- 180
- 102 492
Autonomy Talks
Приєднався 16 чер 2019
This is the official channel of Autonomy Talks.
Autonomy Talks are open to everyone and no registration is required. The goal of the talks is the exchange among different groups working on innovations in autonomy. Autonomy Talks will take place virtually, on Zoom. Check out the official webpage for the details: zardini.mit.edu/autonomytalks
Autonomy Talks are open to everyone and no registration is required. The goal of the talks is the exchange among different groups working on innovations in autonomy. Autonomy Talks will take place virtually, on Zoom. Check out the official webpage for the details: zardini.mit.edu/autonomytalks
Autonomy Talks - Devansh Jalota: Algorithm and Incentive Design for Sustainable Resource Allocation
Autonomy Talks - 03/12/24
Speaker: Devansh Jalota, Stanford University
Title: Algorithm and Incentive Design for Sustainable Resource Allocation: Beyond Classical Fisher Markets
Abstract: Technological advances have opened new avenues for designing market mechanisms for resource allocation, from enhancing resource allocation eDiciency with widespread data availability to enabling real-time algorithm implementation. While these technological advancements hold significant promise, they also introduce new societal challenges pertaining to equity, privacy, data uncertainty, and security that existing market mechanisms often fail to address. My research develops data-driven and online learning algorithms and incentive schemes to address these challenges of traditional market mechanisms, thereby advancing the science and practice of market design for sustainable and society-aware resource allocation.
In this talk, I focus on addressing data uncertainty and privacy issues in the context of Fisher markets, a classical framework for fair resource allocation where the problem of computing equilibrium prices relies on complete information of user attributes, which are typically unavailable in practice. Motivated by this practical limitation, we study a modified online incomplete information variant of Fisher markets, where users with privately known utility and budget parameters, drawn i.i.d. from a distribution, arrive sequentially. In this novel market, we establish the limitations of static pricing and design dynamic posted-price algorithms with improved guarantees. Our main result is a posted-price algorithm that solely relies on revealed preference (RP) feedback, i.e., observations of user consumption, achieving the best-known guarantees for first-order algorithms in the RP setting while providing a regret analysis of a fairness-promoting logarithmic objective, unlike typical non- negative and bounded eDiciency-promoting objectives in online learning.
Speaker: Devansh Jalota, Stanford University
Title: Algorithm and Incentive Design for Sustainable Resource Allocation: Beyond Classical Fisher Markets
Abstract: Technological advances have opened new avenues for designing market mechanisms for resource allocation, from enhancing resource allocation eDiciency with widespread data availability to enabling real-time algorithm implementation. While these technological advancements hold significant promise, they also introduce new societal challenges pertaining to equity, privacy, data uncertainty, and security that existing market mechanisms often fail to address. My research develops data-driven and online learning algorithms and incentive schemes to address these challenges of traditional market mechanisms, thereby advancing the science and practice of market design for sustainable and society-aware resource allocation.
In this talk, I focus on addressing data uncertainty and privacy issues in the context of Fisher markets, a classical framework for fair resource allocation where the problem of computing equilibrium prices relies on complete information of user attributes, which are typically unavailable in practice. Motivated by this practical limitation, we study a modified online incomplete information variant of Fisher markets, where users with privately known utility and budget parameters, drawn i.i.d. from a distribution, arrive sequentially. In this novel market, we establish the limitations of static pricing and design dynamic posted-price algorithms with improved guarantees. Our main result is a posted-price algorithm that solely relies on revealed preference (RP) feedback, i.e., observations of user consumption, achieving the best-known guarantees for first-order algorithms in the RP setting while providing a regret analysis of a fairness-promoting logarithmic objective, unlike typical non- negative and bounded eDiciency-promoting objectives in online learning.
Переглядів: 132
Відео
Autonomy Talks - Cecilia Pasquale: A Multi-Scale Control Framework for Automated Buses
Переглядів 186Місяць тому
Autonomy Talks - 11/26/24 Speaker: Prof. Cecilia Pasquale, University of Genova Title: A Multi-Scale Control Framework for Optimal Charging and Speed Control of Electric Automated Buses
Autonomy Talks - Federico Rossi & Jean-Pierre de la Croix: Multi-agent autonomy on the Moon
Переглядів 416Місяць тому
Autonomy Talks - 19/11/24 Speakers: Dr. Federico Rossi & Dr. Jean-Pierre de la Croix, NASA JPL Title: Multi-agent autonomy on the Moon: NASA' Coordinated Autonomous Distributed Robotics Explorers (CADRE) mission Abstract: This talk will present the multi-agent autonomy architecture of NASA's Cooperative Autonomous Distributed Robotic Explorers (CADRE) mission, a technology demonstration that wi...
