A5043984392- Reinforcement Learning in Robotics
- Robotics and Sensor-Based Localization
- Robotic Path Planning Algorithms
- Artificial Intelligence in Games
- Game Theory and Applications
- Sports Analytics and Performance
- Modular Robots and Swarm Intelligence
- Auction Theory and Applications
- Advanced Bandit Algorithms Research
- Evolutionary Algorithms and Applications
- Spacecraft Dynamics and Control
- Space Satellite Systems and Control
- Evolutionary Game Theory and Cooperation
- Distributed Control Multi-Agent Systems
- Anomaly Detection Techniques and Applications
- Insect and Arachnid Ecology and Behavior
- Astro and Planetary Science
- Sports Performance and Training
- Experimental Behavioral Economics Studies
- Domain Adaptation and Few-Shot Learning
- Complex Systems and Time Series Analysis
- Advanced Vision and Imaging
- Optimization and Search Problems
- Generative Adversarial Networks and Image Synthesis
- Intelligent Tutoring Systems and Adaptive Learning
DeepMind (United Kingdom)
2020-2024
Google (United Kingdom)
2024
Google (United States)
2019-2021
University of Stuttgart
2018-2019
European Space Research and Technology Centre
2014-2018
German Research Centre for Artificial Intelligence
2016-2017
European Space Agency
2015-2016
University of Bremen
2016
Maastricht University
2007-2013
Eindhoven University of Technology
2008-2009
The interaction of multiple autonomous agents gives rise to highly dynamic and nondeterministic environments, contributing the complexity in applications such as automated financial markets, smart grids, or robotics. Due sheer number situations that may arise, it is not possible foresee program optimal behaviour for all beforehand. Consequently, becomes essential success system can learn their adapt new circumstances. past two decades have seen emergence reinforcement learning, both single...
We introduce DeepNash, an autonomous agent capable of learning to play the imperfect information game Stratego from scratch, up a human expert level. is one few iconic board games that Artificial Intelligence (AI) has not yet mastered. This popular enormous tree on order $10^{535}$ nodes, i.e., $10^{175}$ times larger than Go. It additional complexity requiring decision-making under information, similar Texas hold'em poker, which significantly smaller (on $10^{164}$ nodes). Decisions in are...
OpenSpiel is a collection of environments and algorithms for research in general reinforcement learning search/planning games. supports n-player (single- multi- agent) zero-sum, cooperative general-sum, one-shot sequential, strictly turn-taking simultaneous-move, perfect imperfect information games, as well traditional multiagent such (partially- fully- observable) grid worlds social dilemmas. also includes tools to analyze dynamics other common evaluation metrics. This document serves both...
The rapid progress in artificial intelligence (AI) and machine learning has opened unprecedented analytics possibilities various team individual sports, including baseball, basketball, tennis. More recently, AI techniques have been applied to football, due a huge increase data collection by professional teams, increased computational power, advances learning, with the goal of better addressing new scientific challenges involved analysis both players’ coordinated teams’ behaviors. research...
Learning to combine control at the level of joint torques with longer-term goal-directed behavior is a long-standing challenge for physically embodied artificial agents. Intelligent in physical world unfolds across multiple spatial and temporal scales: Although movements are ultimately executed instantaneous muscle tensions or torques, they must be selected serve goals that defined on much longer time scales often involve complex interactions environment other Recent research has...
Abstract Identifying key patterns of tactics implemented by rival teams, and developing effective responses, lies at the heart modern football. However, doing so algorithmically remains an open research challenge. To address this unmet need, we propose TacticAI, AI football assistant developed evaluated in close collaboration with domain experts from Liverpool FC. We focus on analysing corner kicks, as they offer coaches most direct opportunities for interventions improvements. TacticAI...
This paper describes a multi-robot collision avoidance system based on the velocity obstacle paradigm. In contrast to previous approaches, we alleviate strong requirement for perfect sensing (i.e. global positioning) using Adaptive Monte-Carlo Localization per-agent level. While such methods as Optimal Reciprocal Collision Avoidance guarantee local collision-free motion large number of robots, given knowledge positions and speeds, realistic implementation requires further extensions deal...
We present a multi-mobile robot collision avoidance system based on the velocity obstacle paradigm. Current positions and velocities of surrounding robots are translated to an efficient geometric representation determine safe motions. Each uses on-board localization local communication build its surroundings. Our close error-bounded convex approximation density distribution results in collision-free paths under uncertainty. While many algorithms approximated by circumscribed radii, we use...
