Abstract
Ocean processes are dynamic and complex, occurring across multiple spatial and temporal scales. To obtain a synoptic view of such processes, ocean scientists collect data over long time periods. Historically, this has been achieved through ship-based surveys or moored sensors, which are often limited by cost or spatial coverage. This paper introduces a novel framework for "persistent monitoring" using autonomous underwater gliders. By combining a specialized path-planning algorithm (ZZTOPP) with a speed-control strategy that adjusts the vehicle's pitch, the researchers demonstrate how to maximize scientific data collection in high-interest areas while navigating the unpredictable forces of ocean currents.
Introduction
Oceanography faces a fundamental "sampling vs. coverage" dilemma. To understand dynamic phenomena such as Harmful Algal Blooms (HABs), researchers require a broad, synoptic view of a region over extended periods, while simultaneously needing high-resolution data in specific hotspots.
Underwater gliders have emerged as a preferred tool for this task due to their extreme energy efficiency, allowing them to remain at sea for months. However, they possess two significant operational weaknesses:
- Limited Control: Gliders move slowly, driven by buoyancy changes, making them highly susceptible to being pushed off course by strong ocean currents.
- Navigational Uncertainty: While submerged, they rely on dead-reckoning, which leads to cumulative positional errors between GPS fixes at the surface.
Standard "lawnmower" search patterns often prove inadequate because gliders cannot maintain the precise turns and straight lines required. This research seeks to design paths that are robust to currents ("repeatable") while automatically concentrating sampling effort in high-priority zones.
Methodology
The proposed framework utilizes a two-part optimization strategy to address the challenges of persistent monitoring.
ZZTOPP Path Planner
The Zig-Zag in the Tranquil Ocean Path Planner (ZZTOPP) moves away from rigid geometric patterns. Instead, it uses a constrained version of the Bellman-Ford algorithm to identify a closed-loop path. This path is designed to maximize a "reward"—defined as time spent in high-interest zones—while minimizing the "cost," which is the exposure to high-magnitude or high-variance currents.
Sampling Resolution Optimization
Gliders typically collect data in a "sawtooth" vertical pattern. By adjusting the pitch angle, the glider can take more vertical profiles over a shorter horizontal distance, thereby increasing sampling resolution. However, this comes at the cost of horizontal speed. The authors developed a convex optimization tool to determine the optimal pitch for each path segment, balancing the scientific need for detail against a strict total time budget for the mission.
Iterative Integration
These two components are combined into an iterative algorithm. The initial path informs the speed control requirements, and the resulting speed control updates the cost associated with the path. This process continues until the system converges on a globally efficient mission plan.
Experimental Validation
The algorithms were validated through three separate field deployments using Webb Slocum gliders:
- Southern California Bight (SCB): Used the Regional Ocean Model System (ROMS) to predict current magnitudes for path planning.
- Monterey Bay (MB): Utilized the Navy Coastal Ocean Model (NCOM ICON) to predict current speed variance, prioritizing paths through more predictable waters.
These trials served as real-world benchmarks against traditional, human-designed reference paths.
Results and Insights
The field trials demonstrated significant improvements over standard navigational methods. In the Southern California Bight, the ZZTOPP-generated path achieved a navigation score (measuring deviation from the plan) of 56.23 km², compared to the 70.35 km² average of traditional paths.
Key insights from the study include:
- The Importance of Variance: In Monterey Bay, researchers found that avoiding unpredictable (high-variance) currents was as important as avoiding strong ones. Steady currents can be compensated for by onboard systems, whereas erratic currents lead to navigational failure.
- The Zig-Zag Advantage: Zig-zag paths are inherently more "forgiving" for dead-reckoning vehicles. They allow for longer legs that reduce the frequency of time-consuming surfacings while maintaining adequate spatial coverage.
- Sampling Gains: In Monterey Bay, the optimized path increased the number of vertical profiles in the highest-priority region by over 400% compared to the standard reference path.
Conclusion
The research successfully demonstrates that autonomous gliders can be programmed to provide both persistent coverage and high-resolution sampling. However, the system's success remains heavily dependent on the accuracy of underlying ocean models. Future work will focus on "closed-loop" planning, enabling gliders to update their paths in real-time based on locally sensed currents rather than relying solely on a priori model predictions.
Summary updated