Sergey Alatartsev: Robot Trajectory Optimization for Relaxed Effective Tasks. Otto-von-Guericke University of Magdeburg, 2015.

Abstract

Industrial robots are flexible machines that are currently involved in multiple production domains. Mainly their workflow consists of two alternating stages. The first stage is effective movements that are required to perform a task, e.g., welding a seam. The second stage is supporting movements that are needed to move from one effective task to another, e.g., movements between welding seams. Many effective tasks allow a certain freedom during their execution, e.g., the robot’s tool might have a certain deviation during welding. This freedom is often ignored and robots are programmed manually based on the programmer’s intuition. Nonetheless, this freedom can be used as an extra degree of freedom for robot trajectory optimization. In this thesis, we propose a formalization of this freedom for effective tasks. We refer to an effective task with a formalized freedom of execution as a relaxed effective task. Having an infinite number of ways to execute a task raises several research questions: (i) how to optimize a sequence of entry points for relaxed effective tasks? (ii) how to find starting robot configurations for these tasks? (iii) how to optimize a robot trajectory for a certain relaxed task? We propose a solution concept that decomposes a problem containing all three questions into three components that can be applied in combination with each other or with other state-of-the-art approaches. The first component considers the problem of finding a sequence of effective tasks and their entry points. This problem is modeled as the Traveling Salesman Problem with Neighborhoods (TSPN) where a tour has to be found through a set of areas. We propose a Constricting Insertion Heuristic for constructing a tour and a Constricting 3-Opt for improving the tour. In the second component, the problem of adapting a tour for a robot to execute and searching for starting robot configurations is modeled as a Touring-a-sequence-of-Polygons Problem (TPP) where a tour has to be found through a given sequence of areas. We propose a modification of the Rubber-Band Algorithm (RBA). We refer to this extension as a Nested RBA. Optimization of a robot trajectory in the third component is also represented as a TPP. However, in contrast to the classic RBA where areas are constricted with a polyline, we propose an extension of the RBA called Smoothed RBA where areas are constricted with a smooth curve which leads to a minimal cost robot trajectory.

BibTeX (Download)

@phdthesis{Alatartsev2015c,
title = {Robot Trajectory Optimization for Relaxed Effective Tasks},
author = {Sergey Alatartsev},
url = {https://cse.cs.ovgu.de/cse-wordpress/wp-content/uploads/2016/01/Alatartsev_PhD_thesis.pdf},
year  = {2015},
date = {2015-07-07},
school = {Otto-von-Guericke University of Magdeburg},
abstract = {Industrial robots are flexible machines that are currently involved in multiple production domains. Mainly their workflow consists of two alternating stages. The first stage is effective movements that are required to perform a task, e.g., welding a seam. The second stage is supporting movements that are needed to move from one effective task to another, e.g., movements between welding seams. Many effective tasks allow a certain freedom during their execution, e.g., the robot’s tool might have a certain deviation during welding. This freedom is often ignored and robots are programmed manually based on the programmer’s intuition. Nonetheless, this freedom can be used as an extra degree of freedom for robot trajectory optimization. In this thesis, we propose a formalization of this freedom for effective tasks. We refer to an effective task with a formalized freedom of execution as a relaxed effective task. Having an infinite number of ways to execute a task raises several research questions: (i) how to optimize a sequence of entry points for relaxed effective tasks? (ii) how to find starting robot configurations for these tasks? (iii) how to optimize a robot trajectory for a certain relaxed task? We propose a solution concept that decomposes a problem containing all three questions into three components that can be applied in combination with each other or with other state-of-the-art approaches. The first component considers the problem of finding a sequence of effective tasks and their entry points. This problem is modeled as the Traveling Salesman Problem with Neighborhoods (TSPN) where a tour has to be found through a set of areas. We propose a Constricting Insertion Heuristic for constructing a tour and a Constricting 3-Opt for improving the tour. In the second component, the problem of adapting a tour for a robot to execute and searching for starting robot configurations is modeled as a Touring-a-sequence-of-Polygons Problem (TPP) where a tour has to be found through a given sequence of areas. We propose a modification of the Rubber-Band Algorithm (RBA). We refer to this extension as a Nested RBA. Optimization of a robot trajectory in the third component is also represented as a TPP. However, in contrast to the classic RBA where areas are constricted with a polyline, we propose an extension of the RBA called Smoothed RBA where areas are constricted with a smooth curve which leads to a minimal cost robot trajectory. },
keywords = {robotics, travelling salesman problem},
pubstate = {published},
tppubtype = {phdthesis}
}