Abstract
This thesis describes the strategy of a new approach for automated tilesetting currently under development by nLink AS in Sogndal. The main focus of the thesis is the development of an appropriate strategy for path planning for this robotic system.
A mathematical model of a 4-wheel skid-steering mobile platform with a model-based nonlinear controller is presented. A brief introduction to the basics of path planning is offered, before detailing some relevant algorithms. Three different optimization criteria, based on distance time and energy consumption, are defined and a MATLAB implementation of a Sampling-Based Model Predictive Optimization algorithm is presented. The mobile platform is implemented in SIMULINK with input from the SBMPO-algorithm. Simulation results of three different obstacle cases with both constant and variable speed are presented and the performance of the different optimization schemes are reviewed and compared.
The simulations clearly reveal that while optimizing with respect to energy consumption does present some very promising results, the current implementation is far to slow to compute paths with variable speeds. The most promising optimization method is an optimization with respect to time, but with a restriction on the turning radius to avoid sharp, energy-costly turns. This too has some drawbacks, but the ideal optimization strategy can be concluded to be a combination of time and energy efficiency.
A mathematical model of a 4-wheel skid-steering mobile platform with a model-based nonlinear controller is presented. A brief introduction to the basics of path planning is offered, before detailing some relevant algorithms. Three different optimization criteria, based on distance time and energy consumption, are defined and a MATLAB implementation of a Sampling-Based Model Predictive Optimization algorithm is presented. The mobile platform is implemented in SIMULINK with input from the SBMPO-algorithm. Simulation results of three different obstacle cases with both constant and variable speed are presented and the performance of the different optimization schemes are reviewed and compared.
The simulations clearly reveal that while optimizing with respect to energy consumption does present some very promising results, the current implementation is far to slow to compute paths with variable speeds. The most promising optimization method is an optimization with respect to time, but with a restriction on the turning radius to avoid sharp, energy-costly turns. This too has some drawbacks, but the ideal optimization strategy can be concluded to be a combination of time and energy efficiency.
