Humanoid robots’ gait control strategy based on the Lie logic technique and LIPM model Prof. Carlos Balaguer University Carlos III of Madrid IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 http://roboticslab.uc3m.es IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Content 1. 2. 3. 4. 5. 6. 7. 8. 9. 2 Humanoid robots configuration Gait generation Lie logic fundamentals Application to Rh-1 humanoid robot Linear Inverted Pendulum Model (LIPM) 3D Walking pattern generation Simulations of Rh-1 humanoid robot Experiments on Rh-1 humanoid robot Conclusions IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Humanoid robots configuration (I) z Rh-1 humanoid robot developed by University Carlos III of Madrid – – – Full-size: ~1.500 mm ~50 kg including batteries 21 DOF z z z – – – 3 6 DOF each leg 3 DOF each arm 1 DOF head On board computers On-board sensors Wi-fi connection IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Humanoid robots configuration (II) z Rh-1’s cantilever type structure of hip joint: – – – – 4 YPR RPY - Cantilever Lower flexor torque of the robot’s hip → smaller actuator, less forces Lower position of the COM → more stability Wide sphere of leg motion Distribute the flexion wrench of the hip through the body robot → more difficult to control IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Gait generation (I) z Gait types: – – – – – z z 5 Forward motion Backward motion Lateral motion Rotation motion Climbing stairs motion There are infinite implementations of each gait’s type Combining gait types most of the movements must be possible IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Gait generation (II) z Static formulation – 6 During slow walking, COM remains always centered on the soles of the feet. z Dynamic formulation – During fast and smooth walking, COM is not always centered on the soles of the feet. IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Gait generation (III) z For good & fast gait generation are necessary: – Foot position at every moment must be transformed to joint position, i.e. space and time generation of the joints’ paths. z – Kinematics (direct and inverse) transformations at least of 12 DOF with different reference systems depends on the foot support. z 7 Joints’ paths are defined by pre-selected patrons that defines the form of walking (top model, sailor, etc.) Traditional methods, like Denavit-Hartenberg, are difficult to apply: no existence of close solutions and their computation is very time consuming. IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Lie logic fundamentals (I) z 8 CHASLES’ theorem: Every rigid body motion can be realized by a rotation about an axis combined with a translation parallel to that axis, this is a screw motion. The infinitesimal version of a screw motion is the Lie algebra se(3) – special Euclidian: TWIST ξ^. IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Lie logic fundamentals (II) z Screw theory advantages: – – – 9 It allows a global description without singularities due to the use of local coordinates (as Euler angles, Denavit-Hartenberg). It is possible to use only two coordinate frames, the base “S” and the tool “H” ones. Truly geometric description of rigid motion to make easer the kinematics analysis. A very natural and explicit description of the “Jacobian” which has not the drawbacks their local. The same mathematical treatment for the different robot joints: revolute and prismatic. IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Lie logic fundamentals (III) ? z z ξ ^θ How to solve e . p = q ? Using canonicals sub-problems – – – – z 10 Paden-Kahan one Paden-Kahan two Paden-Kahan three Pardos one All of them use exponentials instead matrices IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Lie logic fundamentals (IV) z Paden-Kahan one: rotation about a single axis e ξ ^θ .p = q ⎡ v ⎤ ⎡− w × r ⎤ ξ =⎢ ⎥=⎢ ⎥ w w ⎣ ⎦ ⎣ ⎦ v´= v − wwT v θ = a tan 2[wT (u´×v´), u´T .v´] 11 IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Lie logic fundamentals (V) z Paden-Kahan two: rotation about two ξ1^θ1 ξ 2 ^θ 2 ξ1^θ1 subsequent axis. e .e . p = e .c = q ⎡− w1 × r ⎤ ⎡− w2 × r ⎤ = ^ ξ ⎥ 2 ⎢ w ⎥ 2 ⎣ w1 ⎦ ⎣ ⎦ ξ1 = ⎢ (w w )w u − w v (w w )w v − w u α= (w w ) − 1;β = (w w ) − 1 T 1 2 T 1 T 2 T 1 T 1 2 2 2 T 1 T 1 2 T 2 2 u − α 2 − β 2 − 2αβ w1T w2 2 γ2 = w1 ×w2 2 c = r + αw1 + βw2 ± γ (w1 × w2 ) 12 IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Lie logic fundamentals (VI) z Paden-Kahan three: rotation to a given distance. ξ ^θ e .p − q = δ ⎡v⎤ ⎡− w × r ⎤ ξ =⎢ ⎥=⎢ ⎥ ⎣ w⎦ ⎣ w ⎦ u´= u − wwT u v´= v − wwT v 2 2 2 T δ ´ = δ − w ( p − q) θ 0 = a tan 2[wT (u´×v´), u´T .v´] 2 2 ⎛ + − δ ´2 ⎞⎟ u ´ v ´ −1 ⎜ θ = θ 0 ± cos ⎜ ⎟ 2 u´ v´ ⎝ ⎠ 13 IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Lie logic fundamentals (VII) z 14 Pardos one: traslation to a given distance. IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Lie logic fundamentals (VIII) POE (Product of exponentials) e e e 15 ξ1^θ1 .e ξ 3 ^θ 3 ξ 2 ^θ 2 ξ 2 ^θ 2 .e .e .p ξ 3 ^θ 3 ξ 3 ^θ 3 .p .p = q IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Lie logic fundamentals (IX) n g st (θ ) = ∏ e ξ i ^θ i ⋅ g (0 ) = e ξ1^θ1 ⋅e ξ 2 ^θ 2 ⋅e ξ 3 ^θ 3 ⋅e ξ 4 ^θ 4 ⋅e ξ 5 ^θ 5 ⋅e ξ 6 ^θ 6 ⋅ g st (0 ) i =1 z z z z 16 The above concepts (screw, twis, POE) could be used to solve direct and inverse kinematics of manipulator robot. Remember that Denavit-Hartemberg approach is based on products of matrices. Each robot DOF is an axis (ωi). A Matlab toolbox was implemented. IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Lie logic: Puma robot example (I) Definitions: ⎡0 ⎤ ⎡0 ⎤ ⎡1⎤ ⎡1⎤ ⎡0 ⎤ ⎡1⎤ w1 = ⎢⎢1⎥⎥; w2 = ⎢⎢0⎥⎥; w3 = ⎢⎢0⎥⎥; w4 = ⎢⎢1⎥⎥; w5 = ⎢⎢0⎥⎥; w6 = ⎢⎢0⎥⎥ ⎢⎣0⎥⎦ ⎢⎣0⎥⎦ ⎢⎣1⎥⎦ ⎢⎣0⎥⎦ ⎢⎣0⎥⎦ ⎢⎣0⎥⎦ ⎡v ⎤ ⎡ v3 ⎤ ⎡− w3 × r ⎤ ⎡ v2 ⎤ ⎡− w2 × k ⎤ ⎡− w × k ⎤ ; ξ ; ξ = = = ⎥ 2 ⎢w ⎥ ⎢ w ⎥ 3 ⎢w ⎥ = ⎢ w ⎥ 3 2 ⎦ ⎣ 2⎦ ⎣ ⎦ ⎦ ⎣ 3⎦ ⎣ ⎡v ⎤ ⎡ v5 ⎤ ⎡− w5 × p ⎤ ⎡ v6 ⎤ ⎡− w6 × p ⎤ ⎡− w4 × p ⎤ = = = ; ξ ; ξ 6 5 ⎢w ⎥ ⎢ w ⎥ ⎢w ⎥ = ⎢ w ⎥ w4 ⎥⎦ 6 5 ⎣ 5⎦ ⎣ ⎦ ⎣ 6⎦ ⎣ ⎦ ξ1 = ⎢ 1 ⎥ = ⎢ 1 ⎣ w1 ⎦ ⎣ w1 ξ4 = ⎢ 4 ⎥ = ⎢ ⎣ w4 ⎦ ⎣ 17 IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Lie logic: Puma robot example (II) g sh (θ ) = eξ1 ^θ1 .eξ 2 ^θ 2 .eξ3 ^θ 3 .eξ 4 ^θ 4 .eξ5 ^θ5 .eξ 6 ^θ 6 .g sh (0 ) Initial conditions: ⎡1 ⎢0 g sh (0 ) = ⎢ ⎢0 ⎢ ⎣0 18 ⎤ 1 0 p y − S y ⎥⎥ 0 1 H z − pz ⎥ ⎥ 0 0 1 ⎦ 0 0 0 IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Lie logic: Puma robot example (III) z Solving θ3 by Padan-Kahan three g sh (θ ) ⋅ g sh (0 ) ⋅ p − k = e −1 ξ1^θ1 ⋅⋅⋅ e ξ 6 ^θ 6 ⋅ p−k Twist properties : 1. Rotation over their own axis : e ξ ^θ ⋅ r = r → eliminates angles 1 and 2 2. Rotational screw conservs the norm : e ξ ^θ ⋅ p − r = p − r → eliminates angle 4, 5 and 6 ⎯ ⎯→ δ = e 19 ξ 3 ^θ 3 ⋅ p−k −K −3 ⎯P⎯ ⎯→ θ 3 Etc. IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Humanoid kinematics by Lie logic (I) z 20 Human body locomotion control IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Humanoid kinematics by Lie logic (II) z SKD (Sagital kinematics division) Boundary conditions: same position and orientation for the common parts (pelvis, thoracic, cervical) of the left and right humanoid. 21 Rh-1 IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Humanoid kinematics by Lie logic (III) Each manipulator is treated as an open kinematics chain separating leg and arm Right 12 DOF manipulator with the base in the right foot 22 Left 13 (12+1) DOF manipulator with the base in the left foot IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Humanoid kinematics by Lie logic (IV) z Direct kinematics g st (θ ) = e ξ1^θ1 ⋅e ξ 2 ^θ 2 Λe ξ12 ^θ12 ⎡υ ⎤ ⎡− ωi × qi ⎤ ξ =⎢ ⎥=⎢ ⎥ ω ω i ⎣ ⎦ ⎣ ⎦ ∧ i 23 ⋅ g st (0 ) IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Humanoid kinematics by Lie logic (V) z Inverse kinematics: crucial points t: is a point on the axis of the last DOF s: is a point not on the axis of the last physical DOF (hip) 24 p: is a common point for axes of the last three DOF (femur) q: is a common point for axes of the two first DOF (foot) IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Humanoid kinematics by Lie logic (VI) z e Inverse kinematics (I) −ξ 6 ^θ 6 ⋅⋅⋅ e −ξ1^θ1 ⋅ g st (θ ) ⋅ g st (0 ) ⋅ p − q = e −1 ⎯ ⎯→ δ = eξ9 θ 9 ⋅ p − q ^ e −ξ 6 ^θ 6 ⋅⋅⋅ e −ξ1^θ1 ξ 7 ^θ 7 ⋅⋅⋅ e ξ12 ^θ12 − K −3 ⎯P⎯ ⎯→ θ 9 ⋅ g st (θ ) ⋅ g st (0 ) ⋅ p = e −1 ξ 7 ^θ 7 ⋅⋅⋅ e ξ12 ^θ12 − K −2 ⎯ ⎯→ q ' = eξ 7 θ 7 ⋅ eξ8 θ8 ⋅ p ' ⎯P⎯ ⎯→ θ 7 , θ 8 ^ 25 ^ ⋅ p−q ⋅p IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Humanoid kinematics by Lie logic (VII) z e Inverse kinematics (II) −ξ 9 ^θ 9 ⋅⋅⋅ e −ξ1^θ1 ⎯ ⎯→ q ' = e e −ξ11^θ11 ⋅⋅⋅ e ξ10 ^θ10 −ξ1^θ1 ⎯ ⎯→ q ' = e 26 ⋅ g st (θ ) ⋅ g st (0 ) ⋅ t = e −1 ⋅e ξ11^θ11 ξ10 ^θ10 ⋅e − K −2 ⋅ p ' ⎯P⎯ ⎯→ θ10 , θ11 ⋅ g st (θ ) ⋅ g st (0 ) ⋅ s = e ξ12 ^θ12 ⋅e ξ11^θ11 −1 ξ12 ^θ12 − K −1 ⋅ p ' ⎯P⎯ ⎯→ θ12 ⋅s ξ12 ^θ12 ⋅t IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Lie logic vs Denavit-Hartenberg Y0 k H θ3 θ2 X0 Z0 θ1 Y X Z 27 S 0,4 ms 2,5 ms 0,2 ms 25 ms IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Kinematic (static) gait generation v2 g2 w f2 Foot in fly 28 w h2 g1 + 1 2 r2 v v 1 h1 u w f1 r1 v Marius Sophus Lie (1842-1899) IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Dynamic gate generation z Advantages of the dynamic gate generation: – – – 29 It is not necessary to pass exactly through the ZMP shadow during the walking The robot can mover faster and with more natural and smooth gait The inertial forces can be taken in account for other different applications than walking: seating, stairs climbing, etc. IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Distributed mass model z z 30 To generate the humanoid gait is necessary to take in account its dynamical model Classical Newton-Euler mass distributed model is extremely difficult to compute IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Distributed vs concentrated models z Advantages of the concentrated mass model: – – – z Advantages of the distributed mass model: – – 31 Less computation time with easer algorithm Easer control architecture and strategy Easy analytical model of the robot Exact model of the robot Easer prediction of the robots walking IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Concentrated mass model • Pendulum ball moves like a free ball in a plane following the inverted pendulum laws in the gravity field 32 IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Inverted Pendulum Model (I) z 2D model τ Fx = M .& x&= f . sin θ + . cos θ r τ Fz = M .