So  let's  get  started. So  thank  you  for  attending  the  iron  seminar  today. So  today  we  will  have  Dr.  Elisa  or camera  from  Stanford  as  our  guest  speaker. Dr.  Cabrera  resale  the  budget  degree  from UC  Berkeley  in  1994 and  a  mass  or  a  PhD  degree  from  Stanford. She's  Colonel  Richard  Wilson, Professor  of  Engineering  at Stanford  University  in the  Mechanical  Engineering  Department, was  a  courtesy  appointment  in  computer  science. Previous  SE  was  Professor  and Vice  Chair  Mechanical  Engineering  at  Johns  Hopkins. Data  or  camera,  has  made a  significant  contribution  to that  general  robotics  community, including  serving  as  co-chair  of  arrows and  editor  in  chief  of  REO  and  many,  many. You  can  remember  it.  And  she received  many  prestigious  awards, including  the  I  triple E  engineering  and  medicine  and  biological, Biology  Science  Technical  Achievement  Award. And  I  Tripoli  robotics  automation started  Distinguished  Service  Award. Again,  many,  many  Award. And  she  is  one  of  the  pioneers in  many  research  field  including  medical  robotics, software  robots  and  haptics, rehabilitation  and  prosthetics, and  also  engineering  education. So  today  she  will  give a  talk  on  software  bug  Spock  humanity. So  please  join  me  in  welcoming  Dr.  Nakamura. Awesome.  Thanks  so  much  for  hosting  me UA  and  thanks  everyone  for  coming. I  usually  when  I  come  and  give  a  talk, I  only  spend  one  day, but  here  I  find  it  a  day  and  a  half, a  little  over  than  that  because  that's  how  much  you  need. Probably  I  need  five  days  to  have a  reasonable  visit  to  Georgia  Tech  because there's  so  much  amazing  stuff  going  on  here. So  I  really  appreciate  all  the  labs  again, students  and  faculty  that  gave me  tours  over  the  last  day  and  a  half. So  I  ask  you, what  should  I  talk  about  today? Because  there  are  these  different  topics that  my  lab  works  on. And  it's,  it's,  it's  heart  wrenching  because  there's amazing  biomechanics  and  rehab  folks and  I  could  talk  about  that  stuff. I  could  talk  about  soft  robots  or  haptics. But  I  decided  to  focus  on, on  soft  robotics  because this  is  an  area  that  I'm  especially  passionate right  now  in  terms  of  making  soft  robots  that  are not  only  a  proof  of  concepts  for new  techniques  and  new  materials  methods, but  also  things  that  I  think  we'll  get  out into  the  world  hand  and  help  people. I'm  not  making  a  big  claim  that I've  done  that  yet  in  terms  of  helping  people, but  that's  really  where  we'd  like  to  go. And  so  hopefully  here  you'll  see  a  bit of  a  roadmap  in  a  series of  projects  that  we  worked  on  that  are taking  us  toward  that  direction. So  since  I  know  that  this  is a  pretty  broad  robotics  audience  here, I  thought  I'd  just  give  you  my  definition  for  what a  soft  robot  is  because  there  is  quite  a  spectrum. So  this  is  a  relatively  old  science  robotics  paper  figure from  LacI  and  collaborators. Were  they  really  characterized  soft  robotics  being  from things  that  might  be  some  flexible  elements, but  being  mostly  stiff  all  the  way to  robots  where  even  the  computation  is  soft and  maybe  use  fluidic  logic  and  chemical  reactions  in order  to  make  the  robot  smart  and  do  things. So  soft  robots  can  be  anywhere  along  this  spectrum. And  one  of  the  things  I  think  to  keep  in  mind  when  you see  the  examples  of  the  types  of  soft  robots  we  work  on, is  that  they  can  be  soft  by  material, but  that  doesn't  necessarily  mean  you  have  to  be using  materials  with  great  elasticity, doesn't  all  have  to  be  elastomeric  base  robots. You  can  also  have  things  that  are  soft  by structure  because  they  have  locally  compliant  elements. And  then  even  when  you  have  materials  that  are  soft, you  can  think  of  them  as  soft  because  they're flexible  or  because  they're  stretchable  or  both. And  most  of  the  things  that  we  work  on are  actually  not  that  stretchy. We  see  that  there  are  some  real  challenges with  working  with  elastomeric  base  materials. If  you  want  to  have  impact  on  the  real-world, you  have  to  be  able  to  apply  forces. That's  the  point  of  it  being  a  physical  robot. And  so  you'll  see  a  lot  of  robots  that are  based  on  materials  that  are  flexible, but  purposefully  not  stretchable. And  well,  people  often  say  that  these  robots  are  safe. That's  not  inherently  true. We  need  to  work  to  make  that  happen. They're  also  not  inherently  adaptable. We  need  to  work  that  make,  make,  make  that  happen. And  the  same  thing  with  costs. So  these  benefits  can  include  these  things. But  I  feel  like  we  should  never  assume  as some  people  do  claim  and  the  soft  robotics  community, that  all  of  these  things  are  true. They  have,  they  have  to  be  made  true  by  design. So  I  thought  I'd  review three  different  types  of  projects  today. And  I  have  to  say  this  is  a  very  eye  candy  video. Have  you  talk  of  a  single  equation  in  here? But  hopefully  it'll  give  everyone  a  good  intuition for  the  work  that  we  do. Ranging  and  firms  the  little  bit  of an  intro  and  patient-specific  medical  robotics, which  is  where  we  started when  working  with  flexible  devices. Towards  soft  robots  that  grow  by tip  aversion  and  then  towards soft  wearable  haptic  devices. So  I  just  want  to  kick  this  off  by saying  that  there  is  a  long  history  of  continuing  robots. I  don't  think  I've  done  a  great  job  in  this  talk  of having  references  before  every  topic. So  let  me  just  say  overall  that  there's  been an  amazing  amount  of  work  in this  field  using  different  types  of  materials, really  amazing  controllers  and  planners, all  sorts  of  contributions. And  yet  all  of  these  challenges  remain. And  so  anybody  thinking  about  getting  into  this  field, there  is  plenty  left  to  do. But  we're  also  standing  on  the  shoulders of,  of  giants  here. Another  thing  I  want  to  note  is that  a  lot  of  these  soft  robots  are based  on  biological  inspiration. So  in  particular,  the  soft growing  robots  that  I'll  talk  about for  a  good  portion  today is  based  on  biological  inspiration. And  folks  here  at Georgia  Tech  ranging  from Dan  Goldman's  lab  to  David  who's  lab, are  really  inspiring  mechanical  engineers and  roboticists  to, to  build  on  concepts. And  in  particular,  one  of our  robots  is  based  on  the  idea  of  vines, something  that  grow  that  the  tip, we  don't  use  the  exact  same  mechanism. So  this  is  not  biomedic,  but  rather  bio-inspired. But  to  start  with  the  kind of  medical  robotics  application, I  would  say  one  of  the  first  soft, flexible  robots  that  we  worked on  was  this  concentric  tube  robot. And  this  is  something  that,  UH, one  here  has  made  great  contributions  to, especially  in  terms  of  control  and  planning. And  we're  saying  you  as my  academic  grandchild  was  made  me  feel  awfully  all. But  this  is  the  work  of  Bob  Webster when  he  was  my  PhD  student. And  let  me  get  that  replaying. The  idea  here  is  that  you'd  like  to  have a  robot  that  has  a  really  thin  profile. So  if  you're  gonna  do  this  for  surgery, you'd  like  to  have  the  thinnest, least  invasive  possible  structure  