Type of Document Dissertation Author Caldwell, Charmane Venda Author's Email Address firstname.lastname@example.org URN etd-06172011-144252 Title A Sampling-Based Model Predictive Control Approach to Motion Planning for Autonomous Underwater Vehicles Degree Doctor of Philosophy Department Electrical and Computer Engineering, Department of Advisory Committee
Advisor Name Title Emmanuel Collins Committee Co-Chair Rodney Roberts Committee Co-Chair Linda DeBrunner Committee Member Dave Cartes University Representative Keywords
- Autonomous Underwater Vehicle
- Model Predictive Control
- Sampling-Based Method
- Motion Planning
- Path Planning
Date of Defense 2011-04-21 Availability unrestricted AbstractIn recent years there has been a demand from the commercial, research and military
industries to complete tedious and hazardous underwater tasks. This has lead to the use of
unmanned vehicles, in particular autonomous underwater vehicles (AUVs). To operate in
this environment the vehicle must display kinematically and dynamically feasible trajectories.
Kinematic feasibility is important to allow for the limited turn radius of an AUV, while
dynamic feasibility can take into consideration limited acceleration and braking capabilities
due to actuator limitations and vehicle inertia.
Model Predictive Control (MPC) is a method that has the ability to systematically handle
multi-input multi-output (MIMO) control problems subject to constraints. It finds the
control input by optimizing a cost function that incorporates a model of the system to
predict future outputs subject to the constraints. This makes MPC a candidate method for
AUV trajectory generation. However, traditional MPC has difficulties in computing control
inputs in real time for processes with fast dynamics.
This research applies a novel MPC approach, called Sampling-Based Model Predictive
Control (SBMPC), to generate kinematically or dynamically feasible system trajectories
for AUVs. The algorithm combines the benefits of sampling-based motion planning with
MPC while avoiding some of the major pitfalls facing both traditional sampling-based planning
algorithms and traditional MPC, namely large computation times and local minimum
problems. SBMPC is based on sampling (i.e., discretizing) the input space at each sample
period and implementing a goal-directed optimization method (e.g., A?) in place of standard
nonlinear programming. SBMPC can avoid local minimum, has only two parameters
to tune, and has small computational times that allows it to be used online fast systems.
A kinematic model, decoupled dynamic model and full dynamic model are incorporated
in SBMPC to generate a kinematic and dynamic feasible 3D path. Simulation results
demonstrate the efficacy of SBMPC in guiding an autonomous underwater vehicle from a
start position to a goal position in regions populated with various types of obstacles.
Filename Size Approximate Download Time (Hours:Minutes:Seconds)
28.8 Modem 56K Modem ISDN (64 Kb) ISDN (128 Kb) Higher-speed Access Caldwell_C_Dissertation_2011.pdf 1.83 Mb 00:08:28 00:04:21 00:03:48 00:01:54 00:00:09