Autonomy Talks - Yue Wang: Bridge Spatial Intelligence & Embodied Intelligence w. Foundation Models
Переглядів 224Місяць тому
Autonomy Talks - 12/11/24 Speaker: Prof. Yue Wang, USC Title: Bridge Spatial Intelligence and Embodied Intelligence with Foundation Models
Autonomy Talks - Nadia Figueroa: From Motion to Interaction
Переглядів 374Місяць тому
Autonomy Talks - 05/11/24 Speaker: Prof. Nadia Figueroa, University of Pennsylvania Title: From Motion to Interaction: A Dynamical Systems Approach for Human-Centric Robot Learning and Control Abstract: For the last decades we have lived with the promise of one day being able to own a robot that can coexist, collaborate and cooperate with humans in our everyday lives. This has motivated a vast ...
Autonomy Talks - Ritu Raman: Leveraging Biological Actuators for Soft Robotics
Переглядів 218Місяць тому
Autonomy Talks - 29/10/24 Speaker: Prof. Ritu Raman, MIT Title: Leveraging Biological Actuators for Soft Robotics Abstract: Human beings and other biological creatures navigate unpredictable and dynamic environments by combining compliant mechanical actuators (skeletal muscle) with neural control and sensory feedback. Abiotic actuators, by contrast, have yet to match their biological counterpar...
Autonomy Talks - James Anderson: Collaborative Learning for Control of Heterogeneous Systems
Переглядів 3662 місяці тому
Autonomy Talks - 22/10/24 Speaker: Prof. James Anderson, Columbia University Title: Collaborative Learning for Control of Heterogeneous Systems Abstract: In this talk we consider two problems related to learning a linear system subject to a quadratic cost. We first study a model-free federated linear quadratic regulator (LQR) problem where M agents with unknown, distinct yet similar dynamics co...
Autonomy Talks - Sehoon Ha: Guided reinforcement learning to learn interactive motor skills
Переглядів 4272 місяці тому
Autonomy Talks - 08/10/24 Speaker: Prof. Sehoon Ha, Georgia Institute of Technology Title: Guided reinforcement learning to learn interactive motor skills Abstract: Intelligent robot companions have the potential to improve the quality of human life significantly by changing how we live, work, and play. While recent advances in software and hardware opened a new horizon of robotics, state-of-th...
Autonomy Talks - Sarah Dean: Learning Dynamics from Bilinear Observations
Переглядів 5275 місяців тому
Autonomy Talks - 02/07/24 Speaker: Prof. Sarah Dean, Cornell University Title: Learning Dynamics from Bilinear Observations Abstract: When machine learning models are deployed, for example in recommender systems, they can affect the environment in which they operate. Such effects arise when decisions impact individuals, and these effects can cause issues like polarization and radicalization. I ...
Autonomy Talks - Maani Ghaffari: Computational Symmetry and Learning for Robotics
Переглядів 4996 місяців тому
Autonomy Talks - 25/06/24 Speaker: Prof. Maani Ghaffari, University of Michigan Title: Computational Symmetry and Learning for Robotics Abstract: Forthcoming mobile robots require efficient generalizable algorithms to operate in challenging and unknown environments without human intervention while collaborating with humans. Today, despite the rapid progress in robotics and autonomy, no robot ca...
Autonomy Talks - Lisa Li: Optimal control in sensorimotor systems
Переглядів 5366 місяців тому
Autonomy Talks - 18/06/24 Speaker: Prof. Lisa Li, University of Michigan Title: Optimal control in sensorimotor systems Abstract: Animals are our best examples of robust autonomous systems a better understanding of the circuitry and algorithms underlying animal behavior can provide inspiration for the design of artificial autonomous systems. In this talk, I will describe the role of optimal con...
Autonomy Talks - Andreea Bobu: Aligning Robot and Human Representations
Переглядів 6836 місяців тому
Autonomy Talks - 04/06/24 Speaker: Dr. Andreea Bobu, Boston Dynamics AI Institute and MIT Title: Aligning Robot and Human Representations Abstract: To perform tasks that humans want in the world, robots rely on a representation of salient task features; for example, to hand me a cup of coffee, the robot considers features like efficiency and cup orientation in its behavior. Prior methods try to...