The space close to our planet is getting more and polluted. Orbiting debris are posing an increasing threat operational orbits the cascading effect, known as Kessler syndrome, may result in a future where risk of orbiting at some altitudes will be unacceptable. Many argue that density Low Earth Orbit (LEO) has already reached level sufficient trigger such effect. An obvious consequence we soon have actively clean from debris. Such mission involve complex combinatorial decision choose which...
The use of soft robots in future space exploration is still a far-fetched idea, but an attractive one. Soft are inherently compliant mechanisms that well suited for locomotion on rough terrain as often faced extra-planetary environments. Depending the particular application and requirements, best shape (or body morphology) strategy such will vary substantially. Recent developments robotics evolutionary optimization showed possibility to simultaneously evolve morphology simulated trials....
This paper investigates a population-based training regime based on game-theoretic principles called Policy-Spaced Response Oracles (PSRO). PSRO is general in the sense that it (1) encompasses well-known algorithms such as fictitious play and double oracle special cases, (2) principle applies to general-sum, many-player games. Despite this, prior studies of have been focused two-player zero-sum games, wherein Nash equilibria are tractably computable. In moving from games more settings,...
Although multi-robot systems have received substantial research attention in recent years, coordination still remains a difficult task. Especially, when dealing with spatially distributed tasks and many robots, central control quickly becomes infeasible due to the exponential explosion number of joint actions states. We propose general algorithm that allows for control, overcomes growth by aggregating effect other agents system into probabilistic model, called subjective approximations, then...
We investigate the use of deep artificial neural networks to approximate optimal state-feedback control continuous time, deterministic, non-linear systems. The are trained in a supervised manner using trajectories generated by solving problem via Hermite-Simpson transcription method. find that able represent with high accuracy and precision well outside training area. consider dynamical models under different cost functions result both smooth discontinuous (bang-bang) solutions. In...
Recent progress in deep reinforcement learning has enabled simulated agents to learn complex behavior policies from scratch, but their data complexity often prohibits real-world applications. The process can be sped up by expert demonstrations those costly acquire. We demonstrate that it is possible employ model-free combined with planning quickly generate informative for a manipulation task. In particular, we use an approximate trajectory optimization approach global exploration based on...
The on-chip fabrication and manipulation of microstructures are expected to be applied for single cell analysis system such as measurement tools. In this paper, we previously present a methodology fabricating assembling inside microfluidic channel. By the illumination patterned UV-ray through mask under microscope, with arbitrary shape made photo-crosslinkable resin device. fabricated at desired place channel manipulated by optical tweezers. Based on technique which can manipulate multiple...
The focus of this paper is on situational awareness airborne agents capable 6D motion, in particular multi-rotor UAVs. We propose the fusion 2D laser range finder, altitude, and attitude sensor data order to perform simultaneous localization mapping (SLAM) indoors. In contrast other planar finder based SLAM approaches, we a 3D instead map. To represent environment an octree map used. Our scan registration algorithm derived from Hector SLAM. evaluate performance our system simulation real...
Since early in robotics the performance of odometry techniques has been constant research for mobile robots. This is due to its direct influence on localization. The pose error grows unbounded dead-reckoning systems and uncertainty negative impacts localization mapping (i.e. SLAM). terms residuals, i.e. difference between expected real state, related statistical or probabilistic motion models. A novel approach model errors using Gaussian processes (GPs) presented. methodology trains a GP...
We consider the problem of optimally transferring a spacecraft from starting to target asteroid. introduce novel approximations for important quantities characterizing optimal transfer in case short times (asteroid hops). propose and study detail phasing value φ, maximum initial mass m* arrival m <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">f</sub> . The new require orders magnitude less computational effort with respect state-of-the-art...
In the present work, we propose an active tactile exploration framework to obtain a surface model of unknown object utilizing multiple contacts simultaneously. To incorporate these contacts, strategy is based on differential entropy underlying Gaussian process implicit model, which formalizes with within information theoretic context and additionally allows for nonmyopic multi-step planning. contrast many previous approaches, robot continuously slides along its end-effectors gather stimuli,...
The on-chip fabrication and manipulation of microstructures are expected to be applied for single cell analysis system such as measurement tools. In this paper, we previously present a methodology fabricating assembling inside microfluidic channel. By the illumination patterned UV-ray through mask under microscope, with arbitrary shape made photo-crosslinkable resin device. fabricated at desired place channel manipulated by optical tweezers. Based on technique which can manipulate multiple...
Self-supervised learning is a reliable mechanism in which robot uses an original, trusted sensor cue for training to recognize additional, complementary cue. We study the first time self-supervised how robot’s behavior should be organized, so that can keep performing its task case original becomes unavailable. this persistent form of context flying has avoid obstacles based on distance estimates from visual stereo vision. Over it will learn also estimate distances monocular appearance cues....