- Learning in Control Engineering
- Control Research and Development at IIS Corp:
- Applied to Cart-pole Balancing
- Applied to Attitude control of the Space Shuttle
- Applied to Tether  Control of the Space Shuttle
- Applied to Pitch Control in the Space Shuttle Training Aircraft (STA)
- Biologically-Inspired Robotics

• Two 2 Meters long rails
• A Cart that moves left or right
  
• A computer that issues a command left or right
  
• A Pole that is hinged to a Cart
  
• The system learned in 8 trials using GARIC

A Shuttle attitude controller is expected to perform four basic operations: 

• Attitude Hold or maintaining the desired attitude within a small region of the desired value, typically known as a deadband. 
• Attitude Maneuver or going from one attitude to another. 
• Rate Hold or maintaining a desired rate on a given axis. 
• Rate Maneuver or going from one rate value to another rate value for a given axis.

Its on-orbit controller or Digital AutoPilot (DAP) is based on modern digital control theory and is a highly optimized controller.  It uses two types of thrusters (two levels of jet thrusts), known as primary and vernier, and operates with two different sets of deadband values.
It can perform rate maneuvers in pulse as well as discrete modes.  Typical perturbations acting on the system included gravity gradient, aerodynamic torques, and translational burns.

All Attitude maneuvers were controlled using vernier jets.  In our development we had 3 GARIC controllers for fine tuning membership functions along the three rotational axis, i.e., Pitch, Yaw, and Roll.  We were able to maintain deadbands of 0.5, 0.4, 0.3, and 0.2 deg in the tight deadband region without excessive fuel consumption.

Translational Control

V Bar Approach

V bar Approach is the approach of the space shuttle along the velocity vector, towards a satellite, such as the Solar Max.  This approach is initiated at 400 ft from the satellite and moves into stationkeeping mode as it approaches 50 ft.
All translational maneuvers are controlled using primary thruster only.  In our development we had 3 GARIC controllers executing simultaneously to account for the three axes, i.e., Azimuth, Elevation and Range.

R Bar Approach

R bar Approach is the approach of the space shuttle towards the satellite (SolarMax) along the radius vector. The approach is initiated at 400 ft from the satellite and on reaching 50 ft, the space shuttle starts stationkeeping till the end of observation mission.
All translational maneuvers are controlled using primary thruster only. In our development we had 3 GARIC controllers executing simultaneously to account for the three axes, i.e., Azimuth, Elevation and Range.

Fly Around Approach

Fly Around Approach is the approach from the vbar to rbar, i.e, a quarter circle. The distance to the satellite (Solar Max) is maintained constant.
All translational maneuvers are controlled using primary thruster only. In our development we had 3 GARIC controllers executing simultaneously to account for the three axes, i.e., Azimuth, Elevation and Range.

We investigate the possibility of using Generalized Approximate Reasoning Based Intelligent Control (GARIC) for tether control.  Tether control consists of three main operations 

• Deployment phase: A conducting tether is used to deploy a payload, e.g., Italian satellite weighing 525 kg, deployed to a distance of 20 km.
  
  
• On-station phase: Acquire scientific and operational data.
  
  
• Retrieval phase: Retrieve upto 2.4 km and dampen oscillations.  And finally completely retrieve the payload for reuse.

• In vacuum, zero-g, and under gravitational and magnetic forces.
  
  
• Time varying dynamics of a long, flexible, variable length tether, the orbiter and the payload.
  
  
• Unlike Tether length and Tether tension, oscillation can't be directly measured or controlled.

• Tether oscillation can be indirectly manipulated through tether's length maintenance.
  
  
• During retrieval phase maintain length and length rate of the tether by controlling the voltage applied to the motor.

• Tether Control - Massless Model: An approximate model of tether, here the tether is assumed to be massless. Learning was done during the deployment phase (16200 secs) using GARIC to tune the membership function.
  
  
• Tether Control - Finite Element Model: In the Finite Element Model (FEM), the tether is approximated using 10 beads.  Each bead represents concentration of mass corresponding to a fixed length.  Here we use the use the fine tune labels obtained after learning with GARIC, for testing purposes only.  Testing of the FEM based tether control was done during the retrival phase

The Space Shuttle is both a spaceship and an airplane, hence when training a shuttle crew member, both spaceflight and atmospheric flight must be taught.  The Shuttle Training Aircraft (STA) is NASA's flyable training vehicle that duplicates the Shuttle's approach profile and handling qualities in order that an astronaut pilot could see and feel many simulated landing before attempting an actual Shuttle landing.
STA is a modified Grumman Gulfstream II aircraft with an onboard special computer system to enable the aircraft to simulate the orbiter.  The STA provides orbiter pilots with a realistic simulation of Orbiter cockpit motion, cues, and handling qualities, while simultaneously matching the orbiter's atmospheric decent trajectory from 35,000 ft to the actual Orbiter cockpit height above the runway at touchdown.  There is an onboard computer namely, Advanced Digital Avionics System (ADAS), which during simulation mode controls the Direct Lift Control (DLC) and the in-flight reverse thrust, as well as the conventional aircraft control.

IIS BioBots - Robots that walk and run. It is the goal of a group of engineers and scientists at IIS Corp. to develop these IIS BioBots.
IIS BioBots use artificial muscles and stretch sensors instead of heavy motors and position encoders that are commonly used in traditional robots. 
Biologically Inspired technology (Patent Pending) is used to enable model-free intelligent control and learning in these BioBots. 

“Evolutionary designs” and “Intuitive rules” are the key words in this Biologically Inspired technology.