Vertebrate- or snake-like soft robot based on tensegrity principle Présentation GT5, vendredi 28 novembre 2014 Alex Pitti, phd Maître de Conférence, chaire d'excellence UCP-CNRS Laboratoire ETIS CNRS, ENSEA, Cergy-Pontoise University
Research aims Biologically-inspired solutions to challenge control of dexterous robots - dimensionality (high number of DoF) - exploiting the physics of the material (elasticity, friction) and of the robot morphology - large repertoire of behaviors (walking, breaking, jumping, postural control) Some solutions I attempt to propose: Mechanism of phase synchronization in dynamical systems the controller - control the system's global dynamics - tuning to the material property and system morphology - applied to high dimensional system The body design to process morphological computations - geometry, structure the robot - material properties based on previous works [Pitti, 2005~]
The robot : physical embodiment 1, material The body's material properties are soft (muscles tissues and bones), actual robot are engineered with hard materials (steel, plastic) Hard materials are Hookean (mostly linear) Soft tissues are non-hookean (visco-elastic) '' always in tension (pre-stressed, muscle tone) [Tulving, 2005] [Gordon : «Structures or why things don't fall down»]
physical embodiment 2, the structure The body structure (morphology) plays a role of a function in behavior morphological computations, [R. Pfeifer] Actual robot designer starts to have these considerations in mind. [Gordon, 1994] [Gordon : «Structures or why things don't fall down»] [Lipson, 2004]
Changing the paradigm We may see the musculo-skeletal system as a network of tension links (muscles, tendons) connected to compression structures (bones) : Pre-stressed structure. Pretty much-like tensile structures.
Tensegrity structure Tensegrity = integrity of tension proposed by B. Fuller & Sneil network of tension and compression structure. [Fuller and Sneil]
Tensegrity structure Tensegrity = integrity of tension proposed by B. Fuller & Sneil network of tension and compression structure. No momentum (no tangential force Newton laws & actual robots) Ecological distribution of forces on all the structure: less power consumption to move Exploit fully the physic of the structure: Self-replicative with lots of redundancy Light-weight structure More robust & solid to defects/shocks Self-Balance and neutral posture: return back to its own stable equilibrium for arbitrarily small perturbations [Tulvey 2013]
Tensegrity structure in biology Tensile links (muscles) support the structure weights (bones), not the reverse [Flemons, 2006] Shun Izawaya [Pfeifer Pitti 2012]
Tensegrity robots design principles: Muscles redundancy for compliance Morphological computations: structure = function Weak and loosely distributed units [Riesel 2012] Nasa satellite antenna [SunSpiral, 2012] [Shibata 2009]
Prototypes done [2011~] Joint link device Snake-like or trunk-like tensile robot
Current snake prototype Snake-like or trunk-like tensile robot
Anguilliform models Snake skeleton Representation of the swimmer as a chain of interconnected links. [McMillen 2008]
A proposed model of CPGs for multi-dofs Model-free mechanism For some specific coupling, Chaotic systems will match the system dynamics 3 Synchronization 2 Feedback Unknown dissipative system Chaotic system? 4 Resonance 1 Perturbation γ is a global parameter to control synchronization (motor synergy) State of feedback resonance in the dissipative system [Fradkov, 1999] [Pitti, 2005-2011]
Simulations of multi-dof robots Ring-like mass-spring damper system (30 DoF) [Pitti, 2005] Fréquence (Hz) chaotic controlers Control Parameter Dog-like system (2 DoF) [Pitti, 2006] Frog-like systems [Niiyama, Pitti, 2009]
Snake-like robot Model with 10 servos and 200ms delay sinusoidal oscillators with 5 segments [done with Julien Abadji] top view side view
Snake-like robot, current version Model with solenoids (electro-magnets) and chaotic controllers (logistic map) electro-magnets body chaotic controller
Morphological computation own facial somatopic information Audio spectral filtering done by shape of the ear 3D printed Ear Micro Freq [Hz] can serve for facial processing [Pitti, 2013] Time [s] [Pitti, 2012]
to conclude thinks How to make a robot that perceives like humans? moves Rolf Pfeifer, ETH Zurich Understanding human intelligence(s) by synthetic or constructive approaches. - To bridge the level of explanations between Engineering, Biomechanics Robotics/Comp. Science/Info. Theory Developmental Psychology Cognitive Neuroscience - Seeking for design principles - Reproducing it with robots Rolf Pfeifer and Alex Pitti Manuella Editions, 2012
Save the date: Journée GT8 Robotique et Neuroscience 17 décembre, UPMC Cognitive Robotics and Enactive Systems co-organizers Benoît Girard, Medhi Kamassi, Ghilès Mostafaoui, Alex Pitti Thank you Architect project JB Mouret, 2014