IGM - Gestion

Institute

Related publications (83)

Complexity and Similarity Metrics for Unsupervised Identification of Programming Methods for Robot Cooking Tasks

Josephine Anna Eleanor Hughes, Stefan Ilic

Springer2022

Frequency-domain data-driven position-dependent controller synthesis for Cartesian Robots

Alireza Karimi, Philippe Louis Schuchert

Cartesian robots have position-dependent dynamics that should be accounted for in high performance applications. Traditional methods design Linear Time Invariant (LTI) controllers which are robustly stable with respect to position variations, but require a trade-off in performance to account for the changing dynamics. Advanced methods require Linear Parameter Varying (LPV) models and LPV controller design methods that are not well established in industry. On the other hand, classical model-based gain-scheduled technique requires paramet- ric identification, design of high performance controllers for each position, interpolation of the controller parameters and real-time validation of the gain-scheduled controller, which takes costly engineering time. We propose a new approach, using the frequency response of a system at different operating points, to design a Linear Parameter Varying (LPV) controller. The controller parameters are optimised by a convex optimisation algorithm based on second-order cone programming. The approach is applied to an industrial 3-axis Cartesian robot, showing significant improvements over state-of-the-art control design strategies. Data acquisition and controller design can be performed automatically, reducing significantly the engineering costs for controller synthesis

2022

A convex set of robust D—stabilizing controllers using Cauchy's argument principle

Alireza Karimi, Philippe Louis Schuchert

A new approach is presented to obtain a convex set of robust D—stabilizing fixed structure controllers, relying on Cauchy's argument principle. A convex set of D—stabilizing controllers around an initial D—stabilizing controller for a multi-model set is represented by an infinite set of Linear Matrix Inequalities (LMIs). By appropriate sampling of the D—stability boundary, a Semi-Definite Programming (SDP) is proposed that can be integrated in other synthesis approaches to ensure D—stability along other design specifications. To showcase utility of the proposed approach, two different examples are given: a boost converter with multi-model uncertainty and a laser-beam system modeled by an identified finite impulse response.

Elsevier2022

Data-driven Power Electronic Converter Control Design in Power System Applications

Seyed Sohail Madani

In recent years, power systems have evolved in physical and cyber-physical layers. In the physical layer, the changes are motivated by environmental concerns resulting in the integration of new types of generation/demand/storage into the grid. These integrations offer the opportunity of forming self-sufficient microgrids. Based on the hierarchical microgrid control structure, it should provide a high-performance low-level control while damping the high-frequency oscillations. Moreover, it should share power among DGs at the primary level and restore voltage/frequency at the secondary level. However, achieving these objectives is challenging because of factors such as lack of inertia, system complexity, different line characteristics, high-frequency modes of switching converters and their filters, delay, and incomplete communication graph. With the advent of smart grid technologies, the cyber-physical layer of the grid has transformed, and more measurement data are available. These data are mostly used for monitoring, metering, and protection. This thesis tries to bridge the gap between measurement data and high-performance control design for microgrids to address their operational challenges. The scope of this thesis covers up to the secondary control level. At the lowest level, the passivity theory has been used to tackle the problem of high-frequency oscillation between converters in a decentralized way. To this end, conditions of data-driven robust passivity for a performance channel are proposed. This method is used to make the input admittance of an AC grid-connected converter passive while the tracking performance is optimized. To validate the performance of designed controllers with high fidelity, an experimental Hardware-in-the-Loop (HIL) setup is developed and extended to a Power-HIL (PHIL) setup. However, PHIL tests have a performance limitation due to their stabilization. A data-driven method is proposed to optimize their performance while robust stability is guaranteed. This PHIL setup is employed for the validation of the passivity-based converter control. In addition, inspired by railway system standards, conditions for data-driven partial positive realness over an arbitrary frequency set are developed and used for the traction converter control design. Moreover, the primary/secondary microgrid control problem is formulated as a comprehensive data-driven multivariable synthesis problem. In this method, active power-sharing and frequency/voltage restoration are optimized while the closed-loop stability with a predefined margin is guaranteed. Due to the fixed structure, the controller can be designed based on the available communication while considering its delay. Since there is no need for time-scale separation between primary and secondary, this method leads to better performance. Moreover, because of data-driven property and multivariable structure, there is no need for decoupling or any assumption on the grid impedance. This design is extended to reactive power sharing using the spare capacity of PhotoVoltaics (PVs). In addition, a data-driven Linear Parameter Varying (LPV) multivariable synthesis method is proposed and used for the microgrid control to extend the applicability in different operating points. All the proposed methods are in a data-driven framework, which is suitable for complex power system applications. As well, the design problems are in convex programming form, which can be solved efficiently.

EPFL2022

SPRINGER2022

Many aquatic animals propel themselves by flapping their tails. Leveraging a recently proposed snapping cantilever beam based on the concept of flexural tensegrity, we propose a bio-inspired propulsion device. The design comprises a segmental beam with hollow voussoirs in unilateral contact along tailor-shaped surfaces, held together by a prestressed internal cable. Prescribing relative periodic rotation to a pair of consecutive control segments of the structured beam produces multi-articulated sequential snapping of all joints. We built a series of prototypes and performed precision experimental tests in water to characterize their propulsive capacity. A parametric study was carried out to characterize the dependence of the thrust produced by the oscillating tail on the following factors: number of segments, shape of the internal cavities dictating the cantilever curvature, and mobility constraint of a fin appended at the end of the tail. The results provide a proof of concept that our design for a snapping structured beam can be used as a propulsive device. We further demonstrate the feasibility of this propulsion unit to propel a toy boat in a water basin. (C) 2022 Elsevier Ltd. All rights reserved.

