FFMP: detailed description of the 8 phases

This page changes constantly, it's my scratch area! If you have some comments or ideas, please let me know!


Phase 1: Standing still in air with automatic self stabilization

Build a small and quiet entity (unobtrusive), which is able to stay still in the air: hovering and automatically self-stabilizing. Can move (fly) in three-dimensional space. Make it as autonomous as possible.
Back to phases overview | Back to FFMP homepage

Technical specifications

  • Propulsion: 3 or 4 micro electric ducted fans (impellers)
  • Power: Wires, solid-state lithium ion batteries. Later: thin film lithium batteries, or Wireless Power Transmission (WPT)
  • Auto stabilizing, precision guidance: A helicopter, real or model, is very unstable due to its construction principles. To fly it means first to compensate all the internal and external forces which try to get it out of balance. To automize this process, a lot of sensors are necessary.
    There are two possibilities: relative position measurement (e.g., inertial navigation), and absolute position measurement (e.g., GPS, or active beacons on laser, ultrasonic, or radio-frequency basis)).
    • Relative position measurement: Inertial sensors give acceleration information for angular motions (vertical, longitudinal, lateral) ((Translational-motion or rotational-motion inertial sensors, means accelerometers or gyroscopes? Piezoelectric vs. piezoresistive vs. capacitive?))
    • Absolute position measurement by LPS (Local Area Positioning system, similar to GPS): 3-4 wireless mini transmitters, stuck somewhere on the walls of the room (1 for security). Possible realization: trilateration with active ultrasonic beacons. Depending on how many points of the FFMP are computed for their absolute position, heading sensors like compass and inclinometers might be necessary too. Ultrasonic measurement seems to be the only real option, because beacons on laser basis require direct visual contact, and RF (radio-frequency) does not seem to work reliably indoors (Borenstein, Everett, and Feng, 1996, p. 65). Accuracy requirement: 1 - 5 mm. If the FFMP "floats" away due to internal instability or external influences (airflow), it tries to return to its initial absolute position.
    • Hybrid sensing: combination of both! Outlook: hybrid sensor 3 axis acceleration with GPS: today 8 in. and 3 W power consumption; in 2005: 2 in. and 0.5 W.
  • Security:
    • soft edges
    • airbag, in case of failure of engines
    • large button to shut off all engines (automatically switching off all engines in case of collision?)
  • Ground station: should get obsolete later, all functions should be on board!
    • receiving AV stream
    • translating indirect voice commands
    • logging all activities and behavior
    • making connection to Web
    • making connection between different FFMP

What's the current state of autonomous helicopters? Interesting links:

  • Smallest helicopter with 4 rotors:
    Gyrosaucer II E-570 by KEYENCE.

