Presented by: Rafael O. Batista Jorge Wall Following Robot Path Planning Problem: A Statistical Learning Approach Universidad de Puerto Rico Recinto Universitario.

Slides:



Advertisements
Presentaciones similares
12.5 Completing the Square Goal: Solve a quadratic equation by completing the square.
Advertisements

2003 HYPACK MAX Training Seminar1 Sample Belgium Test Inputting Tag Line information into HYPACK® MAX In the old days, a Tag Line was anchored to.
METABOLOMICS PCA – Principal Component Analysis. Modelos y Datos  Todos los estudios arrojan datos  En cada estudio se miden variables.
Decisiones Robustas en la optimización de portafolio en el Merval Argentina Quants Meetup Xavier Ignacio González Facultad de Ingeniería. Universidad de.
TELEFÓNICA Research (I+D ) © 2008 Telefónica Investigación y Desarrollo, S.A. Unipersonal ICT 2008 – Collective Intelligence Networking Nov. 26, 2008 ©
Rafael Morales Bueno Grupo de la Universidad de Málaga Presentación y posibles contribuciones para la red EUREKA Rafael Morales Bueno.
Lesiones orales y estado inmunológico de pacientes VIH+ expuestos o no al consumo de alcohol. Blanca Lucía Acosta de Velásquez Elisa María Pinzón Gómez.
1 Can Quadratic Techniques Solve Polynomial Equations? PROBLEM 1 Standards PROBLEM 3 PROBLEM 2 PRESENTATION CREATED BY SIMON PEREZ. All rights reserved.
Connected children in Latin America. How close are they from enjoying their right to a safe and responsible use of ICT? Express yourself LATIN AMERICA:
Presentation Created by Dr. Luis O. Jiménez Design & Validation of Classifiers Machine Learning Design & Validation of Classifiers.
Bioinformática Predicción estructural y funcional Máster Biomedicina
Artificial Intelligence Machine learning Fall 2008 professor: Luigi Ceccaroni.
1 Applied biostatistics Francisco Javier Barón López Dpto. Medicina Preventiva Universidad de Málaga – España
Learning Target: I will be able to determine the Difference between different ecosystems around the world.
Selected ShakeAlert Performance Summaries for 1 Jan 2013 through 27 May 2013 CISN Testing Center ShakeAlert Project Philip Maechling, Maria Liukis, Thomas.
sistema nervioso
Presentation Title My name My position, contact information or project description.
UNIVERSITY OF CANTABRIA
INGLÉS BÁSICA PRIMARIA ASSESSMENT IN LANGUAGE TEACHING.
Presentation Title My name My position, contact information or project description.
Presentation Title My name My position, contact information or project description.
Selección de atributos
PRACTICES AND DIDACTIC APPLICATIONS
Engineering, Science and Math along with computing
A) b) c) d) Fig. S1. Representative LC-MRM chromatograms of serum phospholipids obtained using MRM transitions from head group loss a) Phosphatidylcholine.
Click to edit Master text styles Second level Third level Fourth level Fifth level Quantitative Analysis.
Una reflexión acerca del uso de distribuciones aproximadas
CARIBBEAN MARINE ATLAS PHASE 2 – CMA2
Efectos del manejo en la calidad de la madera
Weka & Rapid Miner Tutorial By Chibuike Muoh. WEKA:: Introduction A collection of open source ML algorithms – pre-processing – classifiers – clustering.
Neutron Radiography. Outline Introduction Radiography principle Instrumentation Conclusions.
New Data and How They Change What We Know About Heart Failure.
C. Jaramillo*, J. M. Jaramillo, J. E. Correa  Universidad EAFIT
BI – Business Intelligence
GRAPHIC MATERIALS 1. GRAPHIC MATERIALS. GRAPHIC MATERIALS 1. GRAPHIC MATERIALS.
Youden Analysis. Introduction to W. J. Youden Components of the Youden Graph Calculations Getting the “Circle” What to do with the results.
SAG – Soluciones Avanzadas en Gas SAG vision and mission – Peruvian market We facilitate the encounter between the market and the best technologies.
TASK-BASED LEARNING (TBL) Virtual Tutor: Ivonne Caro Laynes.
JUNIOR MOISES ARAUJO MARCACUZCO C. SEGÚN SU DIRECCION DE FLUJO Unsteady flow simulations of Pelton turbine at different rotational speeds Minsuk.
Presentation Title My name My position, contact information or project description.
University Polytechnic Estatal of Carchi English Investigation Integrants: Guerrero Jilmar Ortega Freddy Paredes Erika Quishpe Sandra Morocho Ronny Class:
Presentation Title My name My position, contact information or project description.
Rational Tool Overview. Introduction Requirements-Driven Software Development with Rational Analyst Studio. Tafadzwa Nzara Analysis & Design Consultant.
Actividades en Big Data
i=2I.
Inverse Modeling in Atmospheric Sciences
Predicción de fallos en redes bancarias mediante Machine Learning
Welcome to Room 202! Bienvenido al salón 202.
Juan Sebastian Alarcon Tomas Vargas Santiago Sanchez Camilo Mendez
Interactive Digital TV in Brazil: Technical and Social Aspects Mauro Oliveira Université d’Ottawa Ecole d’Ingénierie et de Technlogie de l’Information.
Cyclosporine Use in Epidermal Necrolysis Is Associated with an Important Mortality Reduction: Evidence from Three Different Approaches  Carlos González-Herrada,
A PowerPoint Template Your Presentation Name. This text is a placeholder Main Content Page Layout 2 Copyright 2009.
Introduction to CAN. What is CAN and what are some of its features? Serial communication Multi-Master Protocol Compact –Twisted Pair Bus line 1 Megabit.
Fundamentals of Web Development - 2 nd Ed.Randy Connolly and Ricardo Hoar Fundamentals of Web DevelopmentRandy Connolly and Ricardo Hoar © 2017 Pearson.
PST RESOURCE OVERVIEW NAME of the resource / tool
Organización Mundial de la Salud
Climate data in R with the raster package Jacob van Etten.
UNIVERSIDAD NACIONAL DE CAJAMARCA HISTORIA DE LA INVESTIGACION DE OPERACIONES INTRODUCCION Ing. MANUEL ROBERTO AZAHUANCHE OLIVA.
Quasimodo: Translate Sra. Crossett está embarazada.
Welcome to PowerPoint gdskcgdskfcbskjc. Designer helps you get your point across PowerPoint Designer suggests professional designs for your presentation,
How to write my report. Checklist – what I need to include Cover page Contents page – with sections Introduction - aims of project - background information.
Presentation Title My name My position, contact information or project description.
DSP BASED INTERSECTIONAL CONTROL IN MULTIDRIVE SYSTEMS M. Aníbal Valenzuela L., Jimmy Palma H., Ricardo Sanchez Sch. Dept. of Electrical Engg., University.
Types of evaluation. TYPES OF EVALUATION OBJECTIVE PURPOSE CLEAR CRITERIA INSTRUMENT OF MEASUREMENT BASED ON AN QUANTITATIVE POINT OF VIEW SUBJECTIVE.
TexPoint fonts used in EMF.
Sistemas Inteligentes Lógica Difusa y Sistemas Lineales
PRESENTATION: PRESENTATION: MY FUTURE By: Jonathan Julian Ortiz Arango code: 21.
International Civil Aviation Organization Spectrum Seminar Cairo, Egypt June 4-6, 2006 Global Navigation Satellite System (GNSS) Overview and Spectrum.
F. Barrio-Parra, M. Izquierdo-Díaz, D. Bolonio, Y. Sánchez-Palencia,
Churn Norris – Retain your customers with ML
Transcripción de la presentación:

Presented by: Rafael O. Batista Jorge Wall Following Robot Path Planning Problem: A Statistical Learning Approach Universidad de Puerto Rico Recinto Universitario de Mayagüez ININ6048 Final Project Presentation Mayagüez, 2016

Agenda  Problem overview  Dataset description  Methods  Preprocessing  Selected machine learning algorithms  Performance measures  Stacked learner  Results  Preprocessing  Selected machine learning algorithms  Performance measures  Stacked learner  PCA vs LDA  Conclusions

Problem overview  Automatic robot path planning decision  Multiple type of applications  Learning from environment (Sensor measurements)  Different types of approaches o Heuristic approaches (Fuzzy logic) o Geometric base approaches (Optimization based map search) o Machine learning algorithms (Supervised and unsupervised)  Open research problem o Real time execution o Dynamic environments o Unknown environments Image by Simeon87, distributed under a CC-BY 2.0 license.CC-BY 2.0 license

Problem overview  SCITOS G5 robot Photograph by MetraLabs.

Dataset description VariableDescriptionTypeValuesPredictorMissing Values UDS1 Position Angle = 180º Quantitative (0.400,0.401,0.402), (5.000,4.866, 4.860) YES0% UDS3Position Angle = -150º Quantitative (0.470,0.471,0.492), (5.029,5.028, 5.026) YES0% UDS6Position Angle = -105º Quantitative (1.114,1.115,1.118), (5.005,5.000, 4.980) YES0% UDS9Position Angle = -60º Quantitative (0.836,0.854,0.861), (5.000,4.956, 4.955) YES0% UDS12Position Angle = -15º Quantitative (0.778,0.779,0.780), (5.000,4.992, 4.981) YES0% UDS15Position Angle = 30º Quantitative (0.495,0.496,0.497), (5.000,4.921, 4.920) YES0% UDS18Position Angle = 75º Quantitative (0.354,0.355,0.356), (5.000,4.608, 4.591) YES0% UDS21Position Angle = 120º Quantitative (0.380,0.381,0.382), (5.000,4.822, 4.812) YES0% UDS24Position Angle = 165º Quantitative (0.377,0.380,0.381), (5.000,4.871, 4.865) YES0% Move ForwardPath Following Decision Qualitative 2205 (44.41%)NO0% Sharp Right TurnPath Following Decision Qualitative 2097 (38.43%)NO0% Slight Left TurnPath Following Decision Qualitative 328 (6.01%)NO0% Slight Right TurnPath Following DecisionQualitative826 (15.13%)NO0%  Twenty four sensors (Quantitative)  Four classes