for  your  robot. And  so  in  order  to  do  this, we  use  the  inherent  material  properties  of  nitinol  tubes, which  are  super  elastic. Lot  people,  when  they  think  nitinol, they  think  it's  a  shape  memory  alloy,  which  it  is. But  we're  using  it  for  its  super  elastic  properties, which  means  it  can  bend  a  lot without  getting  plastic  deformation. With  these  types  of  robots, what  you  can  do  is  take concentric  tubes  that  are  pre  bent. And  as  you  insert  and  spin these  tombs  with  respect  to  each  other, you  can  get  this  kind  of  tentacle  like  behavior, or  it  depends  on  who  you  are. Some  people  see  an  elephant's  trunk  that  when people  see  a  snake, this  one's  a  little  more  tentacle  like. But  the  idea  is  how  do  we kind  of  cleverly  used  materials  and the  inherent  mechanical  properties  of the  components  in  order  to  get  these  interesting, interesting  multi-device  of  freedom  behaviors. So  this  is  something  we  worked  on  quite  awhile. And  of  course,  the  goal  is  medical  robotics. But  in  this  particular  case, let  me  describe  how  we're kind  of  making  it  more  human-centered. And  I'll  start  just  with  an  example. So  I'm  a  mechanical  engineer, but  as  cars  become  more  complex, more  computer  controlled,  less  accessible  to  me, like,  I  don  t  know  how  to  fix  my  car  anymore. Maybe  it  depends  what  kind  of  car  you  have, who  he's  not  the  kind  of  cars  we  have. Similarly  in  the  medical  field, there's  this  history  of  clinicians ranging  from  surgeons  to  nurses, who  mechanically  went  around  and  design their  own  medical  devices  to  meet  the  needs  at  hand. But  as  we  kind  of  abstract  our  technology from  this  low-level  hands-on  making  capability. We  lose  the  potential  for these  experts  to  come  in  and  help  in  the  design  process. At  least  at  a,  at  a  tight  loop with  the  actual  procedures  that  are  done. So  here's  an  example  about  how  we're  trying  to  make this  particular  software  robot  BB  part  of  humanity, which  is  starting  out  with  3D  models  of a  particular  patient  and then  having  a  virtual  environment. And  this  can  be  a  patient-specific  planning  process  where the  clinician  or  other medical  professional  can  go  and interact  with  that  virtual  environment  and  say, this  is  the  path  that  I  want  to  follow  in  order to  hit  a  particular  target,  Say  e.g. to  get  to  a  kidney  stone  and a  pediatric  patient  where the  organs  are  really  small  and  tight  and  close  together. Then  in  real  time  we  can fabricate  before  we  used  to  use  nitinol, but  lately  we've  been  doing  3D  printed out  of  soft  materials  or  semisolid  materials. These,  these  concentric  tubes. And  then  we  can  add  them  on  to conventional  base  robot  structures and  then  be  able  to  drive  them. And  usually  we  have  a  human  in  the  loop  tele operating  these  in  most  medical  procedures. Although  like  many  labs, we  are  interested  also in  making  autonomous  surgical  robotics. This  is  an  interesting  example  where  of course  you  have  the  medical  application, which  is  just  human-centered. But  how  do  we  think  about  humans  being involved  in  the  design  process? And  I  won't  really  touch  on  this  later. But  even  with  haptic  devices  as well  as  soft  growing  robots, we've  been  looking  at  these  human  in the  loop  design  processes. And  how  do  we  rapidly  fabricate  devices that  someone  can  design  for whatever  their  own  purposes  are. And  I  should  note  that  I'm  going  to  highlight  some  of the  work  of  former  students who  have  gone  on  to  faculty  positions. So  this  was  a  PhD  work  of  Tanya  more emotive  is  now  on  the  faculty  at  UC  San  Diego. So  I  mentioned  that  these, that  these  materials  are  3D  printed. And  so  thinking  about  new  fabrication  techniques is  paramount to  think  making  progress  and soft  robotics  and  3D  printing, of  course,  has  revolutionized  what  we  can  do in  robotics  and  many,  many  different  ways. I  wanted  to  point  out  that  it's  not  just  the, the  tubes  themselves  that  are  3D  printed, but  we  can  even  make  novel  driving  mechanisms,  e.g. using  a  waffle  gear  type  pattern  to  make  it really  compact,  lightweight  driving  mechanisms. In  this  case,  you  have  a  six-degree  of freedom  robot  run  by  a  fairly  small  motors, and  then  they  all  Nest  into  a  really  compact  package. You  can  either  hold  in  your  hand  or  attached to  another  base  robot  if you're  using  a  large  surgical  robot. Also  hidden  pneumatic  devices. We're  often  going  to  gloss  over where  the  power  comes  from. But  this  is  an  example  where  we  really upfront  the  design  in  terms  of  how  do  we, how  do  we  drive  such  a  robot? That  was  just  a  brief  bit of  intro  and  medical  robotics  where we  started  in  these  tentacle  like  devices. But  I'm  going  to  spend  most  of  my  talk  talking about  these  soft  going  vine  robot. And  this  is  a  project.  It  was  really spearheaded  by  Elliot  Hawkes  when  he  was a  postdoc  in  my  lab  and  he's now  on  the  faculty  at  UC  Santa  Barbara. And  doing  all  kinds  of  amazing  things like  jumping  robust  the  jump  five  stories  high. But  I'm  not  going  to  talk  about that  kind  of  thing  today  is  that  I'm going  to  talk  about  these  soft  green  robots. And  some  of  the  stuff  is  very  historical, like  five  years  old  or  so, but  then  some  of  the  latest  things that  haven't  even  been  published  yet. So  in  this  project, what  we  do  is  we  take  a  soft,  thin-walled  material. In  this  case,  it's  plastic like  the  consistency  of  a  Ziploc  bag, usually  LD  PE,  low  density  polyethylene. And  if  we  put  it  in  a  tube  shape  inside  of  itself, and  then  we  apply  air  pressure  or some  other  fluid  pressure, then  it  extends  out  of  the  tip. This  gives  you  this  vine  like  behavior, kind  of  like  a  growth  habit  of trailing  like  these  plants  have. We've  done  a  lot  of  fun  things  with  these. So  in  order  to  get  them  outside  the  lab, you  do  need  to  package  them  and  have  a  way of  actually  delivering  power  on  this  thing  that  can  grow, in  this  case  about  the  length  of  a  football  field. So  we  take  this  thin-walled  tube and  when  it  doesn't  have  air  in  it, it  compresses  to  very  small  so  you  can spool  it  up  and  put  it  into  a  very  compact  base. This  is  a  base  right  here. Let  me  see  if  I  can  get  this  thing  running. So  this  is  the  entire  base  that  has  grown  this  robot. I'm  not  sure  if  you  call  it  a  robot yet.  