Research Article

AUV-Based Plume Tracking: A Simulation Study

Table 2

Different plume tracking strategies in literature.

Article Number of agents Properties Limitations

[23] Multiple Uses CUSUM algorithm, presents the boundary estimation problem as a HMM, and recast as an optimization problem Considered only ellipse-shaped, no path optimization, simulation only, intervehicle communication issues not addressed, and AUV dynamics not considered

[24] Multiple Decentralized, gradient-free algorithm, convergent and stable AUV dynamics not considered, high computation cost, intervehicle communication issues not addressed, and only very simple plume shapes considered

[3] Single, multiple Cooperative, generating polygons to follow based on ocean model predictions, simulation, and practical implementation Trajectory based on (roughly) approximated polygons, temporal constraints not considered, unable to react to the fast moving features, and ignoring the dynamics of the glider

[25] Single Behavior-based approach for plume mapping, subsumption architecture, showing experimental results Limitations of behavior-based approach, no description on adaptive mapping, and implementing simple preplanned lawn-mower strategy

[26] Single Uses colored dissolved organic matter (CDOM) sensor for planed missions, adaptive planning using in situ current, and temperature measurements, and gets the 3D track of the AUV, practical implementation No information on path optimization and no adaptive tracking

[27] Single Based on peak-capture algorithm, it generates a sawtooth trajectory and uses depth information and practical implementation No information on path optimization

[28]Multiple Adaptive behavior-based system, acoustic communication within AUVs, representing the plume using Fourier orders when reconstructing No information on path optimization and communication overhead

[29]Single Uses a plume indicator function and real-time implementation and uses adaptive transects; transect length depends on number of consecutive samples; distance between transects is a percentage of previous samples Less path optimization, not using an AUV, no information about the convergence, and coverage of the used algorithm

[30]Single Uses remote sensing data to detect hotspots, uses surface current to project plumes spatiotemporally, and runs in a lawnmower type pattern, practical implementation No path optimization, only using predefined pattern, and no adaptive tracking

Proposed Single Path optimization based on gradient information, adaptive plume tracking, and centralized approach Only simulation results, no comparison data available, relying on remote sensing data for locating the plume region initially, low performance to noisy plume boundaries, and using only one fluorometer