AI Algorithms and Sensor Integration for Robotic Vessels

Mindchip OÜ is focusing on making maritime industry cost-effective by developing autonomous self-adaptive captains that adjusts to different sea vessels. Robotic vessels are 5 times cheaper as manned vessels, as they won't need staff on the board and can operate 24/7.

Minchip services used within EDIH framework: test-before-invest services x 4 (2x demonstration project 2x DMA); AIRE networking events; training events; support to find funding x 2; EDIH collaboration.

Mindchip OÜ is revolutionising maritime navigation with an AI-based machine vision system for autonomous ships. The primary challenge in this test before invest project was developing and integrating this system to reliably detect and identify other ships and buoys, ensuring safe navigation. Initial plans to train the machine vision model on custom-collected data had to be adjusted due to time constraints, leading to the use of internet-sourced data. This posed challenges, particularly in accurately detecting local Estonian buoys, crucial for the autonomous ship's operational safety and effectiveness. 

Through AIRE's cooperation, MindChip has started collaborating with other EDIHs to develop the vessel: ARIC Hamburg and Northern Netherlands EDIH. 

The CEO of MindChip Heigo Mõlder said: ''Thanks to AIRE's measures, we have been able to put together a team of engineers that otherwise would've not been available for us due to lack of finances. We've done a real thing with the project that has been validated and deployed, and we are demonstrating it now. It's giving us more clients and real value.''

MindChip is and active participants in AIRE's events and AIRE's stakeholders events.

The solution was meticulously crafted by AIRE and Mindchip OÜ around an AI model trained on high-resolution imagery captured by four strategically positioned cameras, seamlessly integrated into the robust Robot Operating System (ROS). Leveraging the YOLOv5m algorithm proved instrumental, as it offered optimal performance in both speed and accuracy, crucial for identifying distant and compact objects in dynamic maritime environments.

Critical to the system's validation was its rigorous testing at sea, where it consistently showcased its capability to detect small boats from distances ranging between 100 to 150 meters, and larger vessels from even greater distances. This capability underscored its potential to enhance navigational safety and operational efficiency in real-world scenarios.

Acknowledging the initial dataset's constraints, particularly in accurately detecting local Estonian buoys, an extensive new dataset tailored to enhance buoy detection accuracy was meticulously compiled. This proactive approach not only addressed the initial challenges but also positioned their AI system to meet stringent safety standards required for autonomous ship navigation.

Through perseverance and innovation, the solution provided by AIRE to Mindchip OÜ not only surpassed technical expectations but also exemplified their commitment to pioneering advancements in maritime technology. By pushing boundaries and embracing challenges head-on, they are charting a course towards safer, smarter, and more autonomous navigation on the high seas.

The AI-based machine vision system significantly enhanced Mindchip's capabilities in autonomous navigation. Key results include:

• Enhanced Detection Capabilities: The system consistently detected small boats from distances spanning 100 to 150 meters, and identified larger ships at even greater ranges. This heightened capability crucially enhances situational awareness and navigational safety.

• Improved Operational Efficiency: By bolstering detection accuracy and reliability, the system significantly enhances the safety and operational effectiveness of autonomous ships. This translates to smoother navigation and reduced risk of collisions or operational disruptions.

• Robust Dataset Development: The compilation of a comprehensive dataset not only addressed initial challenges but also lays the groundwork for ongoing improvements and future advancements in maritime AI technology.

• Versatile Integration: Integrated seamlessly into the Robot Operating System (ROS), the system's applicability extends beyond maritime use cases to encompass diverse applications in land-based robotics and smart city infrastructures. This versatility underscores its potential to drive innovation across various domains.

Quantitative Benefits:

• Environmental Footprint Reduction: Reduction in environmental footprint by 5 times. This reduction is attributed to optimised navigation routes and reduced fuel consumption.

• Cost Savings: Realized substantial cost efficiencies, with savings ranging from 6 to 10 times compared to manned ships. These savings stem from reduced operational overhead, maintenance costs, and minimised risks associated with human error.

• Training Efficiency: Streamlined operational procedures resulted in a remarkable 5-fold reduction in training requirements for personnel, thereby enhancing workforce efficiency and readiness.

Funding and investments:

AIRE support: The company has stated that participating in EDIH services has been a major help in advancing their product and after the resolved challenges with test before invest projects, they have been able to showcase and sell the product to investors further than expected. Furthermore, the public funding service has supported MindChip with next steps to bigger investments in the product and has been helpful in supporting the team to source the right funds.

In conclusion, the developed AI-based machine vision system not only elevates maritime navigation standards but also sets a precedent for sustainable, cost-effective, and safer autonomous operations across global waters and beyond.

Do’s:

• Prioritise Reliability of Core Systems: Ensuring the dependability of fundamental components, such as cameras and image transmission systems, is critical for the overall success and accuracy of the AI-based machine vision system.

• Conduct Extensive Testing: Rigorous testing under a wide range of environmental and operational conditions helps to ensure the model's performance and robustness in real-world scenarios.

Don’ts:

• Don't Underestimate Time and Resources: Tasks that may seem straightforward, such as setting up data collection systems, often require more time and resources than initially anticipated. Properly allocating sufficient time and budget for these tasks is crucial to avoid delays and ensure project milestones are met.

• Avoid Sole Reliance on Internet-Sourced Data: While internet-sourced data can be valuable, it should not be the sole source of training data. Failing to incorporate region-specific requirements can significantly impact the system's accuracy.