    It is the smallest commercially available 4-rotor electro helicopter, as far as I know. Unfortunately, it is not available in the States. The only place outside Japan where you can get one is Intertronics in Germany. Or by mailorder from an R/C model shop in Tokyo called TENSHODO (which turned out to be the cheaper way). It is NOT an autonomous heli, but I guess it would be easy to start with this machine and to modify it!
  • Hovering automatically with 4 rotors: I know only of two projects which deal with that:
    • Hoverbot (here's another page) by Johann Borenstein at University of Michigan. The HoverBot paper (unpublished draft, PDF format) is a very comprehensible and concise introduction to the problems of 4-rotor design helis. There is also a slideshow, as well as a short movie about the HoverBot. Although yaw, pitch, roll, and elevation (motion in the z-direction) were controlled and stabilized by the computer, HoverBot was confined in x and y direction by a support fixture. The control system is quite complex, because not only thrust of the electro motors is controlled, but also rotor pitch. Unfortunately, this very interesting project was discontinued 1992 after only 3 months (funding reasons). Main difference to what I plan to do is size and complexity: HoverBot is much bigger than what I have in mind; I do not plan to use rotor pitch (mechanically complex); Hoverbot has a set of different sensors like accelerometers, gyroscopes, compass, and range detector: I will probably go with only high precision absolute position measurement.
    • Hovering Platform ("Schwebende Plattform"), a 1998 summer term project at the Automatic Control Laboratory (IfA, Institut für Automation) of the Swiss ETH in Zürich. In a previous thesis work, a hovering platform with four electro motors was constructed. An onboard compass module and two inclinometers measured the heading of the platform, and the speed of the motors was controlled to stabilize the heading of the platform. In the new project, the position has to be controlled too. So the project will be about evaluating appropriate sensors like DGPS (differential GPS) or ultrasonic transducers, as well as designing controlling concepts and testing the prototype in free flight. Urs Baumann did this very interesting project, and he calls it Flying Platform ("Fliegende Plattform"). It is mainly about the modeling of the systems dynamics and the design of the controls for stabilizing the platform. He uses three sensors: ultrasonic absolut position sensing (VS-100 by Litec, Inc.); a hybrid sensor which measures roll and pitch acceleration and inclination and yaw angle with a geomagnetic sensor ( TCM2 by Precision Navigation); additional roll and pitch gyro (Gyrostar ENC-05 by Murata). Free flights of a few seconds duration were conducted.
      Another related project (IfA03) uses a simple bar with two motors and 2 DOF to study the MIMO system of a possible autonomous electric helicopter. Here are the manuals for the system as well as for the helicopter modell (partly .PDF, partly .PS format).
  • Aerial Robotics competitions: Here the most successful autonomously flying robots meet:
  • Conferences: First International Conference on Emerging Technologies for Micro Air Vehicles
  • Robot Blimps: Because they are mechanically much simpler than planes or helicopters, they would be much easier to make "stand still in the air." This would be a really elegant solution! Unfortunately, due to their underlying construction principle, they can not be small.
  • Position measurement, relative and absolute: The paper/book/CD-ROM Where am I? -- Systems and Methods for Mobile Robot Positioning, edited and comiled by Borenstein et al., is a very intersting source for all sorts of sensors and techniques for robots.
  • The Autonomously Flying Helicopter of the Measurement and Control Laboratory (IMRT) at the Swiss Federal Institute of Technology (ETH) in Zürich, Switzerland.
  • The DV8 of the University of California, Berkeley. John Koo seems to be interesting.
  • Inertial proprioceptive devices: Self-motion-sensing toys and tools: Chris Verplaetse's paper is about proprioceptive devices which have a sense of their own motion and position. He differentiates absolute position from inertial sensing, and explains the advantages and disadvantages of each, e.g., inertial sensors do not need any external reference, but position errors accumulate. He describes a general system with inertial sensors, where the MIMO unity is not necesseraly pre-programmed, but uses Kalman filter and pattern recognition in a system with neural networks or Hidden Markov models. Finally, he describes a few commercially available inertial sensors (accelerometers and gyroscopes).


  • Phase 1a: testing and evaluating Keyence's Gyrosaucer II E-570.

    • Controlling concept: Elevation, roll, pitch, and yaw are controlled by adjusting the speed of the 4 motors/propellers (two rotating clockwise, two counterclockwise). Through this construction, Gyrosaucer has 4 degrees of freedom (DOF): up/down, sideways, forward/backward, and horizontal rotation (vertical axis). These are controlled manually through a 4CH remote control, with limited support from the built-in gyroscopes.
    • Sensor concept: How is the efficiency of the free gyro and the rate gyro?

  • Phase 1b: building simple FFMP with exobrain and umbilical cord
    The mechanically simplest construction, which enables automatic hovering. It consists of:

    • 4 motors/propellers: ducted fans or impellers
    • single sensor: absolute position sensing
    • controller: processing outboard on a PC ("exobrain")
    • no batteries: power for motors through wires ("umbilical cord")
    • Automatic stabilization: Is a absolute position sensor enough for stabilization, or are gyroscopes, inclinometers, and compass necessary as well?
    • Which are the options for absolute position sensing in 3D space?
      • Ultrasonic. On FFMP: ultrasonic microphone. Within action range (room): 3 ultrasonic loudspeakers (or rather: three clusters) which emit pulsed signals (different frequencies and sequences). The microphone receives these signal and sends them back to ground station (wire, analog signal), where the signal is parsed and time delays between original sequences and received signals are measured. With these three time delays, the absolute distances are computed. After initialization ("learning") the system is able to describe the absolute position of the FFMP in 3D space.
      • RF based. On FFMP: 3 RF receivers (antennas), placed on the most extreme parts of the FFMP. Within action range (room): radio frequency emitter. The emitter sends a high-frequency signal. The signals of the 3 antennas are compared for their phase. From the phase difference, the absolute distance is computed, and from there the absolute position in 3D space. (Weather this is made on board or on ground station is not yet clear).


