PyGemini: Unified Software Development towards Maritime Autonomy Systems

TL;DR

PyGemini is a Python-based framework that unifies maritime autonomy development, enabling flexible, scalable, and maintainable decision-making systems for autonomous surface vessels through innovative configuration-driven processes.

Contribution

It introduces a novel Configuration-Driven Development process that integrates BDD, data-oriented design, and containerization for maritime autonomy software.

Findings

Supports modular and scalable maritime autonomy applications.

Enables versatile deployment as a stand-alone, cloud, or embedded system.

Provides tools for simulation, scenario generation, and AI-based imagery augmentation.

Abstract

Ensuring the safety and certifiability of autonomous surface vessels (ASVs) requires robust decision-making systems, supported by extensive simulation, testing, and validation across a broad range of scenarios. However, the current landscape of maritime autonomy development is fragmented -- relying on disparate tools for communication, simulation, monitoring, and system integration -- which hampers interdisciplinary collaboration and inhibits the creation of compelling assurance cases, demanded by insurers and regulatory bodies. Furthermore, these disjointed tools often suffer from performance bottlenecks, vendor lock-in, and limited support for continuous integration workflows. To address these challenges, we introduce PyGemini, a permissively licensed, Python-native framework that builds on the legacy of Autoferry Gemini to unify maritime autonomy development. PyGemini introduces a novel Configuration-Driven Development (CDD) process that fuses Behavior-Driven Development (BDD), data-oriented design, and containerization to support modular, maintainable, and scalable software architectures. The framework functions as a stand-alone application, cloud-based service, or embedded library -- ensuring flexibility across research and operational contexts. We demonstrate its versatility through a suite of maritime tools -- including 3D content generation for simulation and monitoring, scenario generation for autonomy validation and training, and generative artificial intelligence pipelines for augmenting imagery -- thereby offering a scalable, maintainable, and performance-oriented foundation for future maritime robotics and autonomy research.