Transcript No Slide Title

Systems Research in the
Aerospace Engineering and Mechanics
at the University of Minnesota
Gary J. Balas
Aerospace Engineering and Mechanics
University of Minnesota
Minneapolis, MN
[email protected]
SAE Aerospace Controls and Guidance Meeting
11 October 2006
1
University of Minnesota
Aerospace Engineering and Mechanics
Systems Faculty
• Gary Balas , Department Head
– Robust control, real-time embedded systems, flight control
• William Garrard
– Modeling, flight control, parachutes
• Yiyuan Zhao
– Optimization, air traffic control, rotorcraft
• Demoz Gebre-Egziabher
– Navigation, GPS, sensor fusion
• Bernard Mettler
– Real-time control, planning, rc helicopters and planes
2
Current Research
• “Control Reconfiguration and Fault Detection and Isolation Using Linear,
Parameter Varying Techniques,” NASA Langley Research Center, NASA
Aviation Safety Program, Dr. Christine Belcastro Technical Monitor
• “Stability and Control of Supercavitating Vehicles,” ONR, Dr. Kam Ng
Program Manager
– Special Session at the 2006 AIAA Gudiance, Navigation and Control
Conference entitled “Modeling and Control of High-Speed Underwater
Vehicles.”
 “Development of Analysis Tools for Certification of Flight Control Laws,”
joint work with Andy Packard at UC Berkeley and Pete Seiler at
Honeywell. This research is being funded by AFOSR.
• Workshop on “Real Time Control of Hybrid Systems: Design,
Implementation, Verification, and Validation” in Budapest June 27-28,
2006 sponsored by NSF, Hungarian Academy of Science and Unisaino,
Benevento, Italy.
3
Navigation and Guidance
Research
Demoz Gebre-Egziabher
Department of Aerospace Engineering and
Mechanics
University of Minnesota, Twin Cities
High Integrity Navigation
• Design and validate high
integrity navigation
systems for the DoDs
Joint Precision Approach
and Landing System
(JPALS)
–
Precise
over
bounds
on
Tail-Hook Target = 3’ x 3’
navigation errors.
box which moves with
the aircraft carrier
– Fault detection and
isolation algorithms
JPALS Performance Specifications:
– Methods
for fusing
Accuracy = 1.1 m Vertical Alarm
Limit
information
from
multiple
Integrity = 10-7 Probability Hazardously
Misleading
Information
navigation sensors (GPS,
Time to Alarm = 1 sec.
INS, baro-altimeter)
• Sponsor: Lockheed
5
UAV/RPV as ITS Sensor Platforms
Controlled Airspace
Boundaries (Blue)
Synthetic Vision Display developed at U of M
as part of this research for remote operation
situational awareness of small aerial vehicles.
Display fuses satellite imagery and a GIS data
base with an open source flight simulator
• Explore capabilities
enabled by
Uninhabited Aerial
Vehicles (UAV) or
Remotely Piloted
Vehicles (RPV) in
Intelligent
Transportation
Systems (ITS)
applications
– Data gathering,
surveillance.
• Develop “turn-key”
sensors and systems6
which enable use of
Micro- and Nano-Satellite Design
Minnesat
GPS Antenna
Solar Cell
• Design of systems and
algorithms for ultra-short
baseline GPS attitude
determination systems for
micro- and nano-satellites.
– Baselines on the order of
one-wavelength.
– Non-aligned antenna
arrays.
– Modification of COTS
components
• Sponsor: US Air Force
Research Labs, Space
7
William L. Garrard
Dynamics and Control of Aerospace
Vehicles
Flight Mechanics of Parachute
Systems
8
Prediction of Parachute Opening
Dynamics
9
Azimuth Pointing Control
Of A Balloon-Borne Stabilized
Platform
10
Guidance and Control of Autonomous Vehicles
October 6, 2006
Bernard Mettler
Assistant Professor
Department of Aerospace Engineering and Mechanics
University of Minnesota
Requirements for Autonomous Vehicle Operation
•
•
•
Mobility and agility are fundamental to autonomous vehicles
–
negotiating complex terrain environments
–
handle difficult operational conditions (i.e. atmospheric disturbances)
–
making full use of vehicle dynamic capabilities
Involves interplay between lower-level flight control and higher-level trajectory
planning
Requires technique that integrate local and global scales
Receding Horizon Trajectory Optimization with a Cost-to-go Function
•
•
•
•
Principle
– perform trajectory optimization over finite horizon T
– capture discarded trajectory tail with a cost-to-go (CTG)
J(x(t+T))
Provides a rigorous framework to combine on- and offline
optimization
– Offline (near real time O(secs)): computation of CTG
– Online: computation of the control action
Goal: Approach the performance of infinite horizon optimization
with less computational burden
Key question: how to compute the CTG function ?
Cost-to-go Computation
•
•
Finite-state model to capture vehicle maneuvering
capabilities. Example:
– Quantized speeds: v1, v2, …
– Capabilities at each speed represented by feasible turning
and linear accelerations
– Discrete headings (π/4 resolution)
Compute cost-to-go function with Dynamic Programming
Examples cost-to-go Maps
goal: circle “o” heading North
Open space CTG
Space with Obstacles CTG
Trajectories based on CTG
Key Features
•
Computationally efficient:
– CTG computed O(secs)
No special environment model required (works with digital elevation map)
Captures global features of planning problem from standpoint of vehicle capabilities
– Environment is “resolved” by characteristic vehicle motions
– Provides a way to partition the environment (virtual roadmaps)
Accurate CTG map can be used directly with a control policy
Provides a framework to study interplay between spatial constraints and dynamic
behavior
•
•
•
•
Ongoing Work
1.
2.
Find adequate level of approximations in CTG computation,
i.e., given vehicle capabilities and operational requirements,
determine:
CTG Partitions/Structures
–
Resolution of cost-to-to map, number/type of motion
primitives
–
Length of the optimization horizon and update interval
Exploit global CTG structures
–
–
3.
4.
Partitions and “virtual roadmaps” are based on vector field
characteristics
Enable high-level decision making (tactical)
Develop efficient algorithms for online optimization
Robustness to stochastic effects
–
Uncertainty about environment knowledge
–
Disturbances
Indoor Flight Experiment Facility
•
Vision based tracking system used for aircraft
positioning and attitude
–
–
•
Indoor flight lab setup
No on-board instrumentation required
Makes possible to use micro helicopters and airplanes
Create controlled experimental conditions
–
–
–
Scenarios with real or virtual 3D environments
Jets to simulate wind disturbances
Interactive human-autonomous experiments
Other Aircraft under Investigation
Micron FP Helicopter (50g)
Plantraco Butterfly (3g)
Test Helicopter with cameras
E-Flight Blade CX (200g) with
Vicon’s MX 40 cameras
University of
Minnesota
Distributed
Optimal
Dynamic Optimization Across Airborne Networks
UAV Flights in Wind
16
Distributed Dynamic
Optimization Across Airborne
Networks
 Airborne Networks: Ad Hoc Wireless Networks
Among Aircraft
 Each Vehicle Calculates Its Own Optimal Flight
Trajectory


To optimize a performance index, to achieve a certain team task
To avoid conflicts with each other
17

Optimal Unmanned Aerial
Vehicle
Flight
in
Wind
Potential Benefits of Wind Energy Utilization in UAV Flight

Reduced fuel consumption/prolonged flight/increased range
Wind Gradients
Thermals
Mountain Waves
18