Methods  Preprocessing  Multivariate normality test  Henze-Zirkler’s Multivariate Normality Test  Mardia’s Mutivariate Normality Test  Chi square Q-Q plot  Outlier detection  Mahalanobis distance (Assumes MVN dataset)  Random forest instance proximity (No normality assumption)  Collinearity and dimensionality reduction  Principal components analysis o Correlation matrix (scaling) o Explained variance (number of principal components) o Unsupervised  Linear discriminant analysis o Assumes normality o Class separation o Supervised

Methods  Selected machine learning algorithms ?  Random forest  Non parametric and supervised  Ensemble (Bagging)  k-NN  Non parametric  Supervised

Methods  Selected machine learning algorithms  Support Vector Machine  Cost  Gamma  Kernel Trick  Non linear  Different types of kernel: o Linear o Polinomial o Radial

 Selected machine learning algorithms Methods  Support Vector Machine Cost = 100Cost = 0.1

 Performance measures  Classification error  Kappa statistics  Multiclass Logarithmic Loss Methods  Desirability Function

 Selected machine learning algorithms Methods  Stacked learner Prob. KNN Neighbors = 9 Prob. RF N.Trees = 100 Prob. SVM Degree = 3 Dataset Class Labels PCA Rotated Dataset CV RF Result

Results  Preprocessing: Multivariate normality test.

Results  Preprocessing: Outlier detection

Results  Preprocessing: PCA Scree plot and explained variance PC1PC2PC3PC4PC5PC6PC7PC8PC9PC10PC11PC12PC13PC14PC15 Standard deviation Proportion of Variance Cumulative Proportion

Results  Preprocessing: LDA Histogram First linear discriminant loading’s histogram.

Results  Preprocessing: LDA Histogram Second linear discriminant loading’s histogram.

Results  Preprocessing: LDA Histogram Third linear discriminant loading’s histogram.

Results  Tuning: RF LDA loadings dataset tuning

Results  Tuning: RF PCA rotated dataset tuning

Results  Tuning: SVM with LDA loadings dataset (radial kernel)

Results  Tuning: SVM with PCA rotated dataset (polynomial kernel)

Results  Tuning: SVM with PCA rotated dataset (polynomial kernel)

Results  Tuning: Summary table for SVM DatasetKernelDegreeCostGamma Resulting Error PCA rotated RadialNA PCA rotated PolynomialThird LDA loadings RadialNA

Results  Implementation: PCA rotated dataset and kNN

Results  Implementation: PCA rotated dataset and RF

Results  Implementation: PCA rotated dataset and SVM

Results  Implementation: Dataset for stacked approach

Results  Implementation: PCA rotated dataset stacked learner

Results  Implementation: LDA loadings dataset and kNN

Results  Implementation: LDA loadings dataset and RF

Results  Implementation: LDA loadings dataset and SVM

Results  Implementation: LDA loadings dataset and stacked learner PCA

PCA vs LDA  Comparing kNN performance PCA LDA

PCA vs LDA  Comparing RF performance PCA LDA

Conclusiones  The best performance of our proposal was obtained by using RF and the LDA loadings dataset, with a desirability value of 94.75%.  The stacking approach was successfull, and improved the desirability value to 99.99% for both cases, LDA and PCA.  Machine learning techniques are useful for this kind of path planning problems with dynamical or unknow enviroments.  Future work may considered the use of better optimization strategies for the selection of optimal parameter for SVM.

References Fletcher, T. (2009). Support Vector Machines Explained. Retrieved from Freire, A. Veloso, M. Barreto, G. (2010). UCI Machine Learning Repository [ Irvince, CA: University of California, School of Information and Computer Science. Freire, A. L., Barreto, G. A., Veloso, M., & Varela, A. T. (2009). Short-term memory mechanisms in neural network learning of robot navigation tasks: A case study th Latin American Robotics Symposium, LARS 2009, (4). Hastie, T. J., Tibshirani, R. J., & Friedman, J. H. (2009). The elements of statistical learning: data mining, inference, and prediction. book, New York: Springer. MetraLabs GmbH. (2011). SCITOS G5 Embedded PC and Operating System. Retrieved December 14, 2016, from Otte, M. W. (2015). A Survey of Machine Learning Approaches to Robotic Path-Planning. Cs.Colorado.Edu. Retrieved from Modeling Prelim/Otte.pdf