Email:
Curriculum vitae: [CV]
Twitter:@RCalandra Follow @RCalandra
Linkedin: linkedin.com/in/rcalandra
GitHub: github.com/robertocalandra
Google Scholar: Link
Youtube: Link
Slides: Link

Brief Bio

Roberto Calandra is a Research Scientist at Facebook AI Research. Previously, he was a Postdoctoral Scholar at the University of California, Berkeley (US) in the Berkeley Artificial Intelligence Research Laboratory (BAIR) working with Sergey Levine. His education includes a Ph.D. from TU Darmstadt (Germany) under the supervision of Jan Peters and Marc Deisenroth, a M.Sc. in Machine Learning and Data Mining from the Aalto university (Finland), and a B.Sc. in Computer Science from the Università degli studi di Palermo (Italy). His scientific interests are broadly at the conjunction of Decision-making, Robotics and Machine Learning. Research topics that he is currently developing include Model-based Reinforcement Learning, Tactile Sensing, Morphology Adaptation, and Bayesian Optimization.

I believe in the importance of a balanced life, and thus support Slow Science, Slow food and Sicilian music about slow life.

Research Topics

Upcoming Talks

December 2021: Invited talk at the NeurIPS 2021 Workshop on Robot Learning

14 Oct 2021: Invited talk at the Secondmind research seminar on Bayesian Optimization for Robotics
04 Jun 2021: Invited talk at ICRA 2021 ViTac Workshop
25 May 2021: Guest lecture at Caltech for CS159 on Theory and Practice of Model-based Reinforcement Learning
11 May 2021: Invited talk at Naver Labs
24 Mar 2021: Invited talk at the Keele University
13 Jan 2021: Invited talk at the Mediterranean Machine Learning Summer School [Slides]
21 Dec 2020: "Towards Embodied Intelligence" at the Technische Universität Dresden (Germany)
22 Oct 2020: "Towards Embodied Intelligence" at the University of Stuttgart (Germany)
31 Aug 2020: "Towards Embodied Intelligence" at the Max Planck Institute (Germany)
13 Jul 2020: "Towards a Science of Touch Processing" at the RSS Workshop on Visuotactile Sensors for Robust Manipulation [Slides]
13 Jul 2020: RSS Workshop on Self-Supervised Robot Learning
31 Mar 2020: "Towards In-hand Manipulation from Vision and Touch" at the ICRA Workshop on Closing the Perception-Action Loop with Vision and Tactile Sensing (ViTac 2020) [Slides] [Video]
27 May 2020: "Towards Embodied AI" at the Columbia University (USA)
29 Apr 2020: "Rethinking Model-based Reinforcement Learning" at the University of Edinburgh (UK)
23 Oct 2019: "Rethinking Model-based Reinforcement Learning" at the University of California Berkeley (USA)
08 Oct 2019: "Rethinking Model-based Reinforcement Learning" at the Arizona State University (USA)
27 Jul 2019: Bayesian Optimization for Robotics at the Joint Statistical Meetings (JSM) - Bayesian optimization session
23 Jun 2019: Learning Model-based Control for (Aerial) Manipulation at the RSS 2019 - Workshop on Aerial Interaction and Manipulation
20 Jun 2019: Robots and the Sense of Touch at the Re-Work Deep Reinforcement Learning Summit
31 May 2019: Stanford University
03 Jan 2019: Data, Learning and Inference (DALI)
21 Feb 2018: Stanford University
26 Jan 2018: TU Darmstadt
25 Jan 2018: University of Freiburg
24 Jan 2018: Max Planck Institute (Tuebingen)
23 Jan 2018: ETH
22 Jan 2018: EPFL
11 Jan 2018: Università di Palermo
08 Oct 2017: University of Southern California

Recent News

27 Jul 2021: Our paper Learning Accurate Long-term Dynamics for Model-based Reinforcement Learning has been accepted at the IEEE conference on Decision and Control (CDC) 2021.
21 Apr 2021: Today we released MBRL-Lib: A Modular Library for Model-based Reinforcement Learning [Code].
26 Feb 2021: Our paper On the Importance of Hyperparameter Optimization for Model-based Reinforcement Learning has been accepted at AISTATS 2021. For a brief overview, please read this [Blog post].
03 Feb 2021: Our paper on Model-Based Meta-Reinforcement Learning for Flight with Suspended Payloads has been published by IEEE Robotics and Automation Letters (RAL).
07 Jan 2021: Our paper on Learning Invariant Representations for Reinforcement Learning without Reconstruction has been accepted as an oral presentation to ICLR 2021.

