• Welcome to engineeringclicks.com
  • Autonomous Shipping put to the test

    Discussion in 'Mechanical Design news & events' started by Bill Toulas, Dec 4, 2018.

    1. Bill Toulas

      Bill Toulas Well-Known Member EngineeringClicks Expert

      Joined:
      Sep 2017
      Posts:
      65
      Likes Received:
      0
      What do you think about this? https://www.theengineer.co.uk/falco-autonomous-ferry-rolls-royce/

      I personally believe that autonomous driving will be a lot easier to implement on environments like seas and lakes because they have way less unpredictability, and things happen in slow motion. Even in cases of severe weather conditions, an AI could potentially handle the situation with poise. Where do you stand on autonomous shipping?
       
    2.  
    3. john12

      john12 Well-Known Member EngineeringClicks Expert

      Joined:
      Dec 2018
      Posts:
      301
      Likes Received:
      0
      Very interesting. I think it makes perfect sense on the 'high seas', although maybe harbour pilots will still be needed for congested areas and the final sections.

      I read a story recently about a tiny amount of shipping causing more pollution than millions of cars. I think they're implementing a cleaner fuel standard for ships at the moment but it will remain to be seen how well this is enforced in International Waters.
       
    4. Obashb

      Obashb Member

      Joined:
      Dec 2018
      Posts:
      14
      Likes Received:
      0
      With all the massive investment and research, it is the right time for tested technologies to be put into test. This is a big win for engineers and scientist alike in their quest to develop high-tech systems and mechanisms.
      Much still has to be done to ascertain safety aspects of such autonomous systems, it has to be subjected to virtually all possible sea scenarios. What will be the drive system response to high tides? What safety measures are put in place to avoid near miss and/or close collision? What is the impact of the technology and AI systems on human employees?
       
    5. john12

      john12 Well-Known Member EngineeringClicks Expert

      Joined:
      Dec 2018
      Posts:
      301
      Likes Received:
      0
      Yeah, it might start off like airliners - essentially the 'autopilot' flies for 90% of the route, but it's monitored by human pilots and the critical parts (takeoff and landing/docking) are done by people.

      I wonder how much input a crew typically has on one of these voyages at the moment anyway. I imagine they're not manually steering the ship for a few weeks at a time. Surely there's already a large automated component?
       
    6. MSHOfficial

      MSHOfficial Well-Known Member EngineeringClicks Expert

      Joined:
      Nov 2018
      Posts:
      172
      Likes Received:
      0
      People are scared to sit on a ship that is controlled by AI. That is one reason why airplanes are not autonomous actually. But for me personally, I wouldn’t mind sailing on a ship controlled by AI. But when things go wrong, in such cases, it ends up as a disaster. No in-betweens.

      So imagining a cruise ship would carry 4000-5000 passengers, if anything was to go wrong, it would be a total disaster. However, if its about shipping of goods or oil. I guess it would be a good idea to use autonomous shipping. Except for dangerous terrains with pirated waters. And ships with harsh sea surface conditions or icebergs.

      But I don’t know why autonomous shipping would be important. Of course it saves company the money and all. But there are sea men and sailors who love this job. also, the number of crew required to sail a ship is not a lot. About 20-30 people. That’s 1 crew per 150 passengers.

      Other than the sailing crew, autonomous ship would need all the other crew.

      Also I think if autonomous shipping is going to be a reality. Then the number of crew members on a cargo ship would be reduced to one third. And this would mean the crew get to spend months on end with even less people than it used to be. I think they would prefer the slightly bigger group to sail.
       
    7. john12

      john12 Well-Known Member EngineeringClicks Expert

      Joined:
      Dec 2018
      Posts:
      301
      Likes Received:
      0
      Yeah, I've also heard that the main barrier to planes being completely automated is that passengers wouldn't feel comfortable being flown entirely by a computer and not a person. But this is kind of strange because the majority of accidents are caused by human error!

      I guess it's like self-driving cars - once they're implemented the roads will be way safer (and more efficient!) but people are going to find it weird during the changeover period.

      As you mention too - humans deal well with unusual situations like emergencies and non-standard incidents!
       
    8. MSHOfficial

      MSHOfficial Well-Known Member EngineeringClicks Expert

      Joined:
      Nov 2018
      Posts:
      172
      Likes Received:
      0
      Majority of accidents are caused by human error. But, almost all the times there is human error involved in an accident, there are also other faults or errors that lead to that accident. However, the investigators deem it pilot error.

      There have been disasters which were caused by autopilot or other instances where the pilot was misled to taking decision which doomed their flight.

      As an engineer I know that machines fail, sometimes drastically. Even if everything seems to be fine, one sunny morning it just doesn’t do its job correctly. So imagine we were in an era where all airplanes were flown by AI. Then majority of the accidents would be caused by machine error. Even if the accident was caused by a technician who forgot to cover the pitot tube during a windy day when the airplane was on the ground for a couple of days. Because the machine read the data wrongly and flew the airplane into disaster.

      Humans are scared of the Al because they don’t know what its capable of. Maybe AI will someday surpass the human way of thinking and its computational power would be exceptional. But I still think, even that day we would have accidents. Because that is what they are, an accident.
       
    9. john12

      john12 Well-Known Member EngineeringClicks Expert

      Joined:
      Dec 2018
      Posts:
      301
      Likes Received:
      0
      Yeah, for instance that Air France flight that crashed near Brazil - technically the root cause was the icing up of the pitot tubes (technical failure) but then this caused problems with the instruments which were misinterpreted by the pilots... so you could say that it was pilot error.
      It's hard to say whether a fully autopilot mode would've dealt with the situation better.
       
    10. MSHOfficial

      MSHOfficial Well-Known Member EngineeringClicks Expert

      Joined:
      Nov 2018
      Posts:
      172
      Likes Received:
      0
      It wouldn’t have, I guess. Because you design AI to do what a human would do. Just you make them do it faster. So if you were relying on readings from the same pitot tubes, as in it would have been the same input. The flight would have ended up in the same conditions, just faster.

      So now days we have autopilot that continually monitor the system, all the inputs and outputs, and it runs them through the pilot if anything seems out of the normal range dynamically. For example if your TAS (true airspeed) is increasing while you are climbing (that’s not possible unless you changed throttle settings) it runs the numbers through the pilot (sometimes also the ground) and tells them there might be something really wrong. Either the TAS shown is wrong or the climb angle etc.

      But somehow autopilot could really help, there was a case where a Japanese senior pilot once crashed a plane on purpose. Took the plane into a nose dive and suicided (at least that’s what the investigators propose). So autopilots today would not let the pilot do something so drastic. Inputs like that would be run through the ground and unless preprogrammed in the route or with TCAS (traffic air collision system) these inputs would not be applied.
       
    11. john12

      john12 Well-Known Member EngineeringClicks Expert

      Joined:
      Dec 2018
      Posts:
      301
      Likes Received:
      0
      Oh yeah, was that the one in the Alps somewhere?

      I think with the Air France Brazil one - the autopilot realised there was an error and kicked it over to the pilots but both of the pilots were trying to do to different things so they essentially cancelled each other out.
       

    Share This Page

    By using this website you agree to our Cookies usage. We and our partners operate globally and use cookies, including for analytics, personalisation, ads and Newsletters