• Publications
  • Influence
An Algorithmic Perspective on Imitation Learning
TLDR
This work provides an introduction to imitation learning, dividing imitation learning into directly replicating desired behavior and learning the hidden objectives of the desired behavior from demonstrations (called inverse optimal control or inverse reinforcement learning [Russell, 1998]). Expand
Periodic Finite State Controllers for Efficient POMDP and DEC-POMDP Planning
TLDR
This paper introduces a novel class of periodic FSCs, composed of layers connected only to the previous and next layer, and finds a deterministic finite-horizon policy and converts it to an initial periodic infinite-Horizon policy. Expand
Learning in-contact control strategies from demonstration
TLDR
A model based on the hidden semi-Markov model (HSMM) and Cartesian impedance control is proposed that captures uncertainty over time and space and allows the robot to smoothly satisfy a task's position and force constraints by online modulation of impedance controller stiffness according to the HSMM state belief. Expand
Efficient Planning for Factored Infinite-Horizon DEC-POMDPs
TLDR
This work formulate expectation-maximization based optimization into a new form, where complexity can be kept tractable by factored approximations, and gives results for factored infinite-horizon DEC-POMDP problems with up to 10 agents. Expand
Robotic manipulation of multiple objects as a POMDP
TLDR
The results indicate that: 1) a greedy heuristic manipulation approach is not sufficient, multi-object manipulation requires multi-step POMDP planning, and 2) on-line planning is beneficial since it allows the adaptation of the system dynamics model based on actual experience. Expand
Hybrid control trajectory optimization under uncertainty
TLDR
The method, based on Differential Dynamic Programming (DDP), circumvents this problem by incorporating discrete actions inside DDP: first optimize continuous mixtures of discrete actions, and, subsequently force the mixtures into fully discrete actions. Expand
Learning Intention Aware Online Adaptation of Movement Primitives
TLDR
A goal-based intention prediction model from human motions is learned and intention-aware online adaptation to ProMPs is introduced, which leads to a higher level of perceived safety and felt less disturbed during intention aware adaptation, in particular during spatial deformation, compared to non-adaptive behavior of the robot. Expand
Sparse Latent Space Policy Search
TLDR
A reinforcement learning method for sample-efficient policy search that exploits correlations between control variables, particularly frequent in motor skill learning tasks, and outperforms state-of-the-art policy search methods. Expand
Expectation Maximization for Average Reward Decentralized POMDPs
TLDR
It is shown that under a common set of conditions expectation maximization (EM) for average reward Dec-POMDPs is stuck in a local optimum, and a new average reward EM method is introduced that outperforms a state of the art discounted-reward Dec- POMDP method in experiments. Expand
Model-based Lookahead Reinforcement Learning
TLDR
This work combines MFRL and Model Predictive Control and proposes an approach that can achieve MFRL`s level of performance while being as data-efficient as MBRL. Expand
...
1
2
3
4
5
...