Professional Documents
Culture Documents
Abstract - Carrying out an autonomous object prehension 11. THE TACTILE SENSOR
task, needs some active interaction with the environment,
which only could be done with sensory integration on the The tactile sensing system developed in the ISR-
robot tools. A study of the contact force control performed Coimbra Pole (Institute of Systems and Robotics) [5,6],is
with tactile feedback in a low cost servo-controlled parallel based on a piezoresistive thick-film polymer manufactured
jaw gripper is presented in this article. This gripper by Interlinkm. The distributed pressure is sensed by a
integrates matricial tactile sensors and proximity sensors.
16x16 matrix of force sensing resistors (FSR) pressure
Considering the nonlinear nature of the control system,
hybrid control strategies were used, inter-relating classical transducer (tactile elements or taxels), providing a spatial
digital control strategies with heuristic rules and logic. Two resolution of 1.54 mm with a sensing area of 6.25 mm2.
control algorithms were studied, and experimentally The FSR elements are polymer film gauges that exhibit
validated in a force-compliant mechanical system: hybrid decreasing electrical resistance with applied normal force.
proportional controller; hybrid proportional integral and The acquisition module combines an analog-digital
derivative controller. scanning circuit and a 16 bits microcontroller (Intel
80C196KC) with an analog-to-digital converter, as shown
I. INTRODUCTION in Fig. 2. The scanning array technique implements a
taxel-isolation technique that consists in a voltage
Object grasp stability depends on two factors: 1)- an feedback of the selected taxel into all other not selected
appropriate prehension force; 2)- correct contact points rows and columns taxels [ 5 ] , with a maximum sampling
location. Usually, the contact force control is in the third rate for the entire matrix of 125 Hz. This module
phase of the grasping process, the postcontact phase, when dimensions are small enough to put them on a robot arm
there is a well defined contact between the robot tool and near the wrist.
the object. In this phase the goal is to get contact stability, The two-layered design of the FSR transducer can give
which means the ability for the end-effector to maintain rise to degradation of the sensor response if the two halves
contact with an object without slipping even when the are allowed to slide one relatively to the other. To cope
object is subjected to disturbing external forces, with this problem we placed a 10 mils thick, low
Several researchers have been studying the force control resistance, sheet of anisotropic elastomeric conductive
in robotic mechanisms namely Mason [l], Mason and polymer interconnect (manufactured by AT&T) between
Salisbury [ 2 ] and Nakamura et al. [3]. Contact force the transducer halves likely Speeter [7]. Covering the FSR
control deals namely the transition problem from position sheet there is a compliant rubber, too.
control (precontact) to force control (postcontact). Recent
work has shown that successful transitions from free 111. CONTACT FORCES MODEL
motion to constrained motion can possibly be addressed by
gain scheduling algorithms, and requires the control of the Only the contact normal force distribution is sensed by
contact forces as soon as contact is made [4]. the tactile sensor. Associating this characteristic, to the
This paper addresses a new approach to tactile force
control provided by the analysis of matricial tactile
information.
A. Article Organisation
sensor planar geometry, gives a parallel forces system with (less flexible) and the other as Spring I1 (more flexible).
points of application in the sensors plane. These two
conditions leads to the possibility of force centre (centroid) B. Contact Force Control Methodologies
computation and related resultant force.
Let {S’) denote the coordinates system of the tactile Usually all types of motion transmission systems have
sensor in the left hand finger. The centroid c’ described in several nonlinearities. To deal with these nonlinearities,
{ S’ } is given by a point with coordinates: some researchers have been developing new control
strategies, based on nonlinear programming, hybrid
controllers and heuristics [2].Linear optimal control was
used by Akella and Cutkosky [4], considering negligible
nonlinearities, to deal with contact forces.
R’
Considering the nonlinear nature of the control system,
hybrid control strategies were used in this work, inter-
relating classical digital control strategies with heuristic
R‘ rules and logic. Two control algorithms based on a
modified digital PID controller were studied, and
where R’ represents the resultant force, F’ij represents the experimentally validated in a force-compliant mechanical
contact force in each tactile element (ij) and (X’i, z’j) are system: a hybrid proportional controller and a hybrid
the coordinates of each tactile element (i,j). proportional integral and derivative controller.
The resultant force in each matricial subsystem is
perpendicular to the sen:sors plane XY and is the result of C. Mod$ed Digital PID Controller
the algebraic sum of all contact forces in each subsystem.
The resultant force R’ described in ( S ’ } is given by the Some modifications to the standard discrete PID control
following expression: algorithm have been proposed to improve its performance
(see Fig. 4) [8]. In order to reduce large output changes
16 16
R’= CC F‘, 2’ caused by rapid changes in the reference signal, the
reference variable r(k) is not included in the derivative
;=I j L 1
term (B-Block, in Fig. 4). All the integral parameters are
where z‘denotes the unit vector along 2’-axis. Similarly, only present in the integration term (A-Block, in Fig. 4).
we can obtain the centroid c” relative to the right hand This modified algorithm is less sensitive to higher
finger. frequencies of r(k) than to those of fr(k). The controller gi
parameters, are:
IV.CONTACT FORCE CONTROL Kp.TD
A. Experimental Setup
go = - K , - 7 (3)
hj
' -0
0 10 20 30 40 50 , 70 80 90 100 0 IO 20 30 40 50 60 70 80 90 100
Fig. 7. Proportional step response to a test reference value r(k)=3000, Fig. 8. Digital control values u(k) from a proportional controller without
without actuator management. The proportional gains are Kp=l (-) and actuator management. The reference value is r(k)=3000. The proportional
Kp=0.05 (---). gains ace Kp=l L) and Kp=0.05 (-).
-400 I I
0 10 20 30 40 50 60 70 80 YO 100
Fig. 9. Proportional step reslponse to a test reference value r(k)-3000, Fig. 10. Digital control values u,(k) (---I and ufk) (-), from a
with actuator management. The proportional gain is Kp=0.05. proportianal controller with actuator management. The reference value is
The actuator management uses a 98% criterion to the final value. r(k)=3000. The proportional gain is Kp=0.05.
0 10 20 30
-40 50 60 70 80 90 100
.3"
0 10 20 30 40 50 60 70 80 YO
J
100
EL SAMPLES (Ts=32ms) N.SAMPLES (Ts=32ms)
Fig. 11. PID step response with trapezoidal integration (-) and Fig. 12. Digital control values u,(kj, for trapezoidal integration (-) and
rectangular integration (...), ita a reference value r(k)=3000. The initial rectangular integration (...), to the test conditions of Figure 11.
distance between the fingertip and the compliant test system is 3 mm.