• Publications
  • Influence
Robot trajectory optimization using approximate inference
This work considers a probabilistic model for which the maximum likelihood (ML) trajectory coincides with the optimal trajectory and which reproduces the classical SOC solution and utilizes approximate inference methods that efficiently generalize to non-LQG systems. Expand
Probabilistic inference for solving discrete and continuous state Markov Decision Processes
An Expectation Maximization algorithm for computing optimal policies that actually optimizes the discounted expected future return for arbitrary reward functions and without assuming an ad hoc finite total time is presented. Expand
On Stochastic Optimal Control and Reinforcement Learning by Approximate Inference
We present a reformulation of the stochastic optimal control problem in terms of KL divergence minimisation, not only providing a unifying perspective of previous approaches in this area, but alsoExpand
Multi-class image segmentation using conditional random fields and global classification
This work presents an approach to multi-class segmentation which combines two methods for this integration: a Conditional Random Field which couples to local image features and an image classification method which considers global features. Expand
Probabilistic inference for solving (PO) MDPs
The approach is based on an equivalence between maximization of the expected future return in the time-unlimited MDP and likelihood maximization in a related mixture of finite-time MDPs, which allows to use expectation maximization (EM) for computing optimal policies, using arbitrary inference techniques in the E-step. Expand
Differentiable Physics and Stable Modes for Tool-Use and Manipulation Planning
We propose to formulate physical reasoning and manipulation planning as an optimization problem that integrates first order logic, which we call Logic-Geometric Programming.
Inverse KKT: Learning cost functions of manipulation tasks from demonstrations
A non-parametric variant of inverse KKT that represents the cost function as a functional in reproducing kernel Hilbert spaces is presented, to push learning from demonstration to more complex manipulation scenarios that include the interaction with objects and therefore the realization of contacts/constraints within the motion. Expand
Planning as inference
The topic of planning is brought back to center stage in cognitive science, according to which planning is accomplished through probabilistic inference, and a potentially transformative new idea is introduced. Expand
Gaussian process implicit surfaces for shape estimation and grasping
This work considers Gaussian process implicit surface potentials as object shape representations and validates the shape estimation using Gaussian processes in a simulation on randomly sampled shapes and the grasp controller on a real robot with 7 doF arm and 7DoF hand. Expand
A Sensorimotor Map: Modulating Lateral Interactions for Anticipation and Planning
The proposed sensorimotor map learns a state representation similar to self-organizing maps but is inherently coupled to sensor and motor signals, and encodes a model of the change of stimuli depending on the current motor activities. Expand