|
Humaniform Control System
When compared to modern actuators
human muscles are imprecise, sensory feedback from them is sloppy
and yet in combination a human arm outperforms any artificial robot
arm…….
The difference is the basis of the
Humaniform Robotics approach. The Humaniform Muscle enables a
different type of control system to be used and conversely the
Humaniform Control System allows a different type of actuator to be
used.
Neural networks are now well
established as part of control systems theory, although invariably
they are used to implement more robust versions of traditional
control algorithms. Many control algorithms are sequential in nature
and are suited to the sequential architecture of modern computers.
The human brain, however is entirely different, benefiting from
massive parallelism and new types parallel processing algorithms.
The Humaniform Control System is a hardware and software
implementation of the type of scaleable parallel control algorithms
that are used as part of the human motor control system.
The HCS forms a vital component of
the Humaniform Arm Project. To design the HCS system we had to
develop a new suite of hardware and software tools (DIGILOG
toolset). The HCS incorporates digilogtm (hybrid
digital/analog processing elements) technology.
Shown below is an example of the HCS
Simulation software that depicts a small population of digilog
elements, shown as part of a larger 3-d structure.
The HCS is a fully scaleable control
architecture suited to complex control problems.
Applications
The Digilog toolset is capable of
modelling and simulating any neural network model. Simulations give
a good indication of the dynamic behaviour of a network.
The Humaniform Control System is
designed to control a large number of parallel actuators. In
advanced robotic systems this is increasingly the major challenge
facing researchers. |
