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

The ocean is never still. Even on a calm day, water moves in complex, shifting currents that carry everything floating on its surface. For scientists trying to study the health of our oceans, these currents are both a tool and a challenge. They help spread pollutants, microplastics, and algae blooms, making them visible to sensors, but they also push autonomous robots off course, wasting energy and time. To understand what is happening in the water, researchers need to send out robotic boats that can navigate these flows while simultaneously tracking the drifting particles they are meant to sample. The difficulty lies in the fact that the water moves the robot and the target particles at the same time, but often in different ways depending on the exact shape of the current and the speed of the boat. If a robot is designed without accounting for this dual influence, it might follow a perfect geometric line on a map only to be swept away by a hidden eddy, missing its targets entirely.

This is the problem tackled by a new simulation tool called H-SPAR, developed by researchers at Toronto Metropolitan University and University College London. Rather than trying to build a physical robot to test every possible scenario in the real ocean, the team created a highly detailed digital environment where they could watch how a virtual boat and virtual particles interact with a virtual ocean. The key innovation is that this simulator treats the water current not just as a background disturbance, but as a shared force that drives both the robot's movement and the drift of the particles it is trying to catch. In previous systems, these elements were often modeled separately, which meant researchers could not see how a change in the current would simultaneously alter the boat's path and the location of the floating debris. H-SPAR links them together, allowing scientists to test sampling strategies in a world that behaves much more like the real ocean.

The researchers used this tool to test how well different navigation strategies work when the water is moving. They compared a standard planning method, which simply looks for the shortest distance between two points, against more advanced methods that take the current into account. In the simulation, the standard method produced the shortest line on the map, but when the virtual boat tried to follow it, the current pushed it off course, forcing it to burn extra fuel to fight the flow. The advanced planners, which knew the direction and speed of the water ahead of time, chose slightly longer routes that went with the current instead of against it. This saved a significant amount of energy, reducing the cost of fighting the upstream flow by nearly 70 percent during the planning phase. However, the simulation revealed a crucial nuance: when the current changed over time, as it does in the real world, the advantage of these smart planners dropped to about 42 percent. This showed that while knowing the current helps, the changing nature of the water and the physical limits of the boat's steering still create a gap between the perfect plan and the actual journey.

The study also looked at how the direction a robot moves while scanning an area affects its ability to catch floating particles. The team tested different sweeping patterns, similar to how a person might mow a lawn, but with the robot moving at different angles relative to the flow. They found that the best angle to catch particles was not a fixed rule; it depended entirely on how the particles were moving and how likely the robot was to grab them. In some cases, moving perpendicular to the current caught the most particles, but in others, moving with the current was more effective because it changed the speed at which the water entered the robot's sampling intake. The most important discovery was that the angle that caught the most particles geometrically was not always the one that successfully sampled them. Because the robot's ability to grab a particle depends on the relative speed between the boat and the particle, a path that looks efficient on a map might actually be poor at collecting samples if the water is moving too fast or too slow relative to the intake.

By running these complex scenarios in a single, unified system, the researchers demonstrated that you cannot evaluate a robot's success by looking at its path or its sampling rate in isolation. The water connects everything. A path that saves energy might miss the target particles, and a path that catches many particles might waste too much fuel. The H-SPAR simulator provides a way to see all these trade-offs at once, offering a clearer picture of how to design missions for the real ocean. The findings suggest that future robotic missions will need to be planned with a deep understanding of how the water moves both the robot and the target, rather than just treating the ocean as a static stage. This approach allows for more realistic testing of strategies before a single physical robot ever leaves the dock, ensuring that when they do go out, they are ready for the dynamic, shifting reality of the sea.