upper left) in Louisiana where oil pollutions have been spotted in 2010. Five autonomous missions between surface and 0.5 meters below surface were executed. By interpolating the DVL data, we obtained a bathymetry map for this pond (Figure 1, bottom). The salinity of the lagoon varies between 13ppt and 17ppt under different weather condition.

Two Slocum underwater gliders will be deployed in Long Bay, SC for an NSF project:
“Mechanisms of nutrient input at the shelf margin supporting persistent winter phytoplankton blooms downstream of the Charleston Bump.” PI’s team will control the motion of the two gliders to navigate near the edge of the Gulfstream, where strong current exceeding glider horizontal speed is often observed. We developed a GEM combining a simple tidal and Gulfstream current model based on M2 tide and sinusoidal meandering motion of Gulf Stream as shown in Figure 2. HYCOM (HYbrid Coordinate Ocean Model, http://www.hycom.org) will later be integrated into the GEM. Using the Glider Coordinated Control Systems (GCCS) developed by PI’s team, we simulate a control algorithm to maintain a glider’s position near the edge of Gulfstream. Starting from different positions, the trajectories for station holding are illustrated in Figure 3. It can be observed that the glider is able to escape from a strong northward Gulfstream current and come back to its desired position at the cross hair near the bottom of the figure.

Flow Model of Long Bay, SC

Figure 2. A station-keeping algorithm produces paths of simulated gliders in Long Bay, SC. A GEM is developed to combine tides and the gulfstream.

A dynamic programming approach has been implemented to generate optimal paths for an underwglider to maintain minimal distance from a set goal point under the influence of flow. This approach uses a cost function that integrates the glider’s distance from the goal over a finite time horizon. The domain of operation is discretized in both space and time, and the cost-to-go and associated optimal control actions are computed at each point in the discretized domain, starting at the final time (see Figure 4). The glider’s position is then integrated forward using a simple particle model for the glider dynamics. At each time step in the integration, the glider’s control action (e.g. the choice of heading angle) is taken to be a bilinear interpolation of the optimal control actions at the nearest states in the discretized domain. The glider’s total velocity is taken as a sum of the glider’s through-water velocity and the predicted flow velocity, which is obtained from the GEM used. This gives a near-optimal trajectory that can be converted to a waypoint list to be passed to the glider.


Figure 3. An illustration of dynamic programming for glider path planning. The arrows represent possible motions of the glider over one time step, given different available control actions (heading angles); the glider’s final position at the next time step depends on both the heading angle and the ambient flow. The value function at the current state (blue rectangle in the top layer) is given by the minimum over all possible heading. angles of the sum of the current distance from the goal and the value function at the glider’s final position

The dynamic programming path-planning algorithm has been tested in the simulation module of GCCS, which uses ocean-model flow data and an approximated model of glider dynamics, as well as an implementation of the glider’s on-board control algorithms, to simulate glider motion given waypoint lists passed to the glider. Figure 4 shows a planned path in a simulated flow field.


Figure 4. Dynamic programming-based path planning over a sample domain with a static flow field. The value function for all x,y positions is shown at selected time slices (left). The red on the far left and right sides of the domain marks the infeasible positions from which the glider will be carried out of the domain by the flow within the planning time horizon regardless of the control action taken (the cost at these positions is maximized). Given the value function, a path to the goal can be computed from an arbitrary starting position (if it is feasible). A sample path is shown in the figure on the right. The blue arrows show the flow velocity over the domain. The red arrows show glider headings along the path. The goal position is marked by an asterisk



We use controlled Lagrangian particle tracking (CLPT) to evaluate the accuracy of the simulated glider position. Errors in glider position simulation are due to limited resolution of ocean models, missing physics in the models, and sparseness of available ocean measurements used to drive the model. Using a modified Langevin equation to model the growth of the expected glider position error (termed CLPT error), we have shown that the magnitude of the expected error in simulated position grows exponentially until reaching a lower bound equal to twice the grid size of the ocean model used (see Figure 5). The error growth then slows to a polynomial function of time.


Journal articles:

W. Wu and F. Zhang, “Cooperative Exploration of Level Surfaces of Three Dimensional Scalar Fields,” Automatica, the IFAC Journal 47(9): 2044-2051, 2011. [published, refereed]

K. Szwaykowska and F. Zhang, “Trend and Bounds for Error Growth in Controlled Lagrangian Particle Tracking,” IEEE Journal of Oceanic Engineering, 2011. [submitted, refereed]

Refereed Conference Proceedings:

K. Szwaykowska and F. Zhang, “A Lower Bound for Controlled Lagrangian Particle Tracking Error, ” in Proc.49th IEEE Conference on Decision and Control (CDC 2010), 4353-4358, 2010. [published, refereed]

K. Szwaykowska and F. Zhang, “A Lower Bound on Navigation Error for Marine Robots Guided by Ocean Circulation Models ,” in Proc. 2011 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2011). [published, refereed]


Recipient: Fumin Zhang
Institution: Georgia Institute of Technology
Award: 2010 ONR YIP Award
Sponsor: Office of Naval Research

Recipient: Fumin Zhang
Institution: Georgia Institute of Technology
Award: 2010 Lockheed Inspirational Young Faculty Award
Sponsor: Lockheed Martin Co.

Recipient: Fumin Zhang
Institution: Georgia Institute of Technology
Award: 2011 Roger P. Webb Outstanding Junior Faculty Award
Sponsor: School of Electrical and Computer Engineering, Georgia Tech

Recipient: Fumin Zhang
Institution: Georgia Institute of Technology
Award: Distinguished Lecturer on Cyber-Systems and Control
Sponsor: Zhejiang University, China

Recipient: Georgia Tech Savannah Robotics (Supervised by Fumin Zhang)
Institution: Georgia Institute of Technology
Award: 2011 Martin Klein MATE Mariner Award
Sponsor: Marine Advanced Technology Education (MATE) Center