Sample-based informationl-theoretic stochastic optimal control

By Rudolf Lioutikov, Alexandros Paraschos, Jan Peters, Gerhard Neumann. 2014

32 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computational Theory and Mathematics, Reinforcement Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Sergey Levine
2. Pieter Abbeel
3. Chelsea Finn
4. Trevor Darrell
5. Shixiang Shane Gu
6. Yoshua Bengio
7. Ruslan Salakhutdinov
8. Abhinav Kumar Gupta

In Computational Theory and Mathematics

1. Sergey Levine
2. Pieter Abbeel
3. Chelsea Finn
4. Shixiang Shane Gu
5. Gerhard Neumann
6. Timothy P. Lillicrap
7. Herke Van Hoof
8. Trevor Darrell

In Reinforcement Learning

1. Sergey Levine
2. Pieter Abbeel
3. Chelsea Finn
4. Shixiang Shane Gu
5. Abhishek Gupta
6. Vikash Kumar
7. Ofir Nachum
8. Trevor Darrell

Citations in the corpus

• End-to-End Training of Deep Visuomotor Policies (2015, methodology): “t optimize neural network policies, and we could not obtain reasonable results with REPS. Prior applications of REPS generally focus on simpler, lower-dimensional policy classes (Peters et al., 2010; Lioutikov et al., 2014). 19 Levine, Finn, Darrell, and Abbeel (a)(b) (c)(d)(e)(f)(g)(h)(i) Figure 6:”
• Deep reinforcement learning for robotic manipulation with asynchronous off-policy updates (2016, background): “Model-based algorithms have explored a variety of dynamics estimation schemes, including Gaussian processes [20], mixture models [21], and local linear system estimation [22], with a more detailed overview in a recent survey [23].”
• Algorithmic Framework for Model-based Deep Reinforcement Learning with Theoretical Guarantees (2018, neutral): “…dynamical models ranging from Gaussian process (Deisenroth & Rasmussen, 2011b; Ko & Fox, 2009), time-varying linear models (Levine & Koltun, 2013; Lioutikov et al., 2014; Levine & Abbeel, 2014; Yip & Camarillo, 2014), mixture of Gaussians (Khansari-Zadeh & Billard, 2011), to neural networks…”
• Learning contact-rich manipulation skills with guided policy search (2015, methodology): “While prior policy search methods have been proposed that leverage locally linear models (see, e.g., [25]), the LQR-based update discussed in this section has been previously shown to achieve substantially better sample efficiency on challenging, contact-rich problems [1].”
• Learning of Non-Parametric Control Policies with High-Dimensional State Features (2015, methodology): “The transition dynamics have been approximated by deterministic single-sample outcomes [Daniel et al., 2013, Peters et al., 2010] which only works well for deterministic environments, or by timedependent linear models [Lioutikov et al., 2014].”