from Orca-AI
From Forbes by Yarden Gross
Artificial intelligence is infusing all modes of transit, reshaping how we travel from A to B. Perhaps most evident is in the automotive industry, which saw investors pour $4.2 billion into autonomous driving technologies.
A key emphasis has been the development of Advanced Driver Assistance Systems (ADAS), which could prevent nearly 30 percent of all U.S. car crashes and save over 10,000 lives annually.
But while ADAS has already begun to make the roads safer, the development and adoption of similar technologies have been much slower in the maritime industry, where an ADAS-type innovation is sorely needed.
A report by the European Maritime Safety Agency shows that accidents are on the rise in increasingly crowded waterways, from 2,000 in 2011 to an estimated 4,000 in 2017.
So why can’t the industry simply adopt tools from the automotive market?
Simply put, the maritime industry presents a vastly different set of needs, and training ships to navigate autonomously presents an array of challenges that the automotive industry does not face . But while automotive innovations are not directly transferable to the maritime world, the industry can move closer to “self-sailing ships” by building systems that meet the maritime domain’s unique needs.
Turbulent Waters: Shipping’s Unique Challenges
The water’s inherent environmental and navigational challenges illuminate the key obstacles to innovation confronting the maritime industry, even amid new advances in sensors and data science.
Consider the differences in traffic regulation between cars and ships.
While both can theoretically rely on others to operate according to the established “rules of the road”, the highly defined structure of every inch of the road makes it easier for autonomous and semi-autonomous cars to navigate and to predict the behavior of other cars.
Outside of port areas, traffic for ships is less regulated.
Concepts like traffic lanes, red lights, and other signage simply don’t exist.
Moreover, even in areas with rigorous regulations such as near ports and in crowded waterways, accidents occur with alarming frequency due to overcrowding and human error – the latter accounting for up to 96 percent of maritime accidents.
For cars, safe driving usually requires visibility of a few hundred feet; if visibility is hazy, vehicle-to-vehicle networks can convey information about the car’s surroundings, still allowing for near-instant reaction time.
However, ships, particularly large vessels with slow turn rates, can require up to six miles of open water to adjust course and avoid collisions, so ample advance notice is crucial to shift directions and avoid a collision.
Ships, then, must look after themselves – and while sensors can aid in this task, their reliability hinges on visibility and the ability to identify vessels and hazards at a great distance.
Given the frequency of storms, fog, moonless nights, and other weather events, ships often lack the visibility necessary for sensors to provide the needed information.
All of these sensor problem will be overcome – but that won’t spell the end of the maritime industry’s technological challenges.
For shipping to attain greater autonomy, vessels will need AI systems that can deftly navigate the dynamic and frequently turbulent environments typical of the waters.
Take data – the lifeblood of AI.
Over the course of a weeks-long voyage, a ship will only have limited interaction with other vessels, meaning that relatively little data collected during the voyage will prove useful for training AI.
Whereas cars encounter anomalies and edge cases on a daily basis, thereby guiding the development of autonomous systems, a shipping database with such information is much more difficult to build, and while the process of collection of data is underway, much work remains.
Charting the Course to Autonomy
Overcoming the hurdles to autonomous shipping will entail a process of honing industry-specific technological solutions.
New sensors that provide visibility even in the lowest light, such as thermal cameras, need to reliably perform in the difficult conditions that are commonplace in the oceans, from hail to rain to gale force winds.
Such maritime tailored technologies will be essential to securing industry buy-in.
But the adoption of those sensors and cameras alone are not enough.
They will need to be integrated into navigational systems that are capable of recognizing hazards from miles away in zero visibility conditions.
That requires that the systems and algorithms that operate them be put through extensive real-world training to ensure the systems are able to effectively classify the hazards a ship will face on the water, for both routine and edge case scenarios alike.
Given the current challenges, full autonomy is still years off, and even when it is achieved, crews won’t be eliminated, rather they will continue to play a vital part in shipping operations, from monitoring system performance to delivering goods.
Solutions must be designed with crews’ needs in mind.
Like autonomous cars, autonomous shipping promises to make transportation safer, cheaper, and more efficient.
But while automotive technologies like ADAS and others can serve as a reference point, the maritime industry will need to chart its own course to reach autonomy.
Links :
- VentureBeat : Orca AI raises $2.6 million to stop ship collisions
- EMSA : Annual overview of marine casualties and incidents 2018
- MarineMec : Addressing the ironies of automation
No comments:
Post a Comment