ELSEVIER2022

Experimental data-driven model predictive control of a hospital HVAC system during regular use

Colin Neil Jones, Emilio Maddalena, Christophe Salzmann

Herein we report a multi-zone, heating, ventilation and air-conditioning (HVAC) control case study of an industrial plant responsible for cooling a hospital surgery center. The adopted approach to guaranteeing thermal comfort and reducing electrical energy consumption is based on a statistical non-parametric, non-linear regression technique named Gaussian processes. Our study aimed at assessing the suitability of the aforementioned technique to learning the building dynamics and yielding models for our model predictive control (MPC) scheme. Experimental results gathered while the building was under regular use showcase the final controller performance while subject to a number of measured and unmeasured disturbances. Finally, we provide readers with practical details and recommendations on how to manage the computational complexity of the on-line optimization problem and obtain high-quality solutions from solvers. (c) 2022 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

ELSEVIER SCIENCE SA2022

Arm-wrist haptic sleeve for drone teleoperation

Dario Floreano, Matteo Macchini, Vivek Ramachandran

Teleoperators rely on both visual and haptic feedback to perform drone teleoperation tasks, such as obstacle avoidance. Haptic feedback becomes essential when visual feedback is compromised, either due to visual occlusions or poor depth perception. However, haptic interfaces, are often bulky because they require heavy actuators to provide force feedback. The bulkiness reduces user mobility and makes these interfaces unsuitable for prolonged use. Here, we propose a wearable haptic sleeve that encompasses the wrist and elbow joints of a human arm. The two joint rotations control the motion of a drone in a simulated environment along a horizontal plane. The sleeve is composed of modular electroadhesive clutches that block the joint movement when the drone is in the vicinity of an obstacle. The clutches are lightweight (27 g), require low power (~1 mW) to operate, and can be mounted on the user without affecting the user's mobility. A motor learning subject study is conducted to navigate a drone through a hole in a wall where the depth perception of visual feedback is compromised. The results of the study show that subjects trained with the wearable haptic sleeve learnt the drone obstacle avoidance task and retained the necessary motor skills after haptic training, compared to subjects who received only visual feedback and were unable to learn the motor task.

2021

Stories from the future of intelligent machines—from rescue drones to robot spouses—and accounts of cutting-edge research that could make it all possible. Tech prognosticators promised us robots—autonomous humanoids that could carry out any number of tasks. Instead, we have robot vacuum cleaners. But, as Dario Floreano and Nicola Nosengo report, advances in robotics could bring those rosy predictions closer to reality. A new generation of robots, directly inspired by the intelligence and bodies of living organisms, will be able not only to process data but to interact physically with humans and the environment. In this book, Floreano, a roboticist, and Nosengo, a science writer, bring us tales from the future of intelligent machines—from rescue drones to robot spouses—along with accounts of the cutting-edge research that could make it all possible. These stories from the not-so-distant future show us robots that can be used for mitigating effects of climate change, providing healthcare, working with humans on the factory floor, and more. Floreano and Nosengo tell us how an application of swarm robotics could protect Venice from flooding, how drones could reduce traffic on the congested streets of mega-cities like Hong Kong, and how a “long-term relationship model” robot could supply sex, love, and companionship. After each fictional scenario, they explain the technologies that underlie it, describing advances in such areas as soft robotics, swarm robotics, aerial and mobile robotics, humanoid robots, wearable robots, and even biohybrid robots based on living cells. Robotics technology is no silver bullet for all the world's problems—but it can help us tackle some of the most pressing challenges we face.

MIT Press2022

Bio-inspired Reflex System for Learning Visual Information for Resilient Robotic Manipulation

Josephine Anna Eleanor Hughes, Kai Christian Junge

Humans have an incredible sense of self-preservation that is both instilled, and also learned through experience. One system which contributes to this is the pain and reflex system which both minimizes damage through involuntary reflex ac- tions and also serves as a means of ‘negative reinforcement’ to allow learning of poor actions or decision. Equipping robots with a reflex system and parallel learning architecture could help to prolong their useful life and allow for continued learning of safe actions. Focusing on a specific mock-up scenario of cubes on a ‘stove’ like setup, we investigate the hardware and learning approaches for a robotic manipulator to learn the presence of ‘hot’ objects and its contextual relationship to the environment. By creating a reflex arc using analog electronics that bypasses the ‘brain’ of the system we show an increase in the speed of release by at least two-fold. In parallel we have a learning procedure which combines visual information of the scene with this ‘pain signal’ to learn and predict when an object may be hot, utilizing an object detection neural network. Finally, we are able to extract the learned contextual information of the environment by introducing a method inspired by ‘thought experiments’ to generate heatmaps that indicate the probability of the environment being hot.

2022