Localization of Mobile Robots for Precision Manufacturing

Date
2021
Authors
Paijens, Antonius Franciscus Maria
Supervisor
Huang, Loulin
Al-Jumaily, Ahmed
Item type
Thesis
Degree name
Doctor of Philosophy
Journal Title
Journal ISSN
Volume Title
Publisher
Auckland University of Technology
Abstract

To manufacture a bigger part, you need a larger machine!"

This paradigm is going to be broken through the development of mobile robots that can move around a part and apply a manufacturing operation to it, instead of having to envelop it. Monument size machines will be replaced by a mobile appliance that can do great things. In almost any perspective, mobile appliances outperform big stationary machines sitting in a workshop. The size of the parts that can be produced with a mobile appliance is not constrained by the machine size. It is more flexible in its deployment. A mobile appliance does not take up workshop space permanently but can be used "on site" where the materials are and the part manufactured is needed. It can be taken to a workshop for maintenance and repair. As a consequence the cost of ownership of a mobile appliance is lower than the cost of ownership of a stationary machine. It will make processing of large parts accessible to a wider audience of makers and contribute to the democratization of manufacturing.

To enable a mobile robot to perform precision manufacturing operations, it has to be able to position itself accurately, within tolerances that match the allowable dimensional tolerances of the part to be manufactured by a certain margin. And in order to position itself accurately it has to measure its own location and orientation even more accurately. This challenge is made tangible by focusing on laser cutting as the manufacturing operation in the present research. Within the scope of laser cutting operations a benchmark is defined based on criteria for adoption of a mobile platform for precision manufacturing, derived from manufacturing practices in industries that apply laser cutting in their production. The benchmark states that the allowable manufacturing tolerance of a part to be cut is 0.1% of its maximum dimension. The mobile platform carrying the laser cutter will have to localize itself with a better accuracy by some margin to be able to cut a part within this tolerance. This benchmark defines a measurable target for allowable manufacturing tolerances of parts produced with laser cutting, against which the accuracy of the localization of a mobile robot for precision manufacturing can be evaluated.

The present research project aims to develop a system that can localize a mobile platform within the limits set by the benchmark. Optical flow sensors are selected for incremental odometry of the mobile platform and a laser lighthouse system to measure its absolute location and orientation. Optical flow sensors used to measure the movement of computer mice are very accurate in measuring displacement, but accumulate measurement errors in the course of time. An important source of systematic errors when applying multiple optical flow sensors is the absence of synchronization of the data acquisition by the sensors.

The measurement of a location and orientation in an absolute reference system with a laser lighthouse compensates for the accumulation of random errors generated by the optical flow sensors. The laser lighthouse is a single landmark beacon, emitting rotating light sheets generated with a laser, similar to a lighthouse. A mobile platform to be localized carries light sensors in known locations, that record timestamps when the light sheet from the lighthouse passes over a sensor. The timestamps from at least three sensors on the mobile platform allow for calculation of a planar location and orientation of the platform relative to the lighthouse when it sweeps a light sheet in one orientation. With light sheets swept in two orientations by the lighthouse, a spatial location and orientation can be established for a mobile platform with at least three light sensors.

In the computations for the incremental odometry, the sensitivity of the optical flow sensors and their position and orientation on the mobile platform are cardinal parameters. For the determination of an absolute location and orientation with a laser lighthouse, the location of the light sensors on the mobile platform is the key parameter. Inaccuracies in these parameters will generate a systematic error in the calculated position and orientation of the mobile platform and prevent it from meeting the criteria for adoption of the platform for precision manufacturing.

The present research focuses on experimental methods to establish accurate values for the sensitivity and the location and orientation of the optical flow sensors as well as the location of the light sensors on the mobile platform. Direct measurement of these parameters is difficult and produces uncertain results as for instance the location of the "focal point" of a light sensors with a surface area of several square millimeters can only be established by measurement with the sensor itself. This is exactly what the designed experiments intend to do: measure the magnitude of the cardinal parameters for the localization of the mobile platform with the sensors itself to eliminate systematic errors.

The experiments designed force the mobile platform to travel a constrained motion around a fixed ”anchor point”. The constrain imposed establishes a relationship between the movement the platform and the path traveled by the sensors that allows determination of the location and orientation of the sensors from their recordings while traveling the path in an experiment. Comparison of a reconstruction of the path using the sensor recordings and their measured position in the frame with the constrained motion confirms the efficacy of this approach:

• The computed location of the mobile platform deviated 0.16% of the total distance traveled by the platform (6.7 meter). This is a significant improvement over the next best result of an accuracy of 0.28% published in the open literature. • After measuring the sensor locations on the experimental mobile platform with the experiments designed, the relative error of the location of the mobile platform established with one lighthouse base station, dropped from over 0.3% to below 0.2% (meter error per meter distance) for a slow moving platform.

The experiments demonstrate that when the mobile platform is stationary, the precision of localization of a mobile platform with the HTC-Vive laser lighthouse base station is an order of magnitude better than the benchmark of 0.1%. When the platform is moving however, the accuracy of the laser lighthouse method is insufficient to locate the mobile platform within the required tolerances set by the benchmark.

At the same time the present research has identified many leads for further reduction of systematic errors in the localization of a mobile platform to make precision manufacturing a feasible option.

As the result of the exploration performed to write this thesis, the following original research contribution were made: • Analysis of the propagation of the error in the sensor sensitivity of optical flow sensors into the location of a mobile platform derived from the sensor data. • Design of a data acquisition system that executes the data collection with the optical flow sensors in a schedule of strict coincidence. • Design of a simple experiment to measure sensitivities of optical flow sensors. • Design of an experiment to measure and validate location and orientation of the optical flow sensors on the mobile platform using readings from the sensors itself. • Analysis of the propagation of errors in the location of the light sensors into the location of a mobile platform derived from the sensor data. • Design of an experiment to measure and validate the location of light sensors on the mobile platform through the application of camera calibration techniques, using readings from the sensors itself.

Description
Keywords
Mobile robot , Navigation , Localization , Lighthouse , Light sensor , Optical flow , Time measurement , Real time , Data acquisition , Sensor position , Perspective projection , Conics , Camera calibration , Computer vision
Source
DOI
Publisher's version
Rights statement
Collections