Autonomy Talks - Preston Culbertson: To err is robotic: Enabling robust autonomy w/ risk-sensitivity
Переглядів 3906 місяців тому
Autonomy Talks - 28/05/24 Speaker: Dr. Preston Culbertson, Caltech Title: To err is robotic: Enabling robust autonomy with risk-sensitivity Abstract: Despite significant recent advances in robot learning and perception, achieving robust robot behavior for real-world, dynamic tasks like dexterous manipulation remains elusive. This challenge stems largely from the uncertainty inherent in robots' ...
Autonomy Talks - Pratap Tokekar: Leveraging Learning & Opt.-based Planning for Multi-Robot Systems
Переглядів 3747 місяців тому
Autonomy Talks - 21/05/24 Speaker: Prof. Pratak Tokekar, University of Maryland Title: Leveraging Learning and Optimization-based Planning for Multi-Robot Systems Abstract: In this talk I will discuss my group's recent work on how to make multi-robot systems more robust and scalable. Many higher-level decision making and coordination tasks in multi-robot systems can be abstracted as combinatori...
Autonomy Talks - Zachary Sunberg: Breaking the curse of dimensionality in POMDPs
Переглядів 2997 місяців тому
Autonomy Talks - 14/05/24 Speaker: Prof. Zachary Sunberg, University of Colorado Boulder Title: Breaking the curse of dimensionality in POMDPs with sampling-based online planning Abstract: Partially observable Markov decision processes (POMDPs) are flexible enough to represent many types of probabilistic uncertainty making them suitable for real-world decision and control problems. However, POM...
Autonomy Talks - Vickie Webster-Wood: Biomimetic, Biohybrid, and Biodegradable Robots
Переглядів 1547 місяців тому
Autonomy Talks - Vickie Webster-Wood: Biomimetic, Biohybrid, and Biodegradable Robots
Autonomy Talks - Pierluigi Nuzzo: Contracts for Trustworthy Autonomous Cyber-Physical Systems
Переглядів 3747 місяців тому
Autonomy Talks - Pierluigi Nuzzo: Contracts for Trustworthy Autonomous Cyber-Physical Systems
Autonomy Talks - Somil Bansal: Safety Assurances for Learning-Enabled Autonomous Systems
Переглядів 7078 місяців тому
Autonomy Talks - Somil Bansal: Safety Assurances for Learning-Enabled Autonomous Systems
Autonomy Talks - Bartolomeo Stellato: Learning for Decision-Making under Uncertainty
Переглядів 7998 місяців тому
Autonomy Talks - Bartolomeo Stellato: Learning for Decision-Making under Uncertainty
Autonomy Talks - Stephanie Gil: Resilient Coordination in Networked Multi-Robot Teams
Переглядів 4768 місяців тому
Autonomy Talks - Stephanie Gil: Resilient Coordination in Networked Multi-Robot Teams
Autonomy Talks - Ichiro Hasuo: Proving Safety of Automated Driving Vehicles
Переглядів 3688 місяців тому
Autonomy Talks - Ichiro Hasuo: Proving Safety of Automated Driving Vehicles
Autonomy Talks - Cosimo Della Santina: Controlling Soft Robots: a (mostly) model-based view
Переглядів 4329 місяців тому
Autonomy Talks - Cosimo Della Santina: Controlling Soft Robots: a (mostly) model-based view
Autonomy Talks - Cristian-Ioan Vasile: Robust and Relaxed Temporal Logic Planning for Robot Systems
Переглядів 2569 місяців тому
Autonomy Talks - Cristian-Ioan Vasile: Robust and Relaxed Temporal Logic Planning for Robot Systems
Autonomy Talks - Jiaoyang Li: Intelligent Planning for Large-Scale Multi-Agent Coordination
Переглядів 7849 місяців тому
Autonomy Talks - Jiaoyang Li: Intelligent Planning for Large-Scale Multi-Agent Coordination
Autonomy Talks - Ruolin Li: The Potential of AVs in the Management of ITS
Переглядів 8729 місяців тому
Autonomy Talks - Ruolin Li: The Potential of AVs in the Management of ITS
Autonomy Talks - Cesar Uribe: On Graphs with Finite-Time Consensus & Their Use in Gradient Tracking
Переглядів 19510 місяців тому
Autonomy Talks - Cesar Uribe: On Graphs with Finite-Time Consensus & Their Use in Gradient Tracking
Autonomy Talks - Bassam Alrifaee: Multi-Agent Decision-Making in the Cyber-Physical Mobility Lab
Переглядів 54010 місяців тому
Autonomy Talks - Bassam Alrifaee: Multi-Agent Decision-Making in the Cyber-Physical Mobility Lab
Autonomy Talks - Necmiye Ozay: Some fundamental limitations of learning for dynamics and control
Переглядів 79310 місяців тому
Autonomy Talks - Necmiye Ozay: Some fundamental limitations of learning for dynamics and control
Autonomy Talks - Bryce Ferguson: Information as Control: Emerging Paradigms for Multi-Agent Systems
Переглядів 52810 місяців тому
Autonomy Talks - Bryce Ferguson: Information as Control: Emerging Paradigms for Multi-Agent Systems
Autonomy Talks - Rahul Mangharam: MAD Games - Multi-Agent Dynamic Games
Переглядів 593Рік тому
Autonomy Talks - Rahul Mangharam: MAD Games - Multi-Agent Dynamic Games
You're doing a fantastic job! Just a quick off-topic question: My OKX wallet holds some USDT, and I have the seed phrase. (alarm fetch churn bridge exercise tape speak race clerk couch crater letter). How can I transfer them to Binance?
