Position/force control of robot manipulators without velocity/force measurements

Abstract

I n this aper, we consider the problem of designing a position/'rce controller for robot manipulators during constrained motion. The proposed controllers are based on exact knowledge of the system dynamics and do not require measurements o,f link velocity and end-effector ,force. The elimination of velocity measurements is accomplished by two dierent methods: a) a model-based velocity observer and (ii) a high-pass ltering technique. Both of the controllers also include an open-loop ,force control strategy to achieve semi-global exponential end-effector position/velocity/force tracking.1 Introduction Most of the previous work on the problem of position/force control of robots require feedback of link position, link velocity, and end-effector force. For example, in [7] an adaptive controller was presented which ensured asymptotic position tracking, but only bounded force tracking error (see 20 and 11 also , while recent work which yielded asymptotic position and force tracking. Since tachometer-based velocity measurements are usually very noisy and also add to the cost of the control system, two different methods have been proposed to eliminate velocity measurements with regard to the joint space position tracking control problem. The first method is a model-based velocity observer see [16], [2 , [14], [5 ). The second.method tion to generate a velocity related signal. This approach, which has the advantage that it can be used in robust and adaptive control schemes, can be thought of as an anal0 differentiation of the link position signal (see [22], 41, [3f lem, &U et a1 [SI developed an adaptive position/force controller which eliminated velocity measurements by the use of the filter of [4]. Due to the high cost of a force sensor, the design of position/force control schemes which only require position measurements has also been sought. For example in [9] , Huang et a1 formulated an asymptotic model-basedvariable structure observer for constrained robot systems to estimate velocity and contact force. Later, given that certain assumptions are satisfied, Huang et a1 [lo] used a nonlinear version of the separation principle to combine a velocity/force observer with a computed-torque like, position/force control strategy which produced asymptotic position/velocity/force tracking.In this paper, we design two exact model knowledge, position/force tracking controllers for robot manipulators during constrained motion which do not require measurement of 1ink.velocity and end-effector contact force. The controller strategy design is primarily based on the fact that the total degrees of freedom (denoted by n) for the position/force control problem can be partitioned into m osition control objectives and k force control objectives pi. e. , n = m + k ) . This decouplin is accomplished by employing the transformation of [I4 to develop n openloop equations which will be used to develop the position and force controllers separately (i.e., reduced order model). While the elimination of force measurements is achieved by the design ...

Globally exponential controller/observer for tracking in robots without velocity measurement

“…That extension was in fact made in [8,9] who presented a semi-globally exponential tracking observer/controller for robots with actuator dynamics (brush motor) and no velocity measurement nor current measurement. Reference [13] achieves the semi-globally exponential result for the hybrid control of robot position/force while observing link velocity. [11] also extended the approach of [Lim, et al] above to the robot with flexible joints; only link position and actuator position were assumed available; both unknown velocities were estimated, in a semiglobally exponential result.…”

“…Denote byt 1 the very first jump time after t = 0. Withq = +q d +s − e −aq,q, e, andq d all bounded on (0,t 1 ), this ensures that the gain k(t + 1 ), given in (22) (which satisfies (18)) is bounded (by direct inspection of the arguments of the right hand side of (22) while noting (13) and (14)). min (Q(0)) = 2 k v /2, and min (Q) ≥ 1 k v /2 on (0,t 1 ), we have no finite escape time, per V 0 + V 1 ≤ V 0 (0)+ V 1 (0), on this interval interior, (0,t 1 ), (and per the strictly stable error dynamics in (30) and (32)).…”

“…Additional surveying of output feedback control of flexible joint robots can be found in section 4.5 of [12]. Reference [13] achieves the semi-globally exponential result for the hybrid control of robot position/force while observing link velocity.…”

Adaptive output feedback tracking control of spacecraft formation

“…Using mean value theorem, as in Reference [22], we can bound w n as jjw n jj4r n ðjjg n jjÞjjg n jj ð60Þ…”

Adaptive output feedback tracking control of multiple spacecraft

“…Using mean value theorem, as in[5], we can bound Xn asIlXnII I Pn(II%II)II'YnII, where m(t) E W9 isa collection of error terms defined as yn A [ e : eTn q : ] ; pn is a positive nondecreasing function defined as Pn(ll'Yn1I) Ca + 6 Il'Ynll, and (a and c b are some positive terms. T -…”