& z&= f . cosθ − M .g − . sin θ r 33 IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Inverted Pendulum Model (II) z 34 3D model IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Inverted Pendulum Model (III) z Following the right hand rule, p = (x,y,z) is uniquely specified by a set of state variables q = (qr,qp,r) ⎛ ⎜ 0 ∂p ⎜ = ⎜ − rCr J= ∂q ⎜ − rCr S r ⎜ ⎝ D rC p 0 − rC p S p D S p = sin(θ p ) C p = cos(θ p ) S r = sin(θ r ) Cr = cos(θ r ) D = 1 − sin(θ r ) 2 − sin(θ p ) 2 35 ⎞ Sp ⎟ ⎟ − Sr ⎟ ⎟ D ⎟ ⎠ IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Inverted Pendulum Model (IV) x&⎞ ⎛& ⎜ ⎟ m⎜ & y&⎟ = J T ⎟ ⎜& & z ⎝ ⎠ ⎛τ r ⎞ ⎛ 0 ⎞ ⎟ ⎟ ⎜ −1 ⎜ .⎜τ p ⎟ + ⎜ 0 ⎟ ⎜ f ⎟ ⎜ − mg ⎟ ⎠ ⎝ ⎠ ⎝ ( ) ⎛ ⎜ 0 ⎜ m⎜ rC p ⎜ ⎜ Sp ⎜ ⎝ 36 − rCr 0 − Sr − rCr S r ⎞ ⎛ − rCr S r ⎞ ⎟ ⎜ ⎟ D ⎟⎛ & D x&⎞ ⎛ τ ⎞ ⎜ ⎟ − rC p S p ⎟⎜ ⎟ ⎜ r ⎟ rC S − p p ⎟ y&⎟ = ⎜τ p ⎟ − mg ⎜ ⎜& ⎟ ⎜ D D ⎟ ⎜ ⎟ ⎜ ⎟ ⎜ D ⎟ z&⎠ ⎝ f ⎠ D ⎟⎝ & ⎟ ⎜ ⎟ ⎠ ⎝ ⎠ ⎛ D⎞ ⎟τ p + mgx m( z & x&− x& z&) = ⎜ ⎜C ⎟ ⎝ p⎠ ⎛D m(− z & y&+ y & z&) = ⎜⎜ ⎝ Cr ⎞ ⎟⎟τ r − mgy ⎠ IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Inverted Pendulum Model (V) z Motion constraints: – – – 37 Limits the motion in a plane with given normal vector (kx, ky, -1) and z intersection zc, The normal vector should match the slope of the ground and the z intersection should be the expected average distance of the center of the robot’s mass from the ground, The motion is constraints to the sagital plane g 1 D g & & x= x & τp x&= x + zc zc mzc C p 1 D g & τr y&= y − zc mzc Cr & y&= g y zc IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Linear Inverted Pendulum Model (LIMP) With initial conditions “i” at time ti, the mass concentrated trajectory is: t − ti t − ti ) + Tc x&i sinh( ) Tc Tc x t − ti t − ti x&(t ) = i sinh( ) + x&i cosh( ) Tc Tc Tc x(t ) = xi cosh( y (t ) = yi cosh( y&(t ) = 38 t − ti t − ti ) + Tc y&i sinh( ) Tc Tc yi t − ti t − ti sinh( ) + y&i cosh( ) Tc Tc Tc IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Gait generation by LIPM (I) x Stopping phase x y x Starting phase y 39 y IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Gait generation by LIPM (II) Support foot (left) Support foot (left) 40 Support foot (right) IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Gait generation by LIPM (III) 41 x temporal trajectories (position-blue line, velocity-red line) IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Gait generation by LIPM (IV) y temporal trajectory (position-blue line, velocity-red line) 42 IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Gait generation by LIPM (V) Foot trajectories are computed by single splines taking in account some constraints as, • step length, • max height of foot, • lateral foot motion, • foot orientation and • speed 43 in order to avoid fall down and reduce the impact force on landing foot IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Gait generation by LIPM (VI) Any direction walking patterns could been developed using rotation matrix around z-axis of local frame Pattern Connection 44 IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Simulations of Rh-1 humanoid robot • MATLAB based simulator • VRML toolbox • Lie logic implementation • LIPM model 45 IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Gait generation architecture 46 IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Gait correction procedure (I) z 47 Gait compensation (correction) due the mechanical flexion (flexor torques) IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Gait correction procedure (II) without correction 48 with correction IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Experiments on Rh-1 humanoid robot (I) 49 IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Experiments on Rh-1 humanoid robot (II) 50 IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Experiments on Rh-1 