dynamic filter en

The$humanoid$robot$walking$ andavoidingobstacles.$ Title: Microsoft Word - dynamic_filter_en.docx Created Date: 5/30/2016 3:56:18 PM...

0 downloads 141 Views 942KB Size
     

  Bachelor  Thesis  in  Mathematics,  Physics  or  Computer  Science   “Dynamic  Filter  for  Walking  Motion  Correction“    

The  humanoid  robot  walking   and  avoiding  obstacles.  

Comparison  of  reference,  real  and  corrected   center  of  pressure  trajectories.  

  Abstract  

The  ability  of  autonomous  bipedal  walking  for  humanoid  robots  is  still  an  open  research  topic  for  the  robotics  community.  The   current   state   of   the   art   uses   extensive   model   simplifications   of   the   robot's   dynamics   in   the   form   of   inverted-­‐pendulum-­‐like   models.  The  planned  motion  using  a  simplified  model  can  lead  to  problems  during  its  execution  on  the  robot  due  to  neglected   effects   of   the   dynamics,   most   often   the   change   of   angular   momentum.   A   common   way   to   "smooth"   out   these   trajectories   is   achieved  by  a  so-­‐called  dynamic  filter  that  utilizes  a  given  whole-­‐body  dynamic  model  of  the  robot.  This  leads  to  much  smoother   and  more  reliable  motions  on  the  robot,  cf.  [1].    

Project   The  scope  of  the  project  is  suitable  for  a  beginner  software  practical  or  BSc  thesis.  The  project  covers:   • Implementation  of  a  dynamic  filter  as  proposed  by  „Kajita“  in  [2]   • Comparison  of  dynamic  filter  to  Newton-­‐Raphson  iteration   • Short  paper/BSc  thesis  summarizing  approaches  and  results    

Requirements  

 

• • •

Good  knowledge  of  Analysis  I  &  II,  Linear  Algebra  I   Knowledge  of  Numerical  Optimization  is  helpful   Programming  experience  with  Python  is  helpful  

References  

[1] „NMPC-­‐based   continuous   reactive   walking   pattern   generator“,   Naveau   et   al.,   IEEE   Robotics   and   Automation   Letters,   2016.   [2] „Biped  walking  pattern  generation  by  using  preview  control  of  zero-­‐moment  point“,  Kajita  et  al.,  in  the  proceedings  of   IEEE  International  Conference  on  Robotics  and  Automation,  2003.  

 

Contact  

Prof.  Dr.  Katja  Mombaur,  [email protected]­‐heidelberg.de   Manuel  Kudruss,  [email protected]­‐heidelberg.de     www.orb.uni-­‐hd.de