Incremental Learning for Motion Prediction of Pedestrians and Vehicles

Alejandro Dizan Vasquez Govea

ISBN 10: 3642136419 / ISBN 13: 9783642136412

Editore: Springer Berlin Heidelberg Jun 2010, 2010

Riguardo questo articolo

This item is printed on demand - it takes 3-4 days longer - Neuware -Modeling and predicting human and vehicle motion is an active research domain.Owing to the difficulty in modeling the various factors that determine motion(e.g. internal state, perception) this is often tackled by applying machinelearning techniques to build a statistical model, using as input a collectionof trajectories gathered through a sensor (e.g. camera, laser scanner), and thenusing that model to predict further motion. Unfortunately, most currenttechniques use offline learning algorithms, meaning that they are not able tolearn new motion patterns once the learning stage has finished.This books presents a lifelong learning approach where motion patterns can belearned incrementally, and in parallel with prediction. The approach is based ona novel extension to hidden Markov models, and the main contribution presentedin this book, called growing hidden Markov models, which gives us the ability tolearn incrementally both the parameters and the structure of the model. Theproposed approach has been extensively validated with synthetic and realtrajectory data. In our experiments our approach consistently learned motionmodels that were more compact and accurate than those produced by two otherstate-of-the-art techniques, confirming the viability of lifelong learningapproaches to build human behavior models. 176 pp. Englisch.

Codice articolo 9783642136412

Dati bibliografici

• Titolo: Incremental Learning for Motion Prediction of Pedestrians and Vehicles
• Editore: Springer Berlin Heidelberg Jun 2010
• Data di pubblicazione: 2010
• Lingua: Inglese
• ISBN 10: 3642136419
• ISBN 13: 9783642136412
• Rilegatura: Buch
• Condizione: Neu
• Dimensioni: 241 L x 160 A x 16 P (millimetri) 241 millimetri di larghezza per 160 millimetri di altezza per 16 millimetri di profondità
• Peso: 468 grammi

Descrizione articolo

Riassunto

The Springer Tracts in Advanced Robotics (STAR) publishes new developments and advances in the fields of robotics research - rapidly and informally but with a high quality. The intent is to cover all the technical topics, applications, and multidisciplinary aspects of robotics, embedded in the fields of Mechanical Engineering, Computer Science, Electrical Engineering, Mechatronics, Control, and Life Sciences - as well as the methodologies behind them. Within the scope of the series are monographs, lecture notes, selected contributions from specialized conferences and workshops, as well as selected PhD theses.

Modeling and predicting human and vehicle motion is an active research domain. Owing to the difficulty in modeling the various factors that determine motion (e.g. internal state, perception) this is often tackled by applying machine learning techniques to build a statistical model, using as input a collection of trajectories gathered through a sensor (e.g. camera, laser scanner), and then using that model to predict further motion. Unfortunately, most current techniques use offline learning algorithms, meaning that they are not able to learn new motion patterns once the learning stage has finished.

This book presents a lifelong learning approach where motion patterns can be learned incrementally, and in parallel with prediction. The approach is based on a novel technique - growing hidden Markov models - which extends classical hidden Markov models by giving them the ability to learn incrementally both the parameters and the structure of the model. The proposed approach has been extensively validated with synthetic and real trajectory data. In our experiments our approach consistently learned motion models that were more compact and accurate than those produced by two other state-of-the-art techniques, confirming the viability of lifelong learning approaches to build human behavior models.

Dalla quarta di copertina

Modeling and predicting human and vehicle motion is an active research domain.Owing to the difficulty in modeling the various factors that determine motion(e.g. internal state, perception) this is often tackled by applying machinelearning techniques to build a statistical model, using as input a collectionof trajectories gathered through a sensor (e.g. camera, laser scanner), and thenusing that model to predict further motion. Unfortunately, most currenttechniques use offline learning algorithms, meaning that they are not able tolearn new motion patterns once the learning stage has finished.

This books presents a lifelong learning approach where motion patterns can belearned incrementally, and in parallel with prediction. The approach is based ona novel extension to hidden Markov models, and the main contribution presentedin this book, called growing hidden Markov models, which gives us the ability tolearn incrementally both the parameters and the structure of the model. Theproposed approach has been extensively validated with synthetic and realtrajectory data. In our experiments our approach consistently learned motionmodels that were more compact and accurate than those produced by two otherstate-of-the-art techniques, confirming the viability of lifelong learningapproaches to build human behavior models.