ABSTRACT: This paper describes work un-welding workcell。 This is a complex system derway  to evaluate the  effectiveness of voiceinvolx'ing=  inputs  from  multiple  sensors  and recognition  systems  as  an  element  in  thecontrol of a wide variety' of robot   manipu- control of a robotic welding workcell。   Fac-lator motions and  process  s'ariables。 While tors  being  considered  for  control  includemany functions are automated。 a human op- program editor access security, preoperationeiator sew'es in a supervisor' capacity。 ready' checklist    requirements,     xveldinp=    processto ox’erride  functions v•'hen  necessap'。  In the x'ariable control。 and robot manipulator mo-present investigation。 a commercially' as'ail- tion  overrides。  In the  latter tn'o categories,able s'oice recognition system is being  inte- manual  x'ocal  control  is  being  comparedgrated  svith a robotic  n'elding  workcell  at against manual tactile control and fully’  au-NASA Marshall Space Flight Center。 w’hich tomatic  control  in  terms  of  speed  of   re-is used as a test bed  for evaluation and   de- sponse。  accuracy,  stability,  reliability。 ands'elopment of advanced technologies for use safety。in fabrication of the Space Shuttle Main En- gine。 In the system under development, some76187

ardous situation, it may be necessay' for  him to initiate the E-stop sequence。 Should the operator not be within reach of the button, how。'ever。 he may be unable to take the nec- essay’ action, and。 as a result。 could suffer serious injur)'。 Hax'ing the capabilit)' of stop- ping the robot by issuing a voice command could significantly' improve the operator’s safe)' by enabling him to stop the robot es'en u’hen not within reach of the E-stop button。 Manual corrections are occasionally' re- quired to adjust the location at v•'hich the v。'eld filler v• ire enters the xs'eld pool。 Proper ents' location is absolutely critical to sound weld  quality。  Adjustments  are made either

Introduction

functions do not yet have automatic  closed-loop control。 thus requiring continuous mon- by manually adjusting mechanisms that holdthe v•'irefeed guide tube or by issuing tactile 

Voice recognition  technology' is now "‘itoring and  real-time adjustment by the  hu-

commands  to  a servomechanism。  Use  of a ognized as a potential  means for easing  the workload of operators of complex systems [1]。 Numerous applications have already been implemented, are in various stages of development, or are under consideration [2]—

{5]。 These include data entry。 control of air- craft systems, and voice identification and s'erification for security purposes。

man operator。 Presently', these overrides are input to the system through tactile com- mands (i。e。 。 pushing buttons。 turning knobs for potentiometers, or adjusting mechanical devices)。 Since the operator monitors the process primarily visually, he must either look away from the process to find the proper

button  or  knob  or  rely'  on  “muscular mem-

voice recognition system could eliminate the need for the operator to place his hand within the working envelope of the robot end effec- tor or。 if servomechanisms are employed, could improve speed of response and stabil- ity。

Another aspect of robot operation in an industrial environment that is very important