Indeed, when asked what to improve, there was a clear
desire to include more reactive control to make up for
deficiencies with the sense-plan-act model. The challenge
participants were also not content with the abstraction level
that prepackaged software solutions like MoveIt! provided.
On the one hand, teams wished for the ability to model their
robot hardware more easily and have simple ways to provide
tasks and constraints, much like the way MoveIt! [18] and
OpenRave [19] provide. On the other hand, the tools are not
perfect yet, are difficult to debug, and have a high learning
curve should a solution require the team to make changes
“under the hood” of such tools. A possible solution here might
be not only to continue to improve these tools, but abstract
their lower-level functionality into a higher level language,
making their inner workings more accessible, and making it
easier to attach arbitrary sensing, reactive controllers and logic
to the trajectories they generate.