Micha Hersch

Dynamical System Modulation for Robot Learning via Kinesthetic Demonstrations

We present a system for robust robot skill acquisition from kinesthetic demonstrations. This system allows a robot to learn a simple goal-directed gesture and correctly reproduce it despite changes in the initial conditions and perturbations in the environment. It combines a dynamical system control approach with tools of statistical learning theory and provides a solution to the inverse kinematics problem when dealing with a redundant manipulator. The system is validated on two experiments involving a humanoid robot: putting an object into a box and reaching for and grasping an object.

Phytochrome Kinase Substrate 4 is phosphorylated by the phototropin 1 photoreceptor

Phototropism allows plants to redirect their growth towards the light to optimize photosynthesis under reduced light conditions. Phototropin 1 (phot1) is the primary low blue light‐sensing receptor triggering phototropism in Arabidopsis. Light‐induced autophosphorylation of phot1, an AGC‐class protein kinase, constitutes an essential step for phototropism. However, apart from the receptor itself, substrates of phot1 kinase activity are less clearly established. Phototropism is also influenced by the cryptochromes and phytochromes photoreceptors that do not provide directional information but influence the process through incompletely characterized mechanisms. Here, we show that Phytochrome Kinase Substrate 4 (PKS4), a known element of phot1 signalling, is a substrate of phot1 kinase activity in vitro that is phosphorylated in a phot1‐dependent manner in vivo. PKS4 phosphorylation is transient and regulated by a type 2‐protein phosphatase. Moreover, phytochromes repress the accumulation of the light‐induced phosphorylated form of PKS4 showing a convergence of photoreceptor activity on this signalling element. Our physiological analyses suggest that PKS4 phosphorylation is not essential for phototropism but is part of a negative feedback mechanism.

Nuclear Phytochrome A Signaling Promotes Phototropism in Arabidopsis

This article shows that the phytochrome A photoreceptor promotes reorientation of the hypocotyl toward blue light (phototropism) by regulating the expression of nuclear genes. It also shows that phytochrome A nuclear signaling events still operate in a mutant where phytochrome A does not significantly accumulate in the nucleus. Phototropin photoreceptors (phot1 and phot2 in Arabidopsis thaliana) enable responses to directional light cues (e.g., positive phototropism in the hypocotyl). In Arabidopsis, phot1 is essential for phototropism in response to low light, a response that is also modulated by phytochrome A (phyA), representing a classical example of photoreceptor coaction. The molecular mechanisms underlying promotion of phototropism by phyA remain unclear. Most phyA responses require nuclear accumulation of the photoreceptor, but interestingly, it has been proposed that cytosolic phyA promotes phototropism. By comparing the kinetics of phototropism in seedlings with different subcellular localizations of phyA, we show that nuclear phyA accelerates the phototropic response, whereas in the fhy1 fhl mutant, in which phyA remains in the cytosol, phototropic bending is slower than in the wild type. Consistent with this data, we find that transcription factors needed for full phyA responses are needed for normal phototropism. Moreover, we show that phyA is the primary photoreceptor promoting the expression of phototropism regulators in low light (e.g., PHYTOCHROME KINASE SUBSTRATE1 [PKS1] and ROOT PHOTO TROPISM2 [RPT2]). Although phyA remains cytosolic in fhy1 fhl, induction of PKS1 and RPT2 expression still occurs in fhy1 fhl, indicating that a low level of nuclear phyA signaling is still present in fhy1 fhl.

Light intensity modulates the regulatory network of the shade avoidance response in Arabidopsis

Significance Plants sense foliar shade and neighbors who may become competitors for light. Shade-sensitive species elongate in response to both situations to enhance access to unfiltered sunlight, which is known as the shade avoidance response (SAR). During neighbor detection, plants have access to plenty of light (energy resources), whereas in true shade, light resources are scarce. Our analysis of the molecular mechanisms underlying SAR under these contrasting conditions shows that light intensity balances the production and sensitivity of the growth hormone auxin. In foliar shade, the production of auxin is reduced, whereas the downstream sensitivity to the auxin signal is enhanced. This hints at a resource-aware signaling where the strength of the hormonal signal is tuned to the available resources. Plants such as Arabidopsis thaliana respond to foliar shade and neighbors who may become competitors for light resources by elongation growth to secure access to unfiltered sunlight. Challenges faced during this shade avoidance response (SAR) are different under a light-absorbing canopy and during neighbor detection where light remains abundant. In both situations, elongation growth depends on auxin and transcription factors of the phytochrome interacting factor (PIF) class. Using a computational modeling approach to study the SAR regulatory network, we identify and experimentally validate a previously unidentified role for long hypocotyl in far red 1, a negative regulator of the PIFs. Moreover, we find that during neighbor detection, growth is promoted primarily by the production of auxin. In contrast, in true shade, the system operates with less auxin but with an increased sensitivity to the hormonal signal. Our data suggest that this latter signal is less robust, which may reflect a cost-to-robustness tradeoff, a system trait long recognized by engineers and forming the basis of information theory.