We'll  get  back  to  that  later. That  grows  the  length  of  a  football  field. This  case,  it's  using its  embodied  intelligence  as we  in  soft  robotics  like  to  say, it's  just  bouncing  off  the  walls  and  growing wherever  it  needs  to  grow  in  order to  get,  to  get  somewhere. So  it's  not  yet  a  very,  very  smart  robot, but  nonetheless,  it  could  grow a  very  long  distance  and  deliver  something  somewhere. Then  it  has  all  of  these  other  capabilities which  I'm  gonna  get  to  in  future  slides. So  what  would  make  it  a  robot? It's  probably  not  just  growing  it, that's  just  the  plastic  bag  filled  with  air. So  if  we  want  to  make  it  actually  into  a  robot, we  need  to  have  a  way  of  steering  it  and  using sensor  feedback  in  order  to  control  his  position. And  my  lab  has  looked  at a  lot  of  different  steering  techniques. Many  of  these  were  driven  by  my  former  PhD  student, Lauren  blooming  shine  His  now  at  Purdue. And  what  she  and  other  students  the  lab  has  done  is look  at  what  are  all  the  different  ways  you  can  steer. Sometimes  to  make  the  robots  as  compact  as  possible, it's  useful  to  have  a  set direction  change  like  up  at  the  top  where  we  had these  sort  of  hooks  that  when  they  become unhooked  would  be a  permanent  direction  change  and  you  can't  go  back, but  it's  a  very  compact  way  of  steering. And  then  the  ones  that  we're  showing  here and  videos  are  not looping  the  way  I  was  hoping  they  would. This  is  one  where  we  have  a  tendon  and we've  figured  out  that  the  geometric  geometry  and the  kinematics  for  how  do you  route  these  tendons  in  order to  get  it  to  go  into  particular  shapes and  configurations. These  are,  these  are  set  in  advance, but  then  we  have  other  actuators  that  have  to  be  flexible enough  to  survive  that  eversion  process, but  still  be  able  to  create enough  bending  force  in  order  to  create  movement. And  in  these  cases, these  actuators  allow  you to  have  reversible  direction  change. So  this  is  actually  autonomously steering  towards  this  light. Very  simple  control  algorithm just  to  try  to  keep  it  centered  on  the  light. And  you  have  four  of these  pneumatic  tendons  that  when  you  fill  them  with  air, they  decrease  their  length  and then  it  bends  in  that  direction. So  we  can  get  the  2D  steering  plus  the  growth degree  of  freedom  using  that  type  of  actuator. So  more  recently,  we  have  been  trying  to  really understand  what  is  the  optimal  actuator to  use  for  these  types  of  robots. It  turns  out  that  with  these  kinds  of  robots  there, this  whole  class  of  robots, which  I  would  call  inflated  beam  robots. You  have  this  battle between  applying  air  in  the  main  body, which  gives  it  structure  and  keeps  it  stiff. And  then  the  bending  actuator, which  is  trying  to  bend  against  that  stiff  center. So  figuring  out  how  to  balance  those  pressures and  how  to  structure a  various  pneumatic  actuators  to  make  this  work  right, has  been,  I  think,  a  big  challenge  for all  robotics  researchers  trying  to  do  pneumatic  control. We've  recently  moved  beyond  like my  inherent  skill  set  in  robotics  towards  working closely  with  folks  who  can  do  modeling  of complex  interactions  between  different  materials and  figuring  out  e.g. in  this  case  that  a  particular  structure  of these  goblin  pouch  motors might  work  better  than  a  different, a  different  type  of  structure. And  also  what  materials  can  we  optimize? So  I  think  we're  starting  point,  the  starting  point  in this  field  where  we  can  start  to  optimize the  materials  and  actuator choice  and  pressure  selections for  a  particular  application. And  what  we  want  that  robot  to  do. Another  way  in  which  we've  recently, than  thinking  about  how  do  we  make these  robots  a  little  more  agile, is  to  actually  exploit  dynamics. So  this  is  work  that's  under  review  and collaboration  with  former  Georgia  Tech  person  karen  LU, where  we're  trying  to  take under  consideration  that  sometimes we  don't  have  enough  straying  from these  actuators  to  do  the  kind  of  movements  that  we  want. If  we  want  to  get  to  a  curve  or  hit  a  target  where the  robot  has  to  swing  up and  touch  something  way  up  here. We  just  may  never  be  able  to  get  there  with  the  types  of pneumatic  actuators  and  they  might  have power  that  will  have  on  board. How  do  we  integrate the  dynamics  of  movement  of  the  base  of  the  robot,  e.g. if,  if  it's  hanging  from  a  drone  along  with those  soft  robot  actuators  in  order  to get  it  to  hit  a  particular  target. Along  with  Carolyn  has  group, my  graduate  student  has  done some  really  nice  work  and  trying  to  figure  out, first  of  all,  how  do  you  model these  systems  in  a  dynamic  way? Most  soft  robots  have  static  models  and  not  dynamic  ones. But  also  how  do  you  actually  learn  to  control  policy? And  in  this  case,  they  use reinforcement  learning  from  lots  of iterations  done  in  simulations based  on  the  learned  model. And  then  you  do  learn  this  policy that  can  achieve  various  tasks. And  so  our  goal,  although  we  haven't  done  it  yet, eventually  is  to  be  able  to  put  this  on  a  drone. And  the  idea  is  that  there  are places  where  even  drones  can't  go. Can  you  get  close  enough? Can  you  swing  your  vine  robot in  there  and  can  then  you  then  grow and  drive  down  some  desired  path  to  do  inspection, deliver  something  maybe  in a  search  and  rescue  scenario  or  other  applications. So  one  of  the  big  challenges that  we  have  in  these  robots,  like  I  said, is  to,  in  order  to  make  them  stiffer, you  need  to  apply  more  pressure. So  this  inflated  being  concept, you  apply  more  pressure, you  get  more  stiffness  if  you  make  the  robot  bigger, that  will  also  make  it  stiffer. So  all  the  mechanical  engineers  in  there,  right? You  take  a  cylinder,  you  make it  bigger,  it'll  get  more  stuff. And  then  this  is  not  our  work. This  is  other  words  just  for  reference. Even  things  that  came  out  before  we  did our  soft  growing  robot just  shows  some  really  big  inflated  being  robots. This  one  is  probably  about this  big  around  or  maybe  even  bigger. This  is  a  really  cool  robot. I  think  it  has  a  helium  inside  of  it  to  help  it  go  up. For  plates.  They  putting  out  fires,  e.g. and  this  is  one  type of  application  for  these  inflated  beam  robots. They're  often  driven  in  these  cases  by tendons  in  order  to  create,  to  create  bends. We're  really  interested  in  how  do  you  actually make  these  types  of  robots differ  at  the  smaller  scale without  having  to  scale  up  in  size. So  one  way  to  get  an  increased  resistance  to cantilevered  loads  is  to use  a  technique  we  call  layer  jamming. And  I  apologize,  I thought  I  had  a  slide  in  here  about  it, but  since  I  don't,  I'll  describe  it. You  might  also  be  familiar  with particle  jamming  and  I'll  have  an  example  of  that  later. But  the  idea  with  layer  jamming  is  if you  have  materials  that  are  close  together, if  they  have  a  lot  of  friction  in  between  them, they  resist  bending,  right? So  if  you  vacuum  out  the  air  effectively  in-between  them, then  they  don't  want  to  bend  in  there  stiff, but  if  you  inject  air  than  they're very  loose  and  they  can  bend  easily. We're  using  this  layer  jamming  approach to  increase  the  stiffness  of  the  soft  going  robots. And  when  they're  spooled  up, the  air  and  air, positive  or  negative  pressure  is  taken away  such  that  it  can  easily, easily  curve  and  have this  very  small  stowed  configuration  in  the  base. And  this  is  a  nice  kind  of  stiffness, stiffness  modulating  property  that when  you'd  haven't  made  it  stiff, it  can  still  be  flexible  enough  to  undergo the  inversion  process  and  grow  as  shown  here. And  once  again. Using  SVM  models  and being  able  to  show  both  experimentally  and