16 Dec 2020: New pre-print about Learning Accurate Long-term Dynamics for Model-based Reinforcement Learning
15 Dec 2020: I am excited to share TACTO: A Fast, Flexible and Open-source Simulator for High-Resolution Vision-based Tactile Sensors [Code]
17 Nov 2020: I am happy to announce the launch of a new online space dedicated to the Touch Sensing community. More information at www.touch-sensing.org.
25 Sep 2020: Two papers accepted at NeurIPS 2020: Re-Examining Linear Embeddings for High-dimensional Bayesian Optimization and 3D Shape Reconstruction from Vision and Touch. Congratulations to all the authors!
29 May 2020: I am very excited to share the design of our new tactile sensor DIGIT: A Novel Design for a Low-Cost Compact High-Resolution Tactile Sensor with Application to In-Hand Manipulation. More information are available on the website https://digit.ml/.
04 Mar 2020: Two papers accepted at L4DC 2020: Plan2Vec: Unsupervised Representation Learning by Latent Plans and Objective Mismatch in Model-based Reinforcement Learning. Congratulations to all the authors!
03 Jun 2019: I am guest-teaching two lectures on reinforcement learning at Stanford University in the AA203: Optimal and Learning-based Control course.
16 May 2019: We just released a blog post about robotics at Facebook AI Research. [Fortune] [Wired] [The Verge] [TechCrunch]
07 May 2019: Our paper More Than a Feeling: Learning to Grasp and Regrasp Using Vision and Touch has been nominated a finalist for the best RA-L 2018 paper award.
09 Apr 2019: New pre-print on Learning to Control Highly Accelerated Ballistic Movements on Muscular Robots .
08 Mar 2019: New pre-print on Fast Neural Network Verification via Shadow Prices.
28 Jan 2019: Three papers accepted at ICRA: Data-efficient Learning of Morphology and Controller for a Microrobot with Thomas, Grant, Brian, Rene, Kris, and Sergey; Learning to Identify Object Instances by Touch: Tactile Recognition via Multimodal Matching with Justin and Sergey; Manipulation by Feel: Touch-Based Control with Deep Predictive Models with Stephen, Frederik, Dinesh, Mayur, Chelsea, and Sergey.
11 Jan 2019: New pre-print on Low Level Control of a Quadrotor with Deep Model-Based Reinforcement learning.
13 Dec 2018: I am co-organizing an ICLR workshop on Task-agnostic Reinforcement Learning together with Danijar Hafner, Amy Zhang, Ahmed Touati, Deepak Pathak, Frederik Ebert, Rowan McAllister, Marc G. Bellemare, Raia Hadsell, Alessandro Lazaric, Joelle Pineau. Deadline for submission is 29 March 2019.
04 Dec 2018: Our paper Deep Reinforcement Learning in a Handful of Trials using Probabilistic Dynamics Models has been selected for one of the NVIDIA Pioneer Awards at NeurIPS.
October 2018: I am happy to announce that I joined Facebook AI Research as a Research Scientist.
07 Sep 2018: Our Deep Reinforcement Learning in a Handful of Trials using Probabilistic Dynamics Models [Website] paper has been accepted at NIPS 2018 as a spotlight presentation (~4% acceptance rate). Congratulations to Kurtland for his first paper!
01 Aug 2018: Three journal published. Congratulations to all the authors! More Than a Feeling: Learning to Grasp and Regrasp using Vision and Touch with Andrew, Dinesh, Wenzhen, Justin, Jitendra, Ted, and Sergey; Control of Musculoskeletal Systems using Learned Dynamics Models with Dieter, Bernhard and Jan; Bayesian Multi-Objective Optimisation with Mixed Analytical and Black-Box Functions: Application to Tissue Engineering with Simon, Mohammad, Liesbet, Marc, and Ruth.
30 May 2018: New pre-print available on arxiv about model-based RL: Deep Reinforcement Learning in a Handful of Trials using Probabilistic Dynamics Models [Website]
28 May 2018: New pre-print available on arxiv about learning to grasp with tactile sensing: More Than a Feeling: Learning to Grasp and Regrasp using Vision and Touch [Website]
06 Apr 2018: I am co-organizing the RSS Workshop on Multi-Modal Perception and Control together with Filipe Veiga, Aude Billard and Jan Peters. The submission deadline will be in May 2018!
22 Feb 2018: I am co-organizing the FAIM Workshop on Prediction and Generative Modeling in Reinforcement Learning together with Matteo Pirotta, Sergey Levine, Martin Riedmiller and Alessandro Lazaric. The submission deadline is 01 June 2018!
26 Jan 2018: Learning Flexible and Reusable Locomotion Primitives for a Microrobot accepted at RAL+ICRA. Congratulations to Brian and Grant!
20 Jan 2018: I participated to the Dagstuhl seminar on Personalized Multiobjective Optimization.
03 Jan 2018: The talks from the RSS17 Workshop on Tactile Sensing for Manipulation: Hardware, Modeling, and Learning are now available online
30 Nov 2017: Three papers accepted to the NIPS workshop on Acting and Interacting in the Real World which will take place on Dec. 8th at NIPS: More Than a Feeling: Learning to Grasp and Regrasp using Vision and Touch, Learning Flexible and Reusable Locomotion Primitives for a Microrobot, and On the Importance of Uncertainty for Control with Deep Dynamics Models
20 Nov 2017: Invited talk today at Facebook on "Model-based Policy Search and Beyond"
16 Oct 2017: The Feeling of Success: Does Touch Sensing Help Predict Grasp Outcomes? is now on available on arxiv.
04 Oct 2017: Invited talk today at the University of Southern California on "Learning to Grasp from Vision and Touch".
Sep 2017: I am organizing the NIPS Workshop on Meta-learning (MetaLearn) together with Frank Hutter, Hugo Larochelle, and Sergey Levine. The submission deadline is 01 November 2017!
Sep 2017: Two papers accepted at CoRL 2017: The Feeling of Success: Does Touch Sensing Help Predict Grasp Outcomes?, and MBMF: Model-Based Priors for Model-Free Reinforcement Learning.
Aug 2017: I am editing the JMLR Special issue on Bayesian optimization together with Roman Garnett, Javier González, Frank Hutter, and Bobak Shahriari. The deadline for submissions is 31 March 2018!
Jul 2017: Our paper Goal-Driven Dynamics Learning via Bayesian Optimization has been accepted at CDC.
Apr 2017: Today I gave a talk about “Goal-Driven Dynamics Learning for Model-Based RL” at the DALI 2017 Workshop on Data-Efficient Reinforcement Learning. [Slides][Video]

Representative Publications

In this paper, we present...

For the full list of publications please refer to Google Scholar