Unmanned vehicles

1
Unmanned vehicles?

• Concept manned vehicles your car, remotely
  controlled vehicles your RC toy car, unmanned
  vehicles autonomous car
• Examples drones, airplanes when on autopilot,
  Unmanned Ground Vehicles, Autonomous Marine
  Vehicles

2
Why unmanned vehicles?

• Dangerous mission, keeping humans out of harms
  way
• Long drawn out missions, not necessarily
  requiring constant human monitoring

Applications

• Military scouting hostile environment (air,
  ground), mine countermeasure
• Civil oceanographic or atmospheric surveys
  for monitoring (oil platforms) or data
  collection (ocean floor mapping), missions in
  hostile environments (Mars rover)

3
Motion control of unmanned vehicles

• Concept autopilot, a program on an on-board
  computing device
• How does it work?
• The motion controller monitors and attempts to
  minimize the difference between desired
  trajectory and measured trajectory, providing
  appropriate control command to the vehicles
  actuation system
• How do we construct this motion controller?
• The control design is based on a stability
  analysis of the closed loop system, Lyapunov
  stability theory

4
Challenges

• Uncertain systems
• What are f( . ) and B exactly, how accurately
  can we estimate them?
• Actuator limitations How much torque do we get
  out of these motors? How quickly can we
  change this torque?
• Limited computational power Can embedded
  computers handle this?
• Discrete time environment Algorithm derived
  assuming a continuous environment, an
  embedded computer is discrete

Solutions

• Adaptive control technique, Neural Networks
• Model Reference Adaptive Control (MRAC)
• Dynamic Surface Control
• Inter-sample state estimation, rate saturation
  of the control command

5
Actuator amplitude and rate saturation constraints

• MRAC, concept the actual vehicle tracks a
  virtual vehicle, which tracks the desired
  trajectory
• When the actuation system reaches its limits,
  the control command saturates.
• What do we do about it? The virtual vehicle
  waits for the actual vehicle
• Practically, we reshape the reference trajectory
  in real time, which affects the transient of
  the reference system. Stability properties are
  however preserved.
• Numerical simulation Liénard
  system (Van der Pol-ish oscillator)

6
Dynamic Surface Control (DSC)

• Embedded computers have limited computational
  power
• Motion controller can be complicated algorithms,
  computationally intensive

Issue the more computationally intensive the
algorithm, the slower the control loop, the less
reactive the system, which leads to deterioration
of tracking performance

• Solution we streamline the algorithm design
  through filtering, retaining exclusively the
  meaningful frequencies of the involved
  quantities, thus greatly simplifying the
  algorithm, and allowing the control algorithm to
  operate at reasonable frequencies in spite of
  limited computational power

7
A Discrete Environment

• Inter-sample behavior
• - control algorithm
• - measurements
• - inter-sample estimation linear or
  polynomial interpolation, Kalman filtering

• Inter-sample behavior
• - control algorithm
• - actuation system
• - algorithm needs to account for the command
  remaining constant in-between sample times
• - rate saturation

8
Neural Adaptive Control Algorithm

• System dynamics and control command
• compensates for the system dynamics
• addresses the error dynamics

• Classically, contains arbitrary
  activation functions
• To improve performance, contains activation
  functions reflecting our knowledge of ,
  supplemented by arbitrary activation functions

9
Numerical Simulation Result

• Application Unmanned Aerial Vehicle, quadrotor

Future Work

• Comparison to alternate control methods
• Implementation on actual vehicles