optimizing  the  shapes  of these  stiffened  regions  in  order  to get  certain  desired  stiffness  properties of  certain  locations  on  the  robot. And  it's  important  for  us  not  to  stiffen  the  whole  robot. Because  if  you  do  that, you  can  think  of  an  entire  different  way  of  steering. So  let's  imagine  that  you  have these  different  sections  of  the  robot and  you  start,  well,  that's  animating. I  didn't  expect  that  to  happen. Okay. So  let's  say  you  start  by  growing  from the  tip  and  then you  want  to  aim  it  in  a  particular  direction. So  you  could  then  rotate  the  base. You  could  do  is  you  can  make a  particular  region  less  stiff. And  if  you  use  one  of these  tendon  pool  cables  or really  any  of  the  steering  methods. It  is  going  to  bend  theoretically  at,  at  a  joint. So  you  can  kind  of  think  of  these  as  virtual  joints by  stiffening  at  particular  locations  along  the  robot, you  get  a  place  where  it's  more  likely, it's  like  the  minimum  energy  state, it's  more  likely  to  bend  at  that  location. So  this  is  a  kind  of low  actuation  count  method  of  steering. And  in  fact,  my  grad  student,  Brian, who  I  think  it's  skipped  over  his  slide. See  if  I  can  get  this  shown  here. Really  didn't  want  to  show  that. We  try  it  one  more  time. There  we  go.  There's  Brian. He  calls  these  activators  rather  than  actuators and  other  people  and mechanical  dream  might  be  familiar  with  this  concept. You  have  a  property  of  an  actual, What's  an  actuator  that  actually  doesn't  do  any  work, so  becomes  an  activator. So  this  stiffening  and  Stephanie  and behavior  is  really  an  activator  which then  enables  enables  you  to  really  lower  the  number  of powered  actuators  that  are  required  and therefore  the  overall  power  required  by  the  system. So  these,  these  virtual  joint techniques,  scooting  back  forward. Let  you  create  these  different  robot  shapes with  a  really  minimum  of  actuators. And  so  with  Karen  Lewis  group,  they  do  the  planning. We're  not  really  planning  people. They're  figuring  it  out  if  you  want  to hit  a  certain  target. What? I  hope  this  doesn't  keep  doing  this the  whole  presentation  I  might  have  to  step  out  and  see. But  when  we  add, go  back  one  more  time, when  we  have  these  different  targets, we  figure  out  not  only  the  optimal  series of  actuations  and  activations, but  also  what  range  of  designs  like  what lengths  of  the  stiffening  elements will  actually  make  that  happen? Sorry,  my  slides  are  doing  this. Alright. It  wants  me  to  move  on. So  I'll  move  on  from  stiffening  robots. This  video  will  play  a  while.  So  let  me,  let  me  talk. So  as  I  mentioned  at  the  beginning, one  of  our  goals  is  to  really  get  these  robots out  of  the  lab  and  into  the  field. And  one  of  the  first  things  that  we  did  was really  spearheaded  by  my  former  student, Margaret  code  has  now  at  Notre  Dame. And  she  was  very  adventurous  and  she took  her  vine  robot  to  an  archaeological  expedition  in Kevin  Peru  and  actually put  it  into  these  locations  where  these underground  tunnels  which  are  then  buried  for a  long  time  that have  actually  never  been  seen  by  human  eyes. They've  had,  they've  tried  to  send in  other  wheeled  robots. And  their  best  techniques  so far  is  really  a  GoPro  on  the  end  of  a  long  stick, but  can't  turn  a  90  degree  corner has  all  kinds  of  limitations. So  we  were  able  to  send  her  van  robot, which  can  go  up  to  about  10  m  deep into  these  environments. And  this  is  a  system  where  you  could  see her  a  minute  ago  carrying  it  on  a  wheelbarrow. Like  the  whole  robot  has  to  go  down  there. And  it's  really  hard  to imagine  making  a  portable  system  that can  have  that  kind  of  reach  using  any  other  method. So  you  do  need  power. So  they  do  have  power  at  the  archaeological  site. But  all  of  the  stuff  kinda  has  to  fit  into a  wheelbarrow  in  order  for  you  to  get it  to  the  site  and  be  able  to  implement. So  that  was  super  exciting  to  be  able  to  go  in  and have  a  tip  mount  on  the  robot. I  don't  know  if  you  saw  that  in  any  of  the  pictures, but  with  a  camera  on  it  that  gets  pushed  forward. In  this  case,  our  tip  amount  was  not  very  sophisticated. It's  like  a  Fock,  the  tip  of  the  robot. It  just  gets  pushed  forward,  what  its  grows, and  then  you  just  make  sure  to wheel  it  back  in  when  you're  done. Alright,  so  that  was  one  real-world  application. The  one  that  we're  working  on  now  is  actually  really  fun. I'm  near  the  main  campus  of  Stanford, are  really  within  the  main  campus  of  Stanford. There's  an  area  which  is  the  home  to the  endangered  California  tiger  salamander, which  is  the  cutest  little  salamander. There  it  is,  up  there.  This  is  also  one of  the  dirtiest  projects  my  lab  has  ever  gotten  into, especially  this  year,  it's  a  very  wet  and  muddy. Out  in  California  this  year. But  what  we've  been  doing  is  taking  our  vine  robots  out to  this  area  of  campus  and  then  putting  them  underground. So  I  don't  know  if  you  can  see  it, but  right  on  the  upper  right  is  a  hole  in  the  ground. And  we're  going  to  Irvine  robot  into  there. And  once  again,  this  whole  system. Now  it's  actually  much  smaller. It  even  fits  in  a  largest  backpack, but  not  a  wheelbarrow. And  it's  something  that  we  can  take  around. And  the  reason  why  this  is  important is  that  the  California  tiger  salamander, as  I  mentioned,  is  endangered. And  there  isn't  a  good  way  to characterize  what  their  environment  is  like. What  is  the  temperature  and  humidity  in  their  holes, and  also  how  is  changing weather  and  climate  affecting  their  ability  to  reproduce. So  one  of  the  things  we're  really  excited  about  is  that this  year  we  have  a  ton  of  rain in  California  and  we  think  we're  going  to  have  a  boom  in the  population  of  California  tiger  salamanders. And  that  means  we're  going  to  increase the  likelihood  of  actually  observing them  in  their  habitat  when  we  go  into  these  burrows. And  that  means  we're  gonna  be  able  to  start  correlating their  preferences  in  terms  of temperature  and  humidity  to  their  occupancy. So  we  are  not  the  biologists, but  we  have  amazing  ecology and  conservation  biologists  collaborators  that are  just  so  hungry  for  this  data  so  they  can understand  What's  happening  to  this and  other  subterranean  species. So  how  are  we  accomplishing? So  you  might  think,  on  one  hand, it  is  really  easy  to  grow  vine  robots  through  pipes. You  don't  really  need  to  steer  them  overall because  if  you're  going  through  a  pipe  or  a  tunnel, the  tunnel  walls  kinda force  it  to  go  one  direction  or  another. But  you  do  have  to  make  decisions  at  some  branches. But  at  least  that  means  you  don't  need  to  have a  steering  technique  that  runs the  whole  body  of  the  vine  robot, like  I  showed  in  some  previous  examples. We  really  just  need  steering at  the  tip  and  we  are  using  some, some  tendons  steering  within  a  flexible  joint  at  the  top. And  part  of  the  reason  this  tip  mount  device  has multiple  sections  is  that  we  need  a  camera, we  need  