AEM faculty spotlight:
Bernard Mettler
After driving the same route home time and again, an autopilot of sorts tends to set in. Not to be confused with Highway Hypnosis, where an individual loses driving ability, this autopilot allows for good decision-making though the driver may be unaware of exact details of the drive home after the fact. But beyond the norm, the human on autopilot is able to snap to awareness should the situation demand it. Bernard Mettler, an assistant professor in AEM, hopes to bring insight of these phenomena into the world of autonomous vehicle through his study of the pilot-vehicle interaction. Below is a conversation with Professor Mettler where he discusses dynamic AI, control for micro aerial vehicles, and applying the human mind to flight control.
Bernard Mettler
What types of research interest you?
The general research area I’m interested in is autonomy, mostly for aerial vehicles. Aerial vehicle control is the area where robotic technologies will have the most application in the future. However, there are challenges that need to be solved to implement some of these applications. One is understanding how to fly the vehicle without any human operator. We need to replace the pilot that we’ve known for 100 years with a computer.
How do we replace the human pilot in flight?
There are some fundamental questions that we need to solve. The main challenge has to do idea with the idea of structured operations. We’ve been quite successful in solving autonomous flight problems for areas of flight where the operation is structured. When you land an airplane, there are procedures that can automate the process. If the computer were to fly the aircraft, there are a finite number of things that could happen where the computer would respond appropriately. But if you think about the human and what we do in our daily life, there are lots of mundane situations where we control our car or bicycle, for example, that don’t have this structure. We need to be creative in how we solve many problems. Sometimes we’re not really conscious – but these situations are not trivial to implement on a computer. This is what goes into artificial intelligence, giving the flight computer intelligence so it can handle more than a finite set of scenarios.
How do you give flight computers intelligence?
Once you have different kinds of information available, you need to represent tasks to achieve your goals. Some tasks are trivial to represent – like landing on air strip. The task can be described by going to a specific location using constraints on speed and direction, but some tasks are not describable so easily. Think of the problem of reconnaissance. You are going to find some object or person in an area. You need to formalize that mathematically. This becomes much less trivial.
How does actually flying the aircraft factor in to this problem?
Even if you knew your environment, flying in an appropriate way is something that cannot be done without a lot of computational effort. This is a major issue. A lot of the problems that can be formulated mathematically do not lend themselves to algorithms – that leads to my other direction of research.
Would you talk a bit about this area?
Piloting skills are something that we don’t understand well when we study the human. People have worked a great deal on that, but in recent years with advances in brain sensing tools, there are lots of opportunities that allow us to better understand how humans perform. Maybe we can learn from the brain how these different problems are addressed and base our algorithms on that. We are trying to understand autonomy from an algorithmic and mathematical point but also from the human perspective.
What needs to happen to implement all these problems associated with autonomous flight?
There is a big leap if you want to go from autonomously operating aircraft in a structured fashion to more complex environments and conditions. These tasks are – if you think about search and rescue operations for instance – difficult problems. It challenges our methodologies that we’ve used in the past. That’s where we’ll have to really go through some form of a revolution. It may be difficult to realize how complex these tasks are because, as humans, we tend to take a lot of things for granted – like driving a car. It’s trivial 90 percent of the time, but there is that 10 percent where you have to be intelligent. To be able to move in that area, we’ll really have to develop some fundamentally new tools.
Last Modified: Thursday, 11-Oct-2007 14:25:04 CDT -- this is in International Standard Date and Time Notation