Browser-Native Digital Test Range for Gliders

Summary & Methodology Analysis

The platform architecture leverages browser-native execution environments, specifically Pyodide and WebAssembly, to run simulations directly in the client. Users configure mission parameters by selecting regional bathymetry via GEBCO_2026 grids and defining environmental variables like currents and science objectives. The system executes missions using mission-scale current-advection kinematics rather than high-fidelity nonlinear vehicle modeling, which significantly simplifies the simulation loop for benchmarking purposes. This approach allows developers to import external planner outputs or use built-in baselines to generate 4D trajectories, which are then analyzed by a common evaluator to compute operational, dive-policy, and science-based metrics.

Evaluation relies on a standardized framework where the system maps executed trajectories against specific mission constraints to generate reproducible observation records. By using WebGL for visualization and common kinematics, the platform avoids the need for complex server-side infrastructure. The researchers benchmarked five planners across two OSSE (Observing System Simulation Experiment) episodes. OSSEs are simulation-based procedures used to evaluate the potential impact of new observational data before deployment. In this experiment, all 54 missions reached successful completion without triggering hard violations of mission constraints.

Despite its utility for benchmarking, the system has notable constraints. The simulator does not utilize high-fidelity nonlinear vehicle models, and the current simulation engine failed to perfectly replicate historical field deployment endpoints during audits. Furthermore, the experiments were limited to three random seeds, and the reconstruction skill metrics approached saturation levels. The platform is intended for pre-deployment experimentation and is explicitly not certified for navigation-grade usage.