Trajectory optimization for motion planning requires good initial guesses to obtain performance. In our proposed approach, we build a memory of based on database robot paths provide guesses. The relies function approximators and dimensionality reduction techniques learn the mapping between tasks paths. Three are compared: k-Nearest Neighbor, Gaussian Process Regression, Bayesian Mixture Regression. addition, show that can be used as metric choose several possible goals, using an ensemble...

A common approach to learn robotic skills is imitate a demonstrated policy. Due the compounding of small errors and perturbations, this may let robot leave states in which demonstrations were provided. This requires consideration additional strategies guarantee that will behave appropriately when facing unknown states. We propose use Bayesian method quantify action uncertainty at each state. The proposed simple set up, computationally efficient, can adapt wide range problems. Our exploits...

For a safe and successful daily living assistance, far from the highly controlled environment of factory, robots should be able to adapt ever-changing situations. Programming such robot is tedious process that requires expert knowledge. An alternative rely on high-level planner, but generic symbolic representations used are not well suited particular executions. Contrarily, motion primitives encode motions in way can easily adapted different This paper presents combined framework exploits...

In high dimensional robotic system, the manifold of valid configuration space often has a complex shape, especially under constraints such as end-effector orientation or static stability. We propose generative adversarial network approach to learn distribution robot configurations constraints. It can generate that are close constraint manifold. present two applications this method. First, by learning conditional with respect desired position, we do fast inverse kinematics even for very...

Mapping operator motions to a robot is key problem in teleoperation. Due differences between local and remote workspaces, such as object locations, it particularly challenging derive smooth motion mappings that fulfill different goals (e.g., picking objects with poses on the two sides or passing through points). Indeed, most state-of-the-art methods rely mode switches, leading discontinuous, low-transparency experience. In this letter, we propose unified formulation for position, orientation...

Probability distributions are key components of many learning from demonstration (LfD) approaches, with the spaces chosen to represent tasks playing a central role. Although robot configuration is defined by its joint angles, end-effector poses often best explained within several task spaces. In relevant learned independently and only combined at control level. This simplification implies problems that addressed in this work. We show fusion models different can be expressed as products...

Limited time-resolution in microscopy is an obstacle to many biological studies. Despite recent advances hardware, digital cameras have limited operation modes that constrain frame-rate, integration time, and color sensing patterns. In this paper, we propose approach extend the temporal resolution of a conventional camera by leveraging multi-color illumination source. Our method allows for imaging single-hue objects at increased frame-rate trading spectral information (while retaining...

In this paper we show how different choices regarding compliance affect a dual-arm assembly task. addition, present the parameters can be learned from human demonstration. Compliant motions used in tasks to mitigate pose errors originating from, for example, inaccurate grasping. We analytical background and accompanying experimental results on choose center of enhance convergence region an alignment Then possible ways choosing compliant axes accomplishing scenario where orientation error is...

We propose to formulate the problem of representing a distribution robot configurations (e.g. joint angles) as that approximating product experts. Our approach uses variational inference, popular method in Bayesian computation, which has several practical advantages over sampling-based techniques. To be able represent complex and multimodal distributions configurations, mixture models are used approximate distribution. show while satisfying multiple objectives arises wide range problems...

Learning from demonstration (LfD) is an intuitive framework allowing non-expert users to easily (re-)program robots. However, the quality and quantity of demonstrations have a great influence on generalization performances LfD approaches. In this paper, we introduce novel active learning in order improve capabilities control policies. The proposed approach based epistemic uncertainties Bayesian Gaussian mixture models (BGMMs). We determine new query point location by optimizing closed-form...

This paper proposes an inverse reinforcement learning (IRL) framework to accelerate when the learner-teacher \textit{interaction} is \textit{limited} during training. Our setting motivated by realistic scenarios where a helpful teacher not available or cannot access dynamics of student. We present two different training strategies: Curriculum Inverse Reinforcement Learning (CIRL) covering teacher's perspective, and Self-Paced (SPIRL) focusing on learner's perspective. Using experiments in...

In learning from demonstrations, many generative models of trajectories make simplifying assumptions independence. Correctness is sacrificed in the name tractability and speed phase. The ignored dependencies, which often are kinematic dynamic constraints system, then only restored when synthesizing motion, introduces possibly heavy distortions. this work, we propose to use those approximate trajectory distributions as close-to-optimal discriminators popular adversarial framework stabilize...

We propose a method to approximate the distribution of robot configurations satisfying multiple objectives. Our approach uses variational inference, popular in Bayesian computation, which has several advantages over sampling-based techniques. To be able represent complex and multimodal configurations, we use mixture model as distribution, an that gained popularity recently. In this work, show interesting properties how it can applied wide range problems robotics.

In this paper we show how different choices regarding compliance affect a dual-arm assembly task. addition, present the parameters can be learned from human demonstration. Compliant motions used in tasks to mitigate pose errors originating from, for example, inaccurate grasping. We analytical background and accompanying experimental results on choose center of enhance convergence region an alignment Then possible ways choosing compliant axes accomplishing scenario where orientation error is...