Autonomous driving using Model Predictive Control methods
(2017)Department of Automatic Control
- Abstract
- This thesis explored the path following applications of autonomous driving, where the purpose is to sense the environment and navigate to a goal without human intervention. Autonomous driving enables safer journeys by removing human error in driving, as well as reducing fuel consumption and driving time by optimizing the engine and brake actuation. In 2017 it is a well researched topic in the academic and industrial world.
The scope of this work was to create a reliable and efficient path following solution which enabled a robot car to accurately traverse along any given path. The objective was therefore to minimize path error with the additional objective to minimize the traversal time.
The method used to realize the objective... (More) - This thesis explored the path following applications of autonomous driving, where the purpose is to sense the environment and navigate to a goal without human intervention. Autonomous driving enables safer journeys by removing human error in driving, as well as reducing fuel consumption and driving time by optimizing the engine and brake actuation. In 2017 it is a well researched topic in the academic and industrial world.
The scope of this work was to create a reliable and efficient path following solution which enabled a robot car to accurately traverse along any given path. The objective was therefore to minimize path error with the additional objective to minimize the traversal time.
The method used to realize the objective statement was Model Predictive Contouring Control (MPCC). MPCC uses optimization based prediction to smoothly control the car along a reference path, where actuator constraints and the objective are both handled in the optimization problem. The reference was also defined as a geometric function instead of a specified trajectory of coordinates. With the addition of objective weights it was possible to choose the relative importance of the two objectives, path accuracy and
minimizing time. The resulting problem formulation, which was heavily nonlinear, was linearized to reduce computation time and solved using FORCES Pro. Simulations where made in Simulink, while real-time execution of the MPCC were achieved with a robot prototype, Robotics Operating System (ROS) and Matlab.
The result suggested that MPCC was able to operate efficiently in a real time environment. Manipulating the objective weights resulted in predictable operation, which makes it possible for a user to control whether faster traversal or tracking performance is desired. It can be concluded that MPCC is a viable control method for autonomous driving, in particular for autonomous racing purposes. Additionally, the simulation results regarding the non-linear MPCC showed a surprisingly fast solve time, which suggest that it would be feasible for on-line driving using the solver FORCES Pro. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/8906188
- author
- Fredlund, Johan Kellerth and Sulejmanovic, Kenan Sadik
- supervisor
- organization
- year
- 2017
- type
- H3 - Professional qualifications (4 Years - )
- subject
- report number
- TFRT-6029
- ISSN
- 0280-5316
- language
- English
- id
- 8906188
- date added to LUP
- 2017-05-24 11:53:56
- date last changed
- 2017-05-24 11:53:56
@misc{8906188,
abstract = {{This thesis explored the path following applications of autonomous driving, where the purpose is to sense the environment and navigate to a goal without human intervention. Autonomous driving enables safer journeys by removing human error in driving, as well as reducing fuel consumption and driving time by optimizing the engine and brake actuation. In 2017 it is a well researched topic in the academic and industrial world.
The scope of this work was to create a reliable and efficient path following solution which enabled a robot car to accurately traverse along any given path. The objective was therefore to minimize path error with the additional objective to minimize the traversal time.
The method used to realize the objective statement was Model Predictive Contouring Control (MPCC). MPCC uses optimization based prediction to smoothly control the car along a reference path, where actuator constraints and the objective are both handled in the optimization problem. The reference was also defined as a geometric function instead of a specified trajectory of coordinates. With the addition of objective weights it was possible to choose the relative importance of the two objectives, path accuracy and
minimizing time. The resulting problem formulation, which was heavily nonlinear, was linearized to reduce computation time and solved using FORCES Pro. Simulations where made in Simulink, while real-time execution of the MPCC were achieved with a robot prototype, Robotics Operating System (ROS) and Matlab.
The result suggested that MPCC was able to operate efficiently in a real time environment. Manipulating the objective weights resulted in predictable operation, which makes it possible for a user to control whether faster traversal or tracking performance is desired. It can be concluded that MPCC is a viable control method for autonomous driving, in particular for autonomous racing purposes. Additionally, the simulation results regarding the non-linear MPCC showed a surprisingly fast solve time, which suggest that it would be feasible for on-line driving using the solver FORCES Pro.}},
author = {{Fredlund, Johan Kellerth and Sulejmanovic, Kenan Sadik}},
issn = {{0280-5316}},
language = {{eng}},
note = {{Student Paper}},
title = {{Autonomous driving using Model Predictive Control methods}},
year = {{2017}},
}