humanoid robot (III) 2005 (Static gait) Lp=130 mm, Tp=20 seg (0.02Km/h) 51 2006 (Dynamic gait) Lp=180 mm, Tp=1.25 seg (0.52Km/h) About 20 times faster! IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Experiments on Rh-1 humanoid robot (IV) Hip and Foot spatial trajectories COM trajectory 0 -1 -2 -3 -4 Right foot trajectory -5 -6 -7 -8 1 1.5 0 1 -1 0.5 -2 0 -0.5 -3 -1 -4 52 -1.5 Spatial hip and feet trajectories without correction (black line), with correction (red line) IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 Conclusions z z z z z 53 Locomotion of the humanoid robots is a complex and, in general, not solved problem The small amount of computation for LIPM method, allow us dynamic walking control in real time on actual humanoid robots. Dynamic walking was successfully implemented in Rh-1 humanoid robot using the Lie logic and LIPM model The implemented dynamic gait is smooth and natural, and about twenty times faster than static one It have been demonstrated that any direction patterns could been generated IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 References (I) [1] M. Arbulú, J.M. Pardos, L.M. Cabas, P. Staroverov, D. Kaynov, C. Pérez, M.A. Rodríguez; C. Balaguer, “Rh-0 humanoid full size robot`s control strategy based on the Lie logic technique”, IEEE-RAS International Conference on Humanoid Robots (Humanoids'2005). Tsukuba. Japan. Dec, 2005 [2] M. Arbulú, I. Prieto, D. Gutiérrez, L. Cabas, P. Staroverov, C. Balaguer, “User friendly graphical environment for gait optimization of the humanoid robot Rh-0”, 7tn International Conference on Climbing and Walking Robots (Clawar'2004). Madrid. Spain. Sep, 2004 [3] M. Arbulú, F. Prieto, L. Cabas, P. Staroverov, D. Kaynov, C. Balaguer, “ZMP Human Measure System”, 8th International Conference on Climbing and Walking Robots (Clawar'2005). London. United Kingdom. Sep, 2005. [4] K. Hirai, M. Hirose, Y. Hikawa and T. Takanaka, “The development of Honda humanoid robot”, IEEE International Conference on Robotics and Automation (ICRA 1998) Leuven (Belgium) [5] J. Yamaguchi, E. Soga, S. Inoue A. and Takanishi, “Development of a bipedal humanoid robot control method of whole body cooperative dynamic bipedal walking”, IEEE International Conference on Robotics and Automation (ICRA’ 1999), Detroit, (USA). [6] S. Kajita, F. Kaneiro, K. Kaneko, K. Fujiwara, K. Yokoi and H. Hirukawa, “Biped walking pattern generation by a simple 3D inverted pendulum model”, Autonomous Robots, vol 17, nª2, 2003 [7] K. Löeffler, M. Gienger, F. Pfeiffer and H. Ulbrich, “Sensors and Control Concept of a Biped Robot”, IEEE Industrial Transactions on Industrial Electronics, Vol. 51, Nº 5, October 2004 54 IURS 2006 Robotics Summer School on Humanoid Robots Benicassim, 18-22/9/2006 References (II) [8] Y. Choi, B. You and S. Oh, “On the Stability of Indirect ZMP Controller for Biped Robot Systems”, Proceedings of 2004 IEEE/RSJ International Conference on Intelligent Robots and Systems, Sept. 2004, Sendai, Japan [9] M.H. Raibert, Legged robots that balance, MIT Press:Cambridge, 1986 [10]A.-J. Baerveldt, R. Klang. “A low cost and Low-weight Attitude Estimatin System for an Autonomous Helicopter”. Proc. of IEEE International Conference on Intelligent Engineering Systems, Budapest, Hungary, 391-391, 1997. [11]Q.Huang; K.Kaneko; K.Yokoi; S.Kajita; T.Kotoku; N.Koyachi; H.Arai; N.Imamura; K.Komoriya; K.Tanie. “Balance Control of a Biped Robot Combining Off-line Pattern with Real-time Modification”. Proc. Of IEEE International conference on Robotics and Automation. San Francisco, USA. April, 2000. 55