TODOs for journal submission

[hjsuh94]

Writing

Reviewer 2

• [ ] Clarify equivalence of randomized and analytic smoothing
• [ ] Clarify interior point methods warm starting
• [ ] Clarify timestep vs. Anitescu's relaxation
• [ ] Clarify smoothing coming from Anitescu vs. smoothing
• [ ] Forward mode Autodiff vs. Analytic gradients
• [x] Get rid of statements about B.

Reviewer 6

• [ ] Full dynamics gradients are obtained by chaining dJi/dq and dphi_i/dq, which has collision-geometry discontinuities.
• [ ] Does this scheme differ from relaxing the complementary slackness constraint?
• [ ] Table of hyperparameters
• [ ] Explain more on contact sampling
• [ ] Code release

Reviewer 7

• [ ] Section 1, summary of how each sections are connected.
• [ ] Table 1, show expressions?
• [ ] Our dynamics are not convex, we just approximate it that way.
• [ ] Physical intuition for quasi-dynamic and convexification.
• [ ] Where does smoothed contact mechanics fit in algorithms 2,3,4?
• [ ] Discuss limitations
• [ ] Include a few more references

Reviewer 8

• [ ] Comparison with RL?
• [ ] Validate on real robot

Experiments

Reviewer 2

• [ ] Verify validity of quasidynamic assumption on sim2drake
  • [ ] Planar Hand
  • [ ] Planar Pushing
  • [ ] Allegro Hand
  • [ ] Iiwa-Bimanual setup
  • [ ] Allegro Pen
  • [ ] Allegro Book
  • [ ] Allegro Door
• [ ] Verify validity of quasidynamic assumption on open-loop hardware execution on Bimanual Iiwa

[hjsuh94]

Boilerplate Checklist

• [x] Planar Pushing
• [x] Planar Hand
• [x] Allegro Hand
• [x] Allegro Pen
• [ ] Allegro Plate
• [x] Allegro Door

Drake2Sim Checklist

• [x] Planar Pushing
• [x] Planar Hand
• [x] Allegro Hand
• [x] Allegro Pen
• [ ] Allegro Plate
• [x] Allegro Door

Sim2Real Checklist

• [x] Running iiwa bimanual trajectories on hardware

[pangtao22]

Video:

1. highlights: Robot demo -> 6 sim demos -> allegro hand contact force demo.
2. Sim2Drake (Need a better name?) Show successes and failures. (@hjsuh94 will generate some videos, where reference is overlayed on the MBP trajectory) We need to think more about how to show failures? Think together with the plots for sim2drake results in the paper.
3. More robot demo Video recording and LCM side by side. (@pangtao22 will generate LCM videos)
4. The rest of the RRT results.

[pangtao-BDAI]

Resbumitted!