You were asked to list the specific questions you would like answered during this workshop.
In no particular order, you responded with:
How should we handle modeling error and poor torque control?
◉
What about hydraulic control and hardware limits?
◉
Model-based control?
◉
What's the best way to do force control on Atlas?
◉
How much can we gain from sys id / joint level tuning?
◉
What experience do we have with inverse dynamics approaches on biped robots and how is the performance affected by model and noise uncertainty, imperfect state estimation and a limited control bandwidth?
◉
What is a trajectory representation that bridges the gap between a) providing an intuitive way of expressing costs on the task of interest and b) hinting on a good choice of gains for control?
◉
What are the differences between approaches at attending organizations?
◉
Is there a controller class between heuristics (Raibert’s controllers) controllers (energy efficient/high performance but single task/robot oriented) and model-based controllers (Highly versatile but low energy efficiency)?
◉
Is state estimation a limitation to performance and reliability?
◉
From the DRC experience, is Atlas a suitable humanoid robot for disaster response? What is it missing?
◉
What about the World Wide Humanoid Robots Olympics Games?
◉
Insertion of competition in humanoid research: speed up or slow down research?