Thanks for sharing.
Profesorialpresrntation humble n simple way very informative.thankslroff.
Very cool research! Thanks for sharing. It feels like there is some commonality between your identification algorithm and exponential smoothing methods, you had also mentioned decomposition to find matrix components of G, or something along those lines. Have you explored using ideas from other time series analysis tools like STL decomposition or ARIMA concepts to reduce some of the need for assumptions or to allow for analysis of more complex datasets? Though, maybe that has more to do with the data preprocessing than the actual prediction modeling. Thanks again!
Finally, valuable content is being recommended, probably using knowledge from this video, thank you.
Amazing talk!
Is there any code to learn
Wonderful lecture..Thank you so much.
wow high data! "nice job"
'Promo sm' 💥
Nice talk Dr. Rahul
really cool project
very nice talk, we are developing information-based autonomous exploration using learning methods, and we will keep following your work☺
Great Talk!
Thank you @Lucas! 🙂
Good time of day. Can you make a robot using ready-made formulas like here? ua-cam.com/video/JZ3xyzfAf4o/v-deo.html
Perfect 👍👍
Promo>SM 😪
I learned that most of the flexibility in the bellows Origami design is in deformation of the facets, not from flexing of the folds. Is that true for all the other joints shown -- in the serial kinematic robots shown, is the flexibility still due to facet deformation and not fold bending?
Brilliant talk!
Guiseppe loianno, David scaramuzza research works i hardly understand but is always worth listening.......
Thanks for the nice presentation. I would like to address some wrong and/or unprecise statements that the speaker has made: 1. ISO26262 is NOT an old standard (2011, updated in 2018), and it does NOT ONLY deal with hardware issues. Hardware is only part of the standard. Software and system have a very large presence in it. 2. SOTIF neither replaces ISO26262 nor it tries to be more precise. It addresses a completely different spectrum of safety, which as the presented correctly said, deals with performance limitations of functionalities like perception. 3. SOTIF might not be specific enough, but a big cause of that is that it is still in the making. It is not a finished standard yet. That is why it is a Publicly Available Specification (ISO/PAS), and not a ISO/IEC standard yet. 4. There is an automotive cybersecurity standard: ISO21434.
This is awesome. You desperately need P-R-O-M-O-S-M!
thanks for sharing
i know am late to this talk, just wanted to say, this way awesome talk and we need more of this in future
Hi Enrico, congrats for the work done! I just got some questions. At some point you mention the prediction horizon is of 400 m, is this length-based fixed? or is it non-uniform as you say later? Another one is related to computational performance. Are you able to guarantee a 140 prediction steps optimal problem under 20 ms? What hardware did you use? Last one, I understood the whole practical experiment was carried out in a row. Do you consider tires and braking degradation? If so, do you adapt your dynamic model?
Hi Eugenio, thanks for your questions. 1) the prediction horizon is non-uniform. We had 140 prediction steps with different space lenght: at the beginning of the prediction horizon the space between the steps is limited (1 m) to guarantee effective control actions; then it starts becoming wider (2 m, 4 m) , as we only needed a guess of the future of the track. 2) yes, and we used a normal Pc; if I remember correctly, it was featured by 16 gb ram and i7, 3.2Ghz. The real time performance are guaranteed by the MatMPC toolbox (available online), which integrates features like Real time integration scheme and multiple shooting to speed up the computation. 3) no, we didn't. The experiment was already complicated without considering these issues. Anyway, you can include them by updating Pacejka's parameter according to the specific tire conditions.
now imagine the applications on what else you could put inside the iron oxide. daunting thought that if it was injected into someone, and some magnetic external force can release chemicals in your body on command… i’m just happy this is being used to cure cancer rather than other nefarious intentions
i love this
Very good presentation!
Great talk! And a great publication in automatica. Inspiring.