a  temperature  sensor, we  knew  humidity  sensor, we  need  an  IMU  sensor. There's  no  GPS  underground. So  we  need  to  fit  all  of these  different  sensors  into it  and  we  couldn't  fit  them  all in  one  segment  that  wouldn't  be  so  large  that  it would  prevent  it  from  following  some of  the  turns  and  the  boroughs. So  we  have  a  version in  the  lab  to  actually  see  what  it's  doing. But  we've  also  done  this  underground  and we're  able  to  track  its  position  and  record  data, which  is  very  exciting. And  just  as  an  example  that  we  can't really  see  it  underground. This  is  an  example  of  actually  taking temperature  as  you  go  into  the  burrow. And  you  can  see  the  outside  temperature  versus the  temperature  deep  inside the  borough  is  quite  different. Yeah,  that's  exciting. And  I  told  you  that  for  pipes  and  Burroughs  is relatively  easy  because  you have  this  constraint  of  the  environment, basically  helping  your  robot  go  where  it  needs  to  go. But  for  vine  robots  in  other  types  of  applications. So  one  example  is a  contract  we  have  with  the  FBI  has  interested  in  using it  for  inspection  in  places where  there  might  be  like  a  weapon  of  mass  destruction. And  so  you  might  want  to  cut  a  hole  into a  mystery  van  and  then  put the  robot  in  and  have  it  go  wherever. And  in  that  case,  we  needed  to lift  up  against  gravity  and we  need  it  to  be  able  to  go  have  truly  3D  motion. And  we  don't  necessarily  wanna  do  a  whole  lot  of  relying on  pressing  on  the  environment in  order  to  get  you  where  you  want  to  go, then  you  might  press  on  the  wrong  thing. If  we  want  to  control  this  thing  in  free-space, we  need  to  go  beyond  tip  steering. We  need  to  go  beyond  having one  actuator  that  goes  along the  whole  length  of  the  body  as  I  showed  before. Instead  we  need  multi-segment. So  what  used  to  be  a  very  simple  and  elegant  idea is  quickly  getting  more  and  more  complicated. So  in  order  to  do  multiple  turns, we  wanna  do  this  without  touching  the  environment. We  want  to  have  many  degrees  of  freedom. I'm  trying  to  avoid  having many  different  parts  and  the  vine that  would  make  it  wider. And  we'd  like  to  have  stability without  an  act  of  air  supply. This  is  where  we're  going  to  present  at robust  soft  next  week, next  week,  or  the  week  after.  Very  soon. And  I'm  going  to  play  this  video  and  kind of  talk  you  through  it.  Here. What  we  did  is  we  invoked  a  tip  steering  method. But  what  we  do  is  we  just  locally  turn on  the  pouches  that  create  bends  in  the  robot. That  now  happens  just  as  it  grows. There's  a  mechanism  in  the  tip. There's  one  pressure  supply  line  for  the  whole  body. And  then  at  these  different  pouches  we  have these  special  magnetic  valves  that  there's a  magnet  and  the  tip  that  can open  a  valve  if  we  want  to  turn  it  on. And  then  when  this  valve  turns  on, then  you  can  actually  selectively  turn on  little  pouches  that  will  then  create  a  band. And  I  had  a  fun  time  talking  with  some  of Dan  Goleman  students  about  how  this  might  mimic  some, some  types  of  plant  growth. I  told  you  before  we  had  very  typical  tip  melts. Now  we  have  very  complicated  tip  mounts. And  it's  complicated  because the  vine  robot  doesn't  have  a  constant  tip like  the  material  keeps  changing  so the  material  has  to  be  able  to  slide  through. The  tip  mount. This  is  showing  a  magnet  opening  some  valves. And  then  here  you  can  see the  tip  Mount  going  along  and  then selectively  turning  on  pouches in  order  to  do  a  steering  move. And  when  you  retract, which  I  think  the  video  will  eventually  show, maybe  I'll  skip  forward  to  that. Yeah. When  you  retract,  you  can  also  turn, turn  off  and  release  the  air  from  those  passions  as  well. So  this  thing  has  motors  in  it  that basically  steer  it  backwards so  that  you  can  retract  the  vine  robot  as  well. So  you  can  get  now  lots  of more  complicated  shapes  that  would  actuate  in  free-space. We  have  not  yet  achieved  3D. We  can  do  this  on  the  ground, but  we're  going  to  need  more  pressure  and more  strength  in  order  to be  able  to  lift  it  up  against  gravity. And  we've  tried,  it  turns  out  that there's  not  enough  volume  of  helium  that you  could  put  in  this  thing to  counterbalance  its  own  weight. I  should  say  that  this  work  was  done  in  collaboration with  rolling  secrets  group  at  ETH  Zurich. The  next  thing  to do  is  instead  of  just  staring  at  the  tip, is  to  think  about  what  if  we  want  to  put just  segments  of  steering  and  the  robot? I  don't  think  this  is  hugely  groundbreaking, but  it  has  really  made  us  think about  the  manufacturing  techniques  for  these. So  over  the  last  year, so  we've  gotten  this  really  cool  tool called  an  ultrasonic  welder. Kinda  looks  like  a  sewing  machine. And  you  can  basically  well  materials  together. People  in  the  textile  industry, it  turns  out,  use  this  all  the  time. One  of  our  goals  now  is  to do  these  kind  of  unit  body  fabrications. But  instead  of  my  post-docs and  it's  doing  it  by  hand  here, like  this  should  totally  be  automated. And  so  we'd  like  to  have  an  ultrasonic  welding  technique. Where  are  we  now  automate the  fabrication  of  manufacturing? And  although  vine  robots are  cheap,  the  materials  are  cheap. The  human  labor  costs. Even  graduate  students  and  postdocs are  very,  very  enormous. And  so  we  really  need  to  find ways  as  a  whole  in  the  software  bugs  community. But  in  particular,  there's  a  great  opportunity  for  these inflated  being  timed  robots for  us  to  come  up  with  better, more  efficient  manufacturing  techniques. And  that's  something  we're  aiming  to  work  on  next. Great,  making  good  time. So  I'm  going  to conclude  this  is  the  last  section  of  the  talk. Speaking  about  haptics  and  in particular  soft  wearable  haptic  devices. And  the  motivation  for  this  is  that  haptic  interfaces, interfaces  that  provides  stimulation to  the  human  using  their  sense  of  touch, can  enable  all  sorts  of  useful  communications. I'll  hit  at  the  end  about  social  tasks. So  human,  human  communication  is  really  important. Many  people  realize  this  even  more  over the  pandemic  that  we're  missing, that  aspect  of  human,  human  communication. But  even  now,  it  would  be  great  for people  to  be  able  to  communicate  more by  the  sense  of  touch  when  they're  remote. Human  Asia  communication,  I  like  to, I  like  to  run  and  sometimes  I have  my  watch  giving  me  directions. But  it  usually  vibrates  to  tell  me, oh,  turn  left  or  turn  right. I  have  to  look  at  the  watch  in  order  to  see, is  it  telling  me  left  or  right? So  can  we  increase  the  information  throughput  of these  haptic  devices  to  get useful  information  from  intelligent  agents  and  our  world, as  well  as  autonomous  robots. If  we're  gonna  be  interacting  and crossing  physical  spaces  with  robots, it  may  be  useful  to  have  this  more  natural  form  of communication  that  can  tell  you where  things  are  located  in  your  space. So  when  we  first  started  designing  soft  haptic  devices, this  is  the  first  thing  