PaperlayerView exact revision on arXiv

Source-linked AI summary

H-SPAR: Hydrodynamic-aware Simulation for Particle Transport and Autonomous Robots

Navid Zarrabi, Nariman Yousefi, Sajad Saeedi

arXiv:2610.01985v1cs.RO

TL;DR

Environmental sampling missions must account for currents that affect both robot motion and particle transport, but existing marine simulators do not jointly model these processes. H-SPAR integrates shared flow fields, particle transport, probabilistic sampling, and USV autonomy; its experiments show that planning and sweep-orientation choices must be assessed alongside execution and sampling behavior.

  • Problem

    Marine sampling requires evaluating currents’ joint effects on vehicle motion and particle transport, while existing simulators lack joint modeling of hydrodynamic transport, flow-dependent execution, and sampling-oriented planning.

    Source: §1 · pp. 2–3

  • Method

    H-SPAR integrates a shared flow-field representation with Lagrangian particle transport, probabilistic sampling, and ROS 2/Gazebo-based USV autonomy.

    Source: §1 · pp. 3–4

  • Results

    Experiments show that the shortest geometric path is not necessarily most effective under spatially varying currents, and preferred sweep orientation depends on flow, particle transport, mission duration, and the sampling model.

    Source: §4 · p. 16, §3.2.1 · p. 16

  • Takeaways & Limitations

    Evaluating marine particle-monitoring strategies requires jointly considering hydrodynamics, vehicle execution, particle transport, and probabilistic sampling.

    Source: §3.2.1 · p. 16, §4 · p. 16

    H-SPAR uses one-way coupling, so vehicle-induced wakes, propeller wash, and particle disturbances are not resolved.

    Source: §2.5 · p. 10

Abstract

from arXiv · show

Environmental robotic sampling requires considering the dual influence of water currents on robotic motion and particle transport. Existing marine robotics simulators generally model flow, autonomy, and sampling targets separately, limiting joint evaluation of mission cost and sampling performance. H-SPAR integrates spatially and temporally varying velocity fields, Lagrangian particle transport, probabilistic sampling, and ROS 2/Gazebo-based uncrewed surface vehicle (USV) autonomy. In this work, shared precomputed flow fields drive particle advection and current-induced forces during closed-loop vehicle execution. Path-planning experiments show that the existing current-aware planner SVF-RRT* achieves 69.4% lower upstream cost than conventional RRT* at the planning level, but this reduction falls to 41.7% during execution under time-varying currents, reflecting temporal flow variation, vehicle motion constraints, and path deviation omitted during planning. Coverage experiments show that sweep orientation changes the particle-sampling rate by up to 22.2% under the complete H-SPAR configuration. These findings highlight the importance of evaluating planning, vehicle execution, particle transport, and sampling together under consistent hydrodynamic conditions. The project webpage is available at https://sites.google.com/view/h-spar, and the open-source code is available on GitHub at https://github.com/naviiidz/h-spar-sim.

1 Introduction

Marine currents shape both USV motion and the locations of sampling targets, yet existing approaches do not jointly represent these effects for mission evaluation. H-SPAR combines flow-driven particle transport with current-aware USV simulation and evaluates sampling missions under shared flow conditions.

1 Introduction

  • Currents alter USV trajectories and mission costs while advecting and redistributing the particles being sampled.Jointly modeling these effects is important for evaluating autonomous sampling missions.Source: §1 · pp. 2–3
  • Existing marine simulators and planning approaches lack joint evaluation of fine-grained flow, current-affected vehicle execution, particle transport, and sampling.Prior planners often use point-mass abstractions, while simulator platforms do not support the combined modeling needed to assess trajectory execution, energy, and sampling efficiency.Source: §1 · pp. 2–3, Table 1 · p. 3
  • H-SPAR uses a shared fine-grained velocity field to model both current-induced forces on the USV and particle advection.The framework integrates Lagrangian particle transport with hydrodynamic-aware USV simulation to evaluate sampling missions under marine flow.Source: §1 · p. 3, Figure 1 · p. 4
  • The framework combines ROS 2/Gazebo autonomy, probabilistic particle sampling, planner evaluation across planning and execution, and hydrodynamic coverage benchmarking.Its coverage benchmark examines how sweep orientation affects particle capture, mission duration, and energy expenditure under flow.Source: §1 · p. 4

2 Methodology

H-SPAR connects current-aware planning, hydrodynamic modeling, vehicle simulation, particle transport, and probabilistic sampling through ROS 2 and Gazebo. Shared environmental velocity fields support both vehicle execution and particle motion.

2 Methodology

  • ROS 2 coordinates current-aware planning, flow and particle models, and vehicle commands with Gazebo’s simulated environment, sensors, and USV dynamics.Sensor measurements pass from Gazebo to ROS 2, which returns thruster commands and hydrodynamic forces for closed-loop simulation.Source: Figure 2 · p. 5, §2 · pp. 5–6

2.1 Environmental Velocity Field

  • H-SPAR accepts spatially and temporally varying velocity fields from multiple sources; this study uses depth-averaged shallow-water equations that omit vertical profiles and near-body flow.The fields are exported from an unstructured triangular mesh and used for particle advection, path evaluation, and mission simulation.Source: §2.1 · pp. 6–7

2.2 Current-Aware Path Planning

  • RRT* evaluates path length and obstacles, while VF-RRT* and SVF-RRT* also query local currents to favor flow-aligned paths and avoid strong opposing currents.H-SPAR evaluates upstream cost along planned paths using the local velocity field.Source: §2.2 · p. 7

2.3 Current-Aware USV Dynamics

  • H-SPAR queries current-relative flow at each USV’s position and models hydrodynamic drag, including surge, sway, and yaw effects, in its vehicle dynamics.The framework uses quadratic damping, with current-induced loads combined with propulsion, wind, and wave inputs.Source: §2.3 · pp. 7–8

2.4 Gazebo-Based Simulation, Execution, and Metrics

  • Gazebo simulates vehicle dynamics and sensors, while H-SPAR records planned and executed path lengths, upstream costs, mission duration, and trajectory deviation for comparison.Executed upstream cost is computed along the recorded trajectory using the same formulation as planned cost.Source: §2.4 · pp. 8–9

2.5 Lagrangian Particle Transport

  • Lagrangian particles move with interpolated local flow and an optional stochastic diffusion term; one-way coupling excludes vehicle-induced flow and particle disturbances.This assumption is used for mission-scale simulations; resolving localized vehicle effects is left to two-way coupling or local CFD.Source: §2.5 · pp. 9–10

2.6 Probabilistic Particle Sampling Model

  • The sampling model treats particles within the robot-frame sampling radius as candidates and assigns capture probability based on particle–robot relative speed.Capture probability is Gaussian around the pump’s design intake velocity, with maximum probability at the isokinetic condition; sampled particles are removed to prevent duplicate detections.Source: §2.6 · pp. 10–11

3 Experiments

The experiments assess current-aware planning both during planning and vehicle execution, then examine how sweep orientation and modeling assumptions affect particle monitoring. Together, they show that flow conditions and sampling assumptions shape the preferred mission strategy.

3.1 Experiment 1: Current-Aware Planning and Execution

  • RRT* yields the shortest planned path but highest planned upstream cost, while SVF-RRT* has the lowest planned upstream cost among the evaluated planners.VF-RRT* and SVF-RRT* generate longer paths than RRT*.Source: §3.1 · pp. 11–12
  • Time-varying currents worsen current-aware planners’ execution performance because time-averaged planning omits temporal changes, while vehicle constraints and hydrodynamic forces require corrective motion.Their paths may cross currents more sharply, increasing path deviation and mission time in the time-varying-current case.Source: §3.1 · pp. 12–13
  • Under matched time-averaged flow, current-aware planners reduce mission time and executed upstream cost versus RRT*, but their advantage diminishes under time-varying flow.The comparison includes planning and execution under time-averaged and temporally varying fields.Source: §3.1 · pp. 12–13, Table 2 · p. 13

3.2 Experiment 2: Coverage-Based Particle Monitoring

  • Coverage monitoring uses boustrophedon waypoints in a user-defined region and compares sweep orientations under prescribed flow and particle distributions.To sustain particle influx, 100 particles are released from the upstream boundary every 10 s; transport is simulated under uniform and temporally varying flow.Source: §3.2 · p. 14

3.2.1 Ablation Study

  • The highest sampling-rate orientation depends on flow and sampling assumptions: it is 90° under uniform deterministic flow, 45° under varying flow with deterministic sampling, and 0° with probabilistic sampling.Probabilistic capture depends on relative velocity between the USV and particles.Source: §3.2.1 · pp. 14, 16
  • Hydrodynamic wrench forces change particle counts and sampling rates but do not consistently change the highest-rate orientation; the preferred sweep depends on duration, flow, particle transport, and relative velocity.Sampling rate normalizes for mission-duration effects on particle-encounter opportunities.Source: Table 3 · p. 14, §3.2.1 · pp. 14, 16
  • In the deterministic, no-wrench comparison, the 45° sweep yields the most particle encounters, while the 0° sweep achieves the highest detection rate and lowest upstream cost.The 45° sweep’s higher encounter count is primarily associated with its longer mission duration.Source: Figure 4 · p. 15

4 Conclusion

H-SPAR integrates flow fields, particle transport, and USV autonomy to evaluate marine sampling across planning, execution, and sampling metrics. Its source code and supporting simulation data are publicly available.

4 Conclusion

  • Flow-aware planners can reduce upstream motion and improve mission performance, but may incur tracking deviations; shortest geometric paths are not necessarily most effective under spatially varying currents.Source: §4 · p. 16
  • Sweep orientation preferences depend on mission duration, hydrodynamic cost, particle transport, and the sampling model, particularly when capture probability depends on particle–vehicle relative velocity.Source: §4 · p. 16
  • Marine monitoring strategies should be evaluated with planning-, execution-, and sampling-level metrics rather than geometric measures or total particle counts alone.Source: §4 · p. 16
  • Future work includes validating particle transport and vehicle drift against physical data, extending flow fields, refining sensing and capture models, and testing additional planners.Source: §4 · p. 16

Data availability

  • H-SPAR source code and simulation data supporting the findings are publicly available on Zenodo, and the development repository is on GitHub.Source: Data availability · p. 17
Loading 2610.01985v1…