MAVRIC - General Software Architecture

Hardware Considerations

Mavric is based on an ActivMedia Pioneer I robot platform.  We have modified the sensory systems considerably.  Currently MAVRIC has four semi-directional light sensors and a 'nose'.  The latter is implemented with a microphone and four narrow bandpass filters to simulate four odors.  The tone frequency represents a specific odorant and the amplitude represents the concentration of the odorant along some presumptive gradient.  As MAVRIC approaches a tone source we can increase the volume to simulate the robot moving up a gradient.

Pioneer comes with seven sonar sensors arrayed around the face.  We are not using these in the more traditional approach to robotics (distance-to-target).  Rather we use the information supplied by these sensors to emulate antenae-like feelers that act as proximity detectors at near approach.

Provisions have also been made to incorporate two semi-directional microphones, one on either side of the robot, to implement audition.  Both 'ears' will be tuned to four additional frequencies, thus emulating the Fourier transform process that occurs in the cochlea.  This will provide four additional sensory features to use in object identification as well as the potential for auditory directional processing (probably in the next generation of MAVRIC).

In total, aside from the seven sonar (distance) we have added sixteen channels of greyscale data to improve sensory information .

Figure 1.  The basic Pioneer robot has been enhanced with eyes, a nose, and (in the future) ears.

Software Architecture

ActivMedia's Saphira Environment

The Pioneer robot is controlled by software running on a laptop computer under the Saphira client environment and interacting with the robot via a radio modem link.  This environment provides a library of routines to provide robot state reflection (getting sensor readings and providing motor output commands) and communications packet control.  Additionally, Saphira provides a synchronous real-time, multi-tasking environment which executes all MAVRIC tasks, including the brain, every 100 milliseconds.  Figure 2 shows the relationships of the various components in the system.

Figure 2.  The Saphira environment is used to provide a real-time, multi-tasking environment for executing MAVRIC tasks each 100 msec.

We have significantly modified the on-board microcontroller-based server monitor to accommodate the additional sensors (see above).  The Pioneer uses an on-board 68HC11 microcontroller to handle all sensor and motor control functions as well as the communications through the on-board radio modem.

MAVRIC's Tasks

MAVRIC is controlled by a very different architecture than is found in more typical behavior-based systems.  Saphira, in fact, comes with a rich library of behavior-based tasks and support functions, which we have largely abandoned to accommodate our own architecture.  Figure 3 shows the basic architecture of MAVRIC along with its relationship to the Saphira state reflector task.

Figure 3.  The MAVRIC software architecture is radically different from other behavior-based architectures.

The central feature of this architecture is the brain task.  The brain is composed of a set of structures called inslots (green rectangles in the figure) which provide the data input to the neural network.  The latter is based on a set of adaptrode-based neurons (see MAVRIC Brain Architecture).  Data is output from the brain via a set of outslots (also green).

Another set of external (to the brain) tasks, called 'body functions' (light blue rectangles) mediate signals that emulate sensory transduction, endocrine function and reflex reactions.  These include:

Sensory and motor tasks interface with the Saphira state reflector via a modified packet structure.
 

This material is based upon work supported by the National Science Foundation under Grant No. IIS-9907102.

Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author and do not necessarily reflect the views of the National Science Foundation.