文献
trol of under-actuated dynamic systems.
4. Input-Output Linearbation and Jacobian Matrices According t o the extended dynamic model, the upperactuated and the lower-actuated system equations can be written, respectively, as follows:
2. Extended Dynamic Model Under-actuated dynamic systems consist of active (actuated) and passive (unactuated) joints. The passive joints in such dynamic systems can either be physically real or hyperthetic. For instance, the inverted-pendulum (cart-pole) system has one passive joint which is a real free-rotatible one. While for the space robot, if we model its passive floating base as a virtual six-joint robot arm, then all these six passive joints are hyperthetic, but they uniquely represent the floating base motion with six degrees of freedom[2]. We refer t o the entire system defined to include all actuated (active) and
unactuated (passive) joints, no matter whether they are real or virtual, as an extended dynamic model. Thus, in such an extended model, the number of input channels is equal t o the number of active joints, while the dimension of the system state space is twice of the total number of both active and passive joints.
wq+ci-r9=
3. Inertial Matrices With the extended dynamic model, the n by n inertial matrix W for an under-actuated dynamic system can be partitioned into four blocks,
where W11 is an 1 by 1 symmetric submatrix attributed by the lower body, Wzz is an m by m symmetric block as an inertial matrix of the upper body with respect to the fixed base, and WIZ= W z is an 1 by m submatrix representing the interaction between the lower and upper bodies of the under-actuated system. In other words, there are 1 joints in the lower body and m joints in the upper body of the system, and 1+ m= n. We can now write the inverse of the inertial matrix W[3] for an upper-actuated system by
where uUp E 3" and uio E IR' are the actuated joint torque 2 inputs for the upper-actuated and lower-actuated systems, respectively. Clearly, the location of the zero vector 0 on the right-hand side of each equation is the main distinction between the two cases. To apply the input-output linearization procedure[4,5] into the system equations (6) and (7), we need to define an effective Jacobian matrix for kinematic representations of both the upper- and lower-actuated systems. This can also be developed based on the extended model. Suppose an m-dimensional Cartesian displacement of the system endeffector with respect to the fixed base is chosen as an output vector which is a differentiable function of the joint position q= (q: qTp)T, and is denoted by y= h(q) E 2 . The" Jacobian matrix of y is determined by
where JI= dh/aql, is of m by 1, 5 2= ah/aq,, is of m by m, and qio E%' and qup E IR" are the joint position vectors for the lower body and the upper body of the system, respectively. Similar to the effective inertial matrix definition, we define an effective Jacobian matrix by
j= JT12= J2 - J WG1 Wiz? 1 for the upper-actuated system, and define
(9)
where
(2)
ji= JTzi= J1 - 52 WE1Wzl
(10)
w z z= w z z - wz w W n . 1; (3) It can be seen that W is invertible if both Wii and W~ are Z
for the lower-actuated system, in which the output y= h(q) should be 1-dimensional.
nonsingular. Wl1, as an upper-left block of (I), is positivedefinite and symmetric, and thus invertible. Whereas for the matrix W 2 2 in (3), which is referred to as an effective inertial matrix of the upper-actuated system, we can show that it is also a positive-definite symmetric matrix. Similarly, the inverse of W for a lower-actuated system is as follows:
5. A Direct Adaptive Control Scheme We now show that if pio, qlo and i l o are exactly measurable for an upper-actuated system, then a direct adaptive control scheme can asymptotically stabilize the entire system for accur
ate trajectory-tracking against parameter uncertainty. Let an augmented output error function between the desired (ya)d
=
(E) (;) .
and the actual y,
=
(lo)
be
(4)
e,= ( y, ) d - ya= defined by
Let an extended augmented error be
where
= Wll - w12w~1w21 (5) is called the effective inertial matrix of the lower-actuated system, which is also positive-definite and symmetric. fill
2
1628

Bubbles Adaptive routing... 10页 免费... A dynamic decoupling control scheme for high-...(b) Mixture of direct and cross-coupling PI ...
Direct adaptive control for a class of MIMO nonlinear discrete-time systems_电子/电路_工程科技_专业资料。Journal of Systems Engineering and Electronics Vo1...
Longitudinal Control S... 11人阅读 6页 2.00 Discrete-time Direct M... 5人阅读 5页 2.00 Robust direct adaptive... 2人阅读 8页 2....
Robust adaptive fuzzy control scheme for nonlinear system with uncertainty_专业资料。维普资讯 http://www.cqvip.com JunlCotohoyadApiain (062921 ora nrlTernplts...
Novel control scheme for 3-phase PWM current-source converters under ... Dynamic modeling and d... 20人阅读 10页 2.00 AN ADAPTIVE OPPORTU....
(disturbance) dynamic model are unknown and/or change in time Particular ... Lozano, M’Saad, Karimi How to get a Direct Adaptive Control scheme ? ...
Robust direct adaptive fuzzy control for nonlinear MIMO systems_专业资料。维普资讯 http://www.cqvip.com POGRESl AURLSECERS NTA CIN N V0.116。No0.1....
Direct model reference adaptive control using Kp=LDU factorization for ... A twin-electrode alter... 5人阅读 5页 2.00 Dynamic modeling and ...
Discrete-time Direct Model Reference Adaptive Control: A Systematic Approach_专业资料。维普资讯 http://www.cqvip.com Vb.313 No1 .0 ACTAAUT0MATI CA ICA...
adaptive scheme for global optimization over_信息与...At the heart of every direct search method is ... a control variable for the DE1-scheme. ...
热门文档