A Bayesian Approach to Nonlinear Parameter Identification for Rigid Body Dynamics

Ting, Jo-Anne; Mistry, Michael; Peters, Jan; Schaal, Stefan; Nakanishi, Jun

doi:10.15607/rss.2006.ii.032

Cited by 56 publications

(57 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In robotics it has been used to model contactless motion, e.g. predicting the motion of an object or robot manipulator in free space (Ting et al 2006;Boots et al 2014;Dearden and Demiris 2005). It has also been used to learn the dynamics of an object with a single, constant contact (such as pole balancing) (Schaal 1997;Atkeson and Schaal 1997).…”

Section: Related Workmentioning

confidence: 99%

Learning modular and transferable forward models of the motions of push manipulated objects

et al. 2016

View full text Add to dashboard Cite

The ability to predict how objects behave during manipulation is an important problem. Models informed by mechanics are powerful, but are hard to tune. An alternative is to learn a model of the object's motion from data, to learn to predict. We study this for push manipulation. The paper starts by formulating a quasi-static prediction problem. We then pose the problem of learning to predict in two different frameworks: (i) regression and (ii) density estimation. Our architecture is modular: many simple, object specific, and context specific predictors are learned. We show empirically that such predictors outperform a rigid body dynamics engine tuned on the same data. We then extend the density estimation approach using a product of experts. This allows transfer of learned motion models to objects of novel shape, and to novel actions. With the right representation and learning method, these transferred models can match the prediction performance of a rigid body dynamics engine for novel objects or actions.

show abstract

Section: Related Workmentioning

confidence: 99%

Learning modular and transferable forward models of the motions of push manipulated objects

et al. 2016

View full text Add to dashboard Cite

show abstract

“…It it hard to create sufficiently rich data in order to obtain plausible dynamics parameters. Furthermore, the parameters that optimally fit a data set, are often not physically consistent and, hence, physical consistency constraints have to be imposed on the identification problem [17]. Using data-based nonparametric models for RL -as employed in this paper -for learning optimal control policies can help to overcome these limitations.…”

Section: The Throttle Valve Systemmentioning

confidence: 99%

Learning Throttle Valve Control Using Policy Search

Bischoff

Nguyen-Tuong

Koller

et al. 2013

Advanced Information Systems Engineering

View full text Add to dashboard Cite

Abstract. The throttle valve is a technical device used for regulating a fluid or a gas flow. Throttle valve control is a challenging task, due to its complex dynamics and demanding constraints for the controller. Using state-of-the-art throttle valve control, such as model-free PID controllers, time-consuming and manual adjusting of the controller is necessary. In this paper, we investigate how reinforcement learning (RL) can help to alleviate the effort of manual controller design by automatically learning a control policy from experiences. In order to obtain a valid control policy for the throttle valve, several constraints need to be addressed, such as no-overshoot. Furthermore, the learned controller must be able to follow given desired trajectories, while moving the valve from any start to any goal position and, thus, multi-targets policy learning needs to be considered for RL. In this study, we employ a policy search RL approach, Pilco [2], to learn a throttle valve control policy. We adapt the Pilco algorithm, while taking into account the practical requirements and constraints for the controller. For evaluation, we employ the resulting algorithm to solve several control tasks in simulation, as well as on a physical throttle valve system. The results show that policy search RL is able to learn a consistent control policy for complex, real-world systems.

show abstract

“…Such methods can also be used to identify the dynamic parameters such as the center of mass, the moments of inertia, etc. Ting et al (2006), for example, presented a Bayesian approach for estimating these parameters on two different manipulation robots. In principle, these methods could be applied after our approach has bootstrapped the kinematic model, in order to refine or augment the model and achieve a faster convergence.…”

Section: Related Workmentioning

confidence: 99%

Approaches to Probabilistic Model Learning for Mobile Manipulation Robots

Sturm¹

2013

Springer Tracts in Advanced Robotics

View full text Add to dashboard Cite

Mobile manipulation robots are envisioned to provide many useful services both in domestic environments as well as in the industrial context. Examples include domestic service robots, that implement large parts of the housework, and versatile industrial assistants, that provide automation, transportation, inspection, and monitoring services. The challenge in these applications is that the robots have to function under changing, real-world conditions, be able to deal with considerable amounts of noise and uncertainty, and operate without the supervision of an expert. To meet these challenges, current robotic systems are typically custom-tailored to specific applications in welldefined environments, and therefore cannot deal robustly with changes in the situation. This thesis presents novel learning techniques that enable mobile manipulation robots, i.e., mobile platforms with one or more robotic manipulators, to autonomously adapt to new or changing situations. The developed approaches in this thesis cover the following four topics: (1) learning the robot's kinematic structure and properties using actuation and visual feedback, (2) learning about articulated objects in the environment in which the robot is operating, (3) using tactile feedback to augment the visual perception, and (4) learning novel manipulation tasks from human demonstrations.In the first part of this thesis, we present innovative approaches to learning a robot's own body schema from scratch using visual self-observation. This allows manipulation robots to calibrate themselves automatically and to adapt their body schemata autonomously, for example after hardware failures or during tool use. In the second part, we extend the developed framework to learning about articulated objects -such as doors and drawers -with which service robots often need to interact. The presented algorithms enable robots to learn accurate kinematic models of articulated objects, which in turn allow them to interact with the objects robustly. In the third part, we provide approaches that allow manipulation robots to make use of tactile perception -an ability that is known to play an important role in human object manipulation skills. The main contributions in this part are approaches to identifying objects and to perceiving aspects of their internal states. With this, a manipulation robot can verify that it has grasped the correct object and, for example, discriminate full from empty bottles. Finally, we present an integrated system that allows human operators to intuitively teach a robot novel manipulation tasks by demonstration.All techniques developed in the thesis are based on probabilistic learning and inference. They have been implemented and evaluated on real robots as well as in simulation. Extensive experiments have been conducted to analyze and validate the properties of the developed algorithms and to demonstrate a significant increase in robustness, adaptability, and utility of mobile manipulation robots in everyday life. Zusammenfassung

show abstract

A Bayesian Approach to Nonlinear Parameter Identification for Rigid Body Dynamics

Cited by 56 publications

References 23 publications

Learning modular and transferable forward models of the motions of push manipulated objects

Learning modular and transferable forward models of the motions of push manipulated objects

Learning Throttle Valve Control Using Policy Search

Approaches to Probabilistic Model Learning for Mobile Manipulation Robots

Contact Info

Product

Resources

About