we  thought  of. This  is  really  fun, but  it  is  not  a  good  idea. The  goal  here  was  to  create an  active  surface  like  literally  a  kind  of digital  clay  with  kudos  to wane  for  pushing  this  digital  clay  aspect. Quite  awhile  ago. In  this  case,  we  said,  well, maybe  these  soft  surfaces  can  be  used, in  this  case  for  medical  simulation. We  wanted  to  do  palpation to  find  like  hard  lumps  and  soft  tissue. And  it  may  be  works  okay  for  that. But  the  model  here  is that  in  order  to  provide  haptic  feedback, you  are  literally  trying  to  recreate  the  physical  world. And  that  is  just  a  really  hard  road  to  travel. Instead,  in  haptics,  what  we  prefer to  do  is  use  haptic  illusion. Tried  to  figure  out  how  do  we  trick the  human  brain  into  sensing  something  from the  environment  that's  not  actually  there  and  providing just  enough  stimulation  at the  right  time  and  in  the  right  place  to  do  so. Just  in  case  anyone's  curious  before I  switch  to  how  I  think  we  shouldn't  do  it, just  to  say  how  this  works. So  this  is  a  silicone  elastomer  material that's  divided  into  four  cells. Within  each  cell,  it's  filled  with  coffee  grounds. And  just  like  that  layer  jamming  technique I  talked  about  before, here  we  use  particle  jamming. Coffee  grounds  are  great  for  this  when  you  vacuum air  out  of  a  sealed  chamber  filled  with  coffee  grounds, the  coffee  grounds  locked together  and  they  get  really  stiff. And  people  have  used  this  for  other  types  of cool  surgical  robots  and  other  applications  as  well. But  that's  why  it  got  hard  is  because  it  got vacuumed  and  then  there  was  also  air  pressure  from  below. And  that  would  give  us  a  lot  of  degrees  of freedom  to  try  to  do some  crazy  mimicking  of soft  and  hard  tissues  in  the  human  body. But  nonetheless,  not  the  approach we  typically  want  to  take  with  haptics. Because  what  we'd  really  like  to do  is  utilize  these  haptic  illusions. So  one  way  to  do  that  is  to provide  skin  deformation  and  a  way  to  do  this. This  is  cheaper  and  more  mobile  than  classical  desktop. Rigid  kinesthetic  haptic  devices that  are  more  like  robots. This  case,  this  is  like  a  robot, but  you  wear  it  on  your  fingertip. And  as  my  students, Sam  touches  these  virtual  environments. He  gets  skin  stimulation, not  a  net  kinesthetic  feedback, but  local  skin  stimulation  that  mimics the  contact  that  he  would  have  with the  environment  if  he  was manipulating  those  objects  directly. And  it  turns  out  that  if  you're lifting  fairly  light  objects, like  maybe  a  little  bit  lighter  than  this  thing  of  water. You  would  mostly  be  relying  on  your  cutaneous  or  skin based  mechano  receptors  in  order  to  sense the  mechanical  properties  of that  object  and  not  muscle  spindles, spindles  and  Golgi  tendon  organs or  joint,  or  joint  movements. So  it  turns  out  that  as  a  haptic  illusion, this  skin  deformation  is  a  great  idea. But  those  are  really,  really  hard  to  make. They're  tiny.  They  involve  putting little  tiny  pins  and  the  little  tiny  holes. And  so  our  next  step  was,  well, maybe  we  should  try  to  make  an  origami  haptic  device, one  that  lays  flat  and  then  you  fold  it  up. And  it  turns  out  this  was  a  pretty  good  idea. It  was  easier  to  make. However,  that  still  the  whole  thing was  overall  bulkier  than  we  would  have liked  to  have  in  order  to  truly explore  the  ability  of  someone  to  unencumbered, manipulate  a  virtual  environment  and  especially put  them  on  multiple  fingers  at  the  same  time. The  origami  solution  is  where  we're  getting  there and  we're  continuing  to  try  to  improve  this. But  what  I  want  to  talk  about  today  is  using fully  3D  printed  soft  haptic  devices  that  use  some of  these  kind  of origami  approaches  combined  with  pneumatic  soft, soft  robotics  approaches  in order  to  get  a  multi  degree  of  freedom  system. And  my  postdoc  scientists  did  this  work, TBD  because  he's  interviewing  for  faculty  jobs  right  now. But  hopefully  you'll,  you'll  see him  as  soon  in  a  faculty  position. And  this  is  an  example  of  what  we're  calling  fingerprint. It's  a  monolithic  lead  3D  printed  haptic  device done  on  a  Form  Labs  three  for  anybody, that's  a  Form  Labs  3D  printer  fan. One  of  their  standard  resins, which  is  kind  of  semi  soft. That  is,  if  you  make  a  piece  of  that  material  very  thick, It's  basically  rigid,  but  you  can  also  print very  thinly  and  that  material  will  bend  and  flex. So  it  kind  of  becomes  like the  flexor  elements  in  an  origami  device. Unlike  the  other  devices  I  showed  on  the  last  slide, which  were  three  degrees  of  freedom. This  one  has  four  degrees  of  freedom, can  provide  some  local  skin  stretch  and pressure  as  well  as  vibration. It  has  this  kind  of  monolithic  construction. You  just  print  the  whole  thing. Saying  just  prints  might  be  a  bit of  an  oversimplification  like you  do  have  to  clean  it  out. It's  not  that  easy, but  it  is  fully  printed  at  one  time  out  of  one  material. It  just  requires  some  cleaning  steps  afterwards. Even  though  it  looks  and  is a  soft  robot  which  is  driven  by  pneumatics. And  in  fact,  the  actuators  here, which  I'll  show  on  the  next  slide  are  driven  by vacuum  rather  than  positive  pressure. Underline,  it  really  is  that  it's an  origami  mechanism  with folding  actuators  rather  than  stretching  actuators. It's  really  small  and  lightweight. If  you  don't  count the  giant  building  compressor  that  is  driving  it. So  what  does  this  look  like  in  a  little  bit  more  detail? So  this  is  what the  whole  device  looks  like  in  a  nice  CAD  drawing. And  just  some  things  to  point  out. So  you  have  to  have  channels  for  the  air  that  gets embedded  in  the  thicker  parts  of the  material  because  we're  vacuuming  air  out. Vacuum  is  a  little  tougher  than positive  pressure  because  if  you  vacuum, you  can  collapse  the  channels  that  you're trying  to  use  to  put  the  air  in. So  those  all  have  to  be  in  a  very  thick  area. It  can  easily  be  worn  on the  fingertip  and  because  it's  a  little  bit  soft, it's  actually  very  forgiving for  a  number  of  different  fingers  sizes. Then  it  was  hard  to  see  from the  pictures  on  the  previous  slide. But  there's  the  vector  which  is the  point  that  contacts  the  skin. Just  like  the  other  one  I  showed that  had  that  the  red.in  the  middle, it  contacts  the  finger  pad  right  here. So  it's  as  effective  as  those  kind  of traditionally  constructed  robotic  fingertip  devices. But  now  it  is  based  on  this  softer  material. So  the  origami  for  anybody  that's a  fan  of  these  kinds  of  devices  and wonders  out  what  is  the  fold  pattern? Well,  there's  a  famous  mechanism known  as  though  the  water  bomb. Basically  it's  to  make  a  thing  that  you  can  fill  with water  if  you  make  a  little  origami  thing  out  of  paper. And  so  we  use  the  same  fold  pattern to  have  those  degrees  of  freedom. And  then  you  get these  different  roll  pitch  yaw  degrees  of  freedom. And  if  you  assemble  them  all  correctly, then  you  can  get  the  overall  four  degrees of  freedom  of  motion  that  