time is up
Ask or upvote questions you have for the speakers below. Please indicate Panel A, B, or C.
Ideas
Pros and cons
Votes
How can you guarantee safety throughout the layers of a robot?
It is fairly common to see that each layer of a system, e.g. motion planning algorithm, controlle...
more
3
Vote
Ali
Lukas
SiQi
How can data be used to verify the assumptions? (Panel B)
1
Vote
Jacopo
One of the key factors that led to the success of machine-learning-oriented methods like Deep RL in recