The US Government wants robots to fight their battles (most technically advanced countries are considering the development of autonomous armies). Anyone can purchase a UAV (unmanned aerial vehicle) on the open market today. What remains the challenge is to create the control system that will define where it goes and what it does and how it does it. The FAA is concerned with how these UAVs will participate in the same air space with commercial piloted aircraft. In military situations, when they (the UAVs) start shooting at people, there will be some people who will want to know why they did (or did not) shoot at specific targets.
One obvious objective of UAVs is to be able to fly from one point to another. This can be facilitated with GPS navigation systems, or the like. It is a more complex objective to have the UAVs navigate around unexpected obstacles (weather conditions, undocumented building, mountains, or other aircraft).
When a human is piloting a plane, that human is constantly watching radar or looking out the windows for other air traffic or hazardous weather conditions. The pilot considers what the obstacles are, where they are, and how/if they should be avoided.
The human pilot considers the capabilities of his/her aircraft and balances the different alternatives to choose the most appropriate action. The pilot considers “general” advice from the FAA (NOTAMs) about how to react to different situations. Each of the alternatives may have different benefits and at the same time will likely have some negative attributes. Some of the drivers may be conflicting; safety versus time, for instance. What is safe enough? In a military situation, there may be consequences for delaying a mission. What about the impact of fuel? Diverting around a large storm may take a long time, resulting in a failed mission. And when deciding on how to avoid the aircraft, the pilot will “assume” the other pilot does the “expected” maneuver. However, the human pilot will attempt to recover, even if the other pilot makes a mistake.
The use of some form of artificial intelligence is required to make judgmental decisions for these autonomous aircraft. I would suggest that it would be ineffective, or impossible, to attempt to “script” solutions to all (or even many) of these judgmental types of situations. These are not decisions about whether to do one thing or another; but “how much”.
A technology like Compsim’s KEEL® (Knowledge Enhanced Electronic Logic) Technology may be the best approach for defining auditable, executable policies that can control these UAVs.
Subscribe to:
Post Comments (Atom)
2 comments:
When a human is piloting a plane, that human is constantly watching radar or looking out the windows for other air traffic or hazardous weather conditions.
What about when the plane's on autopilot? Also, I thought when it comes to weather, air traffic control has a lot of say, both in alerting pilots to weather conditions and telling them how to avoid bad weather.
I guess my real question is how much intelligence needs to be onboard the aircraft vs at a central location (such as air traffic control)?
Information can come from anywhere. One might assume that (eventually) air traffic control, just like aircraft pilots, will be automated (or at least be supported with extensive AI services). There will not be an overnight switch to all autonomous aircraft, however.
Regarding how much is on the aircraft and how much is at a central location: Individual manufacturers of UAVs may do this in different ways. In some very small UAVs (hand launched) the manufacturer (may?) build an intelligent base station that automatically controls the UAV. In larger UAVs with slightly more payload, more intelligence could be located on board.
It is likely that the first steps with piloted aircraft will be to enhance the pilot information displays with enhanced graphics that are automatically generated. This would show the pilots what the autonomous system would do and potentially warn the pilot if he/she chose an alternative maneuver. It will be a race, however, because the demand for UAVs is extensive. They will have to be able to co-exist with conventional aircraft. To do so, they (the UAVs) will have to have the onboard intelligence.
Utilizing artificial intelligence in conventional (piloted) aircraft, even when it is on autopilot, could allow conventional aircraft to automatically provide collision avoidance services. Conventional aircraft could "automatically" navigate automatically around weather systems (per FAA or individual airline policies). The information about the weather systems could come from manned weather stations, or automated weather stations. All of these are opportunities to automatically interpret information and enforce reactions according to an approved policy.
KEEL Technology is 100% explainable and auditable, which might make it the most appropriate way to define and execute these policies (where information has to be interpreted in real-time relative to a changing environment). This could be in the aircraft or used in intelligent stations on the ground.
Post a Comment