Phase 2: Passive vision

Adding a wireless camera for conveying pictures from otherwise impossible camera angles.
Back to phases overview | Back to FFMP homepage

Technical specifications

  • Payload capability: must be able to carry a lipstick camera with transmitter

What's the current state of wireless micro cameras and aerial photography? Interesting links:


Phase 3: Listening

Listens to simple verbal instructions of humans like Up! Down! Turn left! Zoom in!
Back to phases overview | Back to FFMP homepage

Technical specifications

  • Possible speech commands in phase 3, most probably indirect commands over a remote listening device like a walkie-talkie:
    • linear movement commands: up, down, left, right, forward, backward
    • turning commands: turn left, turn right (tilt forward, tilt backward)
    • amount commands: slower, faster, stop
    • camera commands: zoom in, zoom out

What's the current state of speech recognition in small appliances? Interesting links:

  • Traditional speech recognizers


Phase 4: Active vision, simple tasks, simple morality

Implementing intelligence so that it behaves like an simple artificial life form: Human and non-human obstacle avoidance, evasive behavior, performing tasks like Come here! Leave me alone!, very simple moral prime directive Do not harm anybodyor anything.
Back to phases overview | Back to FFMP homepage

Technical specifications

  • Active vision and situational awareness: Environmental sensors, mainly distance sensors like sonar, infrared, ultrasonic, and tactile, for obstacle avoidance; perhaps radar (for a more spatial awareness). They enable the FFMP to move away from things or people coming to close.
  • Possible additional speech commands in phase 4, most probably direct commands which FFMP recognizes itself (without its remote listeing device):
    • Go away! or Shhhhh!: FFMP should move to free space
    • Come here!: FFMP approaches very carefully the caller up to an arm length. Automatic shut off of all engines when someone grabs it within 15 seconds after this command!
  • The above two points should lead to a Simple Morality in the form of the Prime Directive Do not harm anybody or anything!

What's the current state of sensors of interest for FFMP? Interesting links:


Phase 5: Complex tasks

Behaves like an artificial pet: understands complex verbal requests like Follow me! Follow this man in a distance of 5 meters! Give me a close up of person XY!
Back to phases overview | Back to FFMP homepage

Technical specifications

  • Vision processing as well as
  • Natural language processing for complex commands like:
    • Follow me!
    • Follow this man in a distance of 5 meters!
    • Give me a close up of person XY!

What's the current state of vison and natural language processing which might be of interest for FFMPs? Interesting links:

  • SmartCam by Claudio Pinhanez: "A SmartCam is a robotic TV camera which can operate without a cameraman, changing its attitude, zoom, and position to provide specific images upon verbal request from TV director."


Phase 6: Adaptation to environment, emergent robotic behavior

Create an adaptive autonomous robot, means, let it learn from the interaction with the environment about dangerous objects and situations, as well as about its power management and flying behavior.
Back to phases overview | Back to FFMP homepage

Technical specifications

What's the current state? Interesting links:


Phase 7: Use of tools

Modify so that it can use external physical tools for simple self repair and (symbolic) reproduction.
Back to phases overview | Back to FFMP homepage

Technical specifications

What's the current state? Interesting links:


Phase 8: Intellect and cross cultural morality

Improve intelligence up to Artilect stage (artificial intellect, ultra-intelligent machine), as well as implement an Artificial Multi Ethical Advisor System (AMEAS, Cross Cultural Ethical Knowledge) to make sure its behaviour is always ethically correct.
Back to phases overview | Back to FFMP homepage

Technical specifications

What's the current state? Interesting links:

  • A CNN article about the problematic aspects of ultra-intelligent machines, the Robot Armageddon.
  • Artilect: Hugo de Garis paper about The 21st. CENTURY ARTILECT: Moral Dilemmas Concerning the Ultra Intelligent Machine. In this phase, the mobot's intelligence should be on a high level, ideally on an Artilect stage. "An Artilect ('artificial intellect'), according to Dr. Hugo de Garis, is a computer intelligence superior to that of humans in one or more spheres of knowledge together with an implicit will to use the intelligence. Artilects are the concern of artificial intelligence specialists (or 'intelligists') like de Garis, who speculates that human society may soon have to face the question of whether and how we can restrain artificial intelligence from making decisions inimical to humans."
    "Dr. de Garis assumes that within one or two generations, we will have computers that are more sophisticated than human brains with the ability to experimentally evolve their intelligence into something much beyond what humans might contemplate or understand. De Garis wonders whether such machines would consider human beings important enough to preserve. He speculates that society will soon need to face the question of whether we should permit Artilects to be built. He foresees two factions arising: the Cosmists, who argue that they should be built, and the Terras, believing that they should not. The Cosmists might believe that Artilects would probably want to leave our planet to seek intelligence elsewhere in the universe. The Terras believe that it would be too dangerous for the human race to allow Artilects to be developed." (Taken from What Is...an artilect (a definition).)
  • Artificial Multi Ethical Advisor System (AMEAS). The second important goal of phase 8 would be to connect or add an Artificial Multi Ethical Advisor System (AMEAS). At this stage, an FFMP has already a considerable amount of autonomy. Therefore, it should be aware of all consequences of its behavior. For this purpose, I suggest to develop and implement an Artificial Multi Ethical Advisor System.
    This idea was originally inspired by a Science Fiction movie in which the most ethical being turned out to be an android (quote from Alien Resurrection: "No human can be that humane!"). Although this situation might sound not very plausible, it is actually not so far fetched if one takes a closer look at the problem. The domain of ethics is not as fuzzy as one might expect. Based on my ethical-philosophical studies at the University of Bern (my minor was Ethics), I think that it is theoretically possible to implement the whole domain of ethics in an expert system. The reason why I think it is possible is that most ethical systems are based on some sort of rule based system anyways. One of the problems that has to be solved might be related to the fact that most ethical systems are described using a proprietary and incompatible terminology-learning and understanding this terminology is an important step towards understanding the ethical system itself. Therefore, it is possible that ethical constructs of different ethical systems seem to be incompatible, but are the same, just described in different words.
    Such and expert system should be able to give advice on complex ethical questions, considering not only one ethical system, but several. An expert system can contain several ethical positions in parallel. Since there is no "claim for truth" anymore in modern ethical discussions, ethical advice from this system would not be a "right or wrong" answer, but more like suggestions towards a decision as opposed to another one.
    Given a simple rule based ethical system like the 10 commandments, a very simple AMEAS would be able to provide pieces of advice like Do not kill. However, such an answer would not be specific enough. Therefore, part of an AMEAS would be to provide an understandable explanation for this rule, and based on that, practicable advice in today's world.
    Obviously, a simple question for advice has to be submitted together with extensive information about the personal and general situation of the asking person, since ethical advice may depend crucially on individual situational circumstances.
    I expect that most people will dislike deeply the idea of "being said by a machine what I have to do, what is good and what is bad for me." However, the AMEAS is not meant to patronize human mankind. Every single person has to be kept responsible for the consequences of his/her decisions-this fact will never change. But in today's multi cultural world where our behavior easily can have global impact, it is nontrivial to consider the consequences of our behavior. E.g., what is appropriate to be put on a homepage on the World Wide Web? What might be perfectly appropriate for one culture might be highly offensive and blasphemous in another one. Therefore, a Multi Ethical Advisor System can give advice and explain why one decision might be more favorable than another. However, the asking person does not necessarily have to follow this advice if s/he thinks that it is not appropriate. AMEAS just enables people to ask several competent and accepted philosophers like Kant, Socrates, etc. for their advice on an actual ethical problem.
    Not much work has been done to set up an AMEAS. However, computer based ethical reasoning systems are described by J. Gips, "Towards the Ethical Robot," in Android Epistemology, K. Ford, C. Glymour and P. Hayes (eds.), MIT Press, 1995. Here's a very interesting discussion of it.
    Interesting is also a case-based knowledge representation for practical ethics by Ashley, K. D., & McLaren, B. M. (1994). A CBR Knowledge Representation for Practical Ethics. In the Proceedings from the Second European Workshop on Case-Based Reasoning, November, 1994, Chantilly, France. To be published in M. Keane (ed.) Lecture Notes in Artificial Intelligence, Springer Verlag: Berlin.
    In the context of FFMP, having an AMEAS available means that an FFMP can be aware of the consequences of its behavior. If carrying out a request of a human being would violate clearly ethical rules in most ethical systems, the FFMP has to refuse this request. It would be interesting to see if such an FFMP would accept a military mission at all.

    Note: Keith Emnett and I think that after Affective Computing, the next big thing could be Ethical Computing. It has a huge potential for controversy, and it touches all people at a very sensitive point: We might accept that machines are stronger, faster, and more intelligent than humans are. The only thing that makes us superior to them is our possibility of a moral awareness. If we lose this domain too, we might lose our position as the most evolved creature on this earth. And intuitively, everybody knows that.

Back to FFMP homepage

Send me some comments! Stefan Marti Last updated Dec 20 1999.

Copyright © 1997 - 2007 by Stefan Marti and MIT Media Lab. All rights reserved.