you  want. For  a  close-up  of  each, each  pneumatic  actuator,  it  looks  like  this. You  vacuum  air  out  and  essentially  it  folds  closed. And  again,  we're  using  a  judicious  selection  of material  thicknesses  in  order  to  get  the  behavior  that  we want  so  that  we  can  make  it  out of  a  single  type  of  material. The  thing  does  pretty  well  compared to  basically  any  wearable  haptic  device  you  can  imagine. It  can  get  forces  on  the  order  of  newtons  in  x  and  y, which  is  actually  pretty  good. We're  very  sensitive  in shear  were  actually  a  little  less  sensitive. And  the  normal  force,  which  is  the  z-direction, actually  works  out  pretty  well  for  this  device  because this  is  where  this  device  is  really  powerful. Because  in  this  normal  direction, all  of  these  pneumatic  actuators  can  work  in  parallel and  you  can  get  tens  of  newtons  of  force. So  this  is  really  exciting. The  caveat  that  we  are  attached  to  a  building  compressor. So  we  are  connecting  it  now  to  portable  pumps. It's  still  not  great  because you  have  to  carry  them  along  with  you. So  you  really  have  to  be  cognizant  of  folks. They  say,  Oh,  I  made  this tiny  haptic  device  or  soft  robot. Then  if  it's  required  to  be  attached  to  air, that  is  really  going  to  limit  its  applications. I  think  that's  another  thing  in  general  that we're  exploring  is  how  do you  use  some  clever  techniques to  really  make  these  things  portable? So  far  I've  been  talking  about  the  fingertips, but  there  are  a  lot  of  reasons  to  go  away from  the  fingertips  and  put haptic  devices  on  other  parts  of  the  body. What  is  generally  for  wearability  and  portability. But  another  is  we're  very  interested  in augmented  reality  and  augmented  reality from  a  visual  point  of  view, you  see  your  visual  world  and  you  get visual  overlays  that  kind  of  cover  up  the  real-world. The  problem  with  doing  this  for  haptics  and the  fingertips  is  that  if  I  want  to  touch  this  desk, but  then  I  want  to  somehow have  augmented  haptic  feedback. I  can't  have  that  at  the  fingertips because  then  I  can  know  hunger  touch  the  desk. So  what  we've  been  exploring  in  my  lab  is  how  do  we  put haptic  devices  on  other  locations  of  the  body, both  in  general  for  portability, but  also  how  can  we  use  this  for augmented  reality  so  that  we  touch  things  in  there. We'll  grow  their  fingertips  and  we  augment  that  with additional  haptic  feedback  at  these  other  body  locations. This  is  difficult  for  a  couple  of  reasons. One,  and  these  purple  dots are  not  meant  to  signify  some  kind  of  disease. They're  supposed  to  demonstrate  the  distribution, the  density  of  mechano  receptors, sensing  within  our  skin. And  we  have  very  dense  coverage on  the  palm  and  the  fingers. This  is  known  as  the  glamorous  or  non-hairy  skin, and  we  have  much  more  sparse distribution  and  other  locations like  the  forearm. And  then  another  challenge, which  is  maybe  a  little  more  subtle, is  that  we  are  very  used  to  explain the  world  around  us  by active  exploration  with  our  fingertips. I  don't  learn  about  the  mechanical  properties  of this  table  by  leaning  over  and rubbing  my  forearm  against  it. My  body  is  not  designed  to  do  that  kinematically. But  also  the  mechano  receptors  are. So  one  of  the  things  that  makes  haptic  devices  general, so  challenging  is  that  you  have  to  let  a  person explore  and  then  the  haptic  feedback  has  to  come  in, in  a  way  that's  appropriate  for whatever  exploration  that  human  wants  to  do. You  typically  don't  want  to  constrain. You  can  only  touch  it  this  way. That  becomes  really  unnatural. So  this  sort  of  passive  versus active  touch  means  that at  least  for  this  augmented  reality  idea, we're  doing  active  touch  with  our  fingers. But  the  feedback  is  coming  elsewhere  and we're  not  really  sure  how  well  that  is  going  to  work. So  I  would  say  that's  still  an  open  question. But  we're  trying  to  drive towards  technology  that  will  help  us answer  this  active  versus  passive  touched  question. So  soft  wearable  haptic  devices that  now  go  off  the  fingertips. What  might  those  look  like? These  are  both  pretty  old  projects, 2,019.20,  18  respectively, where  we  tried  to  take  other  ideas in  soft  robotics  and  put  them  out  there. So  this  is  one  where  we  use the  classic  fiber  reinforced  soft  robotic  actuator. And  the  idea  is  you  have  a  single  point  of  contact  and  it pushes  in  multiple  directions like  we  did  on  the  fingertip. And  then  we  also  had this  nice  work  by  Nathaniel  to  try  to repurpose  our  vine  robot  to grow  out  of  a  bracelet  onto  the  arm, and  then  have  distributed pneumatic  pouches  that  could  squeeze. And  provide  communication.  And  in  this  case, it's  a  single  direction  like  it's squeezes  or  it  provides  pouches  that  inflate. But  each  one  of  those  is  very  simple. It  just  provides  a  single  direction. So  these  are  a  couple  of  approaches  which  were fun  and  kind  of  helped  expand  our  ideas  about  this, but  I  don't  really  think  they're  practical. One  of  the  big  challenges  I'll  say  here, in  particular  is  that  you  need  something  to  push  against. So  a  beautiful  soft  robotic  device now  has  this  rigid  frame  around  it, which  is  allowing  us  to  have a  grounding  against  which  to  push. It's  not  really  a  soft  robot  any  longer. But  with,  with  vanishes  ideas. I  showed  you  this,  this  fingerprint  here. We  found  that  we  can  put  these  kinds  of  things  all  over the  body  and  they  actually  strap  on  quite  easily. So  we've  moved  them  to  the  thumb  and  the  index  finger. The  first  time  we  showed  this  one, someone  told  us  that  looked  like  a  diamond  ring. And  so  now  we  know  students  like  to  swim  around the  lab  on  their  fingertips. And  then  we  have  larger  force  versions  that  use a  different  mechanism  that go  on  other  locations  in  the  body. So  to  put  them  here, we  need  to  have  even  larger  forces. Forces,  these  go  up  to  20  Newtons  of  pressure. That  might  sound  like  a  lot, but  it  actually  doesn't  hurt  at  all. That's  kind  of  what  you  need  to  be able  to  have  a  range  of  forces  that you  display  that  give a  lot  of  information  on  the  forearm. So  this  is  where  we're  going. Dentist  calls  them  haptic  voxels  or  hydroxyls. Not  sure  if  that's  going  to  catch  on, but  we're,  we're  selling  it. And  the  idea  is, can  we  take  these  monolithic  lead  II pre  3D  printed  devices  along  with  what  we  hope  will  be new  techniques  and  portable  pneumatic  power to  start  being  able  to make  multiple  devices  be mountain  in  the  bodies  in  different  locations. And  when  we  do  that, a  couple  of  more  slides, we're  going  to  have  to  figure  out  a  couple  of  things. So  these  are  really  open  problems that  I'm  going  to  throw  out  here. One  is  understanding  how  do  we  map  locations for  this  augmented  reality  problem  that  I  talked  about. If  you  touch  something  with  these  fingertips, where  should  the  feedback  come  in?  My  suit? And  Jasmine  just  published  a  paper  in  Iraq  where she  demonstrated  that  actually, it  doesn't  matter  where  you  put  the  feedback. It  turned  out  people  could  interpret  all  of  them, but  they  did  have  a  preference. They  preferred  a  mapping, which  was  the  one  that  was obvious  to  us  from  the  beginning. You  should  make  them  geometrically  aligned. So  that's  nice,  but we  would  still  like  to  understand  a  little better  how  do  we  optimize  the  placement? So  that's  ongoing  work. Then  another  aspect  of  this  work  is  social  touched. I  kind  of  preface  all  of  this  talk  about haptics  with  the  need  for social  touched  during  the  pandemic. And  this  is  still  true. And  so  we've  done  some  fun  work  where  we've  characterized social  touches  between  people using  a  capacitive  array  sensor  worn  on  the  arm. And  then  we've  mapped those  social  touches  onto  a  wearable  haptic  device. This  one's  not  really  soft  Other  than  the  fact  that their  voice  coil  motor  is  mounted  in  a  soft  sleeve. Not  really  a  soft  device. But  the  idea  is  it's  wearable. And  it  turns  out  that  even  with a  very  sparse  distribution  of  stimulation  locations, you  can  very  well  communicate  a  number  of  social  touches, ranging  from  sadness  and  love,  attention. I'll  give  some  examples.  Pay  attention  to  me, is  attention.  I  love  you. Might  be  a  stroke  like  this  and  it doesn't  feel  like  individual  touches. It  will  feel  like  a  stroke  if  you  get  the  timing  right. And  using  our  data-driven  approach based  on  our  sensing  work  here, we've  been  able  to  figure  out  what  actually  feels  good, what  feels  continuous  and pleasant  and  compelling  to  users. And  this  is  work  that's  going  to  be  ongoing. We're  interested  in  more  real-time tele  operation  of  touch. In  this  case,  we  curated haptic  emojis  based  on  acquired  data. So  that  concludes  my  story about  soft  robotics  going  from medicine  to  these  vine  robots. And  now  the  software  will  haptic  devices. Really  want  to  thank  all  my awesome  students  and  postdocs  who, who  did  all  the  work  and  our  sponsors  and  all  of you  for  listening  and  I  hope  there's time  for  questions.  Thanks. Oh,  yes. Thank  you.  So  I I've  been  following  your  vine  robots  for  awhile. I  was  actually  an  intern  with  Margaret  2020. I  don't  remember  how  you have  at  least  complicated  cameras or  magnetic  valves  that  you  can stick  on  that  on  the  front  of  your  mind  robot, how  do  you  control  them? How  do  you  pass  wires? Is  it  all  wireless? And  how  do  you  keep  that  all  connected? And  how  do  you Why  routing  is,  is  an  issue  that  I've  had for  many  of  my  soft  robots. So  I'm  just  curious. Yeah,  absolutely. Great  question.  So  the  particular  one  that  has the  valves  where  we  do  this  kind of  addressable  individual  actuators. The  valves  are  actually passive  devices  that  are embedded  in  the  vine  robot  itself. So  all  that  work  is  done  upfront. When  you  manufacture  that  robot, that  every  pouch  has  its  own  valve  made, made  Alex's  life  a  living  ****.  Put  that  together. So  that  is  really  difficult, but  it  all  comes  on  the  body  of  the  vine  robot  itself. And  then  that  cap  that  you  saw  on that  particular  robot  has  batteries within  it  that  drive  a  motor that  push  it  forward  and  backward. Then  also  change  the  orientation  of  the  magnets that  switch  on  and  off the  passive  valves  that  are  already  embedded  in  a  body. So  I  hope  I  answered  your  question,  but  yeah, we  always  struggle  with, we  have  this  great  robot,  it  grows  forever. You  think  you  could  just  put  cables  down  the  center, but  you  can't  because  the  center  actually  goes forward  twice  as  fast  as  the  main  body. You  can  all  think  about  that  a  little  bit. The  geometry  of  it  is  such  that the  stuff  inside  goes  twice  this  part. So  it's  still  a  challenge. And  there's,  I'm  sure  there  can  be  lots more  brilliant  ideas  out  there  to  be  discovered on  how  do  we  handle  power  and  as we  increase  the  complexity.  Thanks  for  your  question. Any  questions? It  would  seem  from  you're  getting your  axes  of  rotation  aligned  in a  desirable  direction  that  you  could  rotate  about the  center  line  of  the  structure. So  with  my  elbow, I  if  I  wanted  to  go  to  the  now, I  have  to  rotate  it  this  way  and  go  this  way. Does  that  is  that  incorporated  in  some  way  or  do  you mean  for  the  vine  robots  or  perhaps  for  write  binary. But,  oh,  well,  so  most  of  our  vine  robots, like  they  steer  in  3D  by  having  three  actuators, a  line  on  the  side. And  so  any  two  of  those actuating  can  get  it  to  bend  in  a  particular  direction. I'm  not  sure  if  that's  what  you're  asking. Yeah. I  guess  I  didn't  feed  or  destroys the  ability  to  rotate  about  the  line. Yeah. I  see  what  you  mean.  Yeah.  No,  we haven't  we  haven't  tried  to  spin  it. We're  usually  trying  to  avoid  having  a,  having  a  spin. But  that  said,  if  we  have  the  distributed  actuation, the  hope  is  that  you  can  point  in any  direction  and  the  sensors  at  the  tip, if  it's  vision  data,  you  can  just  rotate. It  is  the  hope. But  there  might  be  some  utility. I  mean,  we  have  created caps  like  tip  mounts  that  have  grippers  on  them. And  then  they've  had  a  little  motor  that  just rotated  the  gripper  at  the  very  tip, like  a  little  wrist  so  that  we  can orient  the  gripper  the  right  way. Any  other  questions? I'm  just  curious. How  do  you  avoid  having  the  tip fall  off  as  you  retracting? Yeah. So  the  first  one  is  did  fall  off like  there  was  like  a  sock  and it  would  retract  and  it  would  come  off. The  newer  ones,  the  material  of  the  vine  robot  actually interdigitate  with  an  inside  piece  and  an  outside  piece. So  it's  like  the  outside  pieces  apart, like  locks  in  and  the  material  of the  vine  robot  kind  of  snakes  its  way  through  that. So  it's  like  physically  impossible  for  it  to  come  off. But  that  does  increase  friction and  that  means  we  need  more  pressure to  grow  and  these  sort  of  interlocking  things. It's  also  a  challenge  when  you have  various  steering  mechanisms, like  thicker  and  more  complex the  steering  mechanism  mechanism. And  the  more  likely  it's  good  to  get  gummed up  in  that  mechanism. But  that's  the  idea. That's  cool. Thanks. Thanks.  So  I'll  ask  the  last  question. Folks,  a  lot  about  the  mechanism, but  I'm  just  curious  if  you  want  to  reach  tracker. So  what's  a  protocol  to  wear? Just  a  vacuum  in  it  and  it's  adjust  it  back. Yeah.  Vacuum  actually  definitely  work  it  just  flat. Yeah. So  one  thing  you  will, you  typically  try  to  do  to  gracefully  retract  is  our  tip. Nouns  have  motors  in  it  that  kind  of steer  backwards  and  kind  of shoved  the  material  back  inside. But  you  can  also  do  it  by  just pulling  on  the  tail  which  is  inside. That  doesn't  work  great  when  there's  a  tip  mountain because  usually  there's  too  much  friction. But  other  groups  like  my  tallies  group  at UC  San  Diego  has  shown  that,  yeah, if  you  just  pull  on  the  tail  and there's  no  tip  mount,  it  will, it  might,  it  might  first  do  a  weird  thing, but  eventually  it  will  it  will  go  back  in. Okay.  Thank  you.  Thank  you. Okay.  So  let's  Dr.  or  camera.  Yeah. Okay.  Thank  you.