Reaching with multi-referential dynamical systems

Abstract We study a reaching movement controller for a redundant serial arm manipulator, based on two principles believed to be central to biological motion control: multi-referential control and dynamical system control. The resulting controller is based on two concurrent dynamical systems acting on different, yet redundant variables. The first dynamical system acts on the end-effector location variables and the second one acts on the joint angle variables. Coherence constraints are enforced between those two redundant representations of the movement and can be used to modulate the relative influence of each dynamical system. We illustrate the advantages of such a redundant representation of the movement regarding singularities and joint angle avoidance.

Distinct levels in Pom1 gradients limit Cdr2 activity and localization to time and position division

Where and when cells divide are fundamental questions. In rod-shaped fission yeast cells, the DYRK-family kinase Pom1 is organized in concentration gradients from cell poles and controls cell division timing and positioning. Pom1 gradients restrict to mid-cell the SAD-like kinase Cdr2, which recruits Mid1/Anillin for medial division. Pom1 also delays mitotic commitment through Cdr2, which inhibits Wee1. Here, we describe quantitatively the distributions of cortical Pom1 and Cdr2. These reveal low profile overlap contrasting with previous whole-cell measurements and Cdr2 levels increase with cell elongation, raising the possibility that Pom1 regulates mitotic commitment by controlling Cdr2 medial levels. However, we show that distinct thresholds of Pom1 activity define the timing and positioning of division. Three conditions—a separation-of-function Pom1 allele, partial downregulation of Pom1 activity, and haploinsufficiency in diploid cells—yield cells that divide early, similar to pom1 deletion, but medially, like wild-type cells. In these cells, Cdr2 is localized correctly at mid-cell. Further, Cdr2 overexpression promotes precocious mitosis only in absence of Pom1. Thus, Pom1 inhibits Cdr2 for mitotic commitment independently of regulating its localization or cortical levels. Indeed, we show Pom1 restricts Cdr2 activity through phosphorylation of a C-terminal self-inhibitory tail. In summary, our results demonstrate that distinct levels in Pom1 gradients delineate a medial Cdr2 domain, for cell division placement, and control its activity, for mitotic commitment.

Cell-Cell Contact Area Affects Notch Signaling and Notch-Dependent Patterning

Summary During development, cells undergo dramatic changes in their morphology. By affecting contact geometry, these morphological changes could influence cellular communication. However, it has remained unclear whether and how signaling depends on contact geometry. This question is particularly relevant for Notch signaling, which coordinates neighboring cell fates through direct cell-cell signaling. Using micropatterning with a receptor trans-endocytosis assay, we show that signaling between pairs of cells correlates with their contact area. This relationship extends across contact diameters ranging from microns to tens of microns. Mathematical modeling predicts that dependence of signaling on contact area can bias cellular differentiation in Notch-mediated lateral inhibition processes, such that smaller cells are more likely to differentiate into signal-producing cells. Consistent with this prediction, analysis of developing chick inner ear revealed that ligand-producing hair cell precursors have smaller apical footprints than non-hair cells. Together, these results highlight the influence of cell morphology on fate determination processes.

Learning Dynamical System Modulation for Constrained Reaching Tasks

In this paper we combine kinesthetic demonstrations and dynamical systems to enable a humanoid robot to imitate constrained reaching gestures directed toward a target. Using a learning algorithm based on Gaussian mixture regression, the task constraints are extracted from several demonstrations. Those constraints take the form of desired velocity profiles for the end-effector and joint angle variables, with associated covariance matrices describing the variations allowed around the desired profiles. Those constraints are then used to modulate a dynamical system which has the reaching target as attractor. This way, the reaching trajectory can be reshaped in order to satisfy the constraints of the task, while preserving the adaptability and robustness provided by the dynamical system. In particular, the system can adapt to changes in the initial conditions and to target displacements occurring during the movement execution. We first evaluate the potential of this method on experiments involving the Hoap3 humanoid robot putting an object into a box. We then show how a manipulation tasks can be executed as sequences of such constrained reaching movement. This is illustrated on a packaging task performed by the robot

A Biologically-Inspired Controller for Reaching Movements

This paper presents a robotic arm controller for reaching motions. This controller applies two principles inspired by current theories of human reaching motions, namely the multi-referential control of movements and the dynamical system approach to biological control. The controller consists of a stable dynamical system active in a hybrid cartesian-joint angle frame of reference. Our results show that this controller has interesting properties in terms of stability and robustness to perturbations, and that its redundancy can be exploited for a simple solution to the joint limit avoidance problem

Extended hopfield network for sequence learning: application to gesture recognition

In this paper, we extend the Hopfield Associative Memory for storing multiple sequences of varying duration. We apply the model for learning, recognizing and encoding a set of human gestures. We measure systematically the performance of the model against noise.