NCSRR 5 week plan.
NCSRR Biweekly Plan
Started on Monday July 14
Thursday Jul 17
- Prepare and give kickoff talk
- Setup latest Opensim/Simbody dev environment
- Read some papers to get an idea of scale limits in NLP problems with IPOPT and/or SNOPT
- Complete all the orientation stuff for Stanford
- Complete initial test run of inverted pendulum on cart problem with 10,000 nodes in IPOPT
- Talk with Jack Wang about using his existing code (if I can use the code, below plans may be altered)
- Discuss direct collocation vs shooting with Ton and others
Monday Jul 21
- Test the limits of scaling in the inverted pendulum code, make computational speed improvements.
- Build (or obtain) a basic torque driven 9/10 DoF model in Opensim
Thursday Jul 24
- Implement gain scheduled black box control structure for the planar walking model in Opensim
Monday Jul 28
- Start development of direct collocation code in C++ (port stuff from my current Python code) using numerical functionality from Simbody
Thursday Jul 31
- Complete initial direct collocation implementation
Monday Aug 4
- Iterate on direct collocation solutions: testing initial guesses, minimal data sets, # discretization nodes, improve computational speed, etc.
- Implement IPOPT + paradiso on an AWS instance
Thursday Aug 7
- Either polish things that work or fix things that don't.
Monday Aug 11
- Either polish things more that work or fix things that don't.
Thursday Aug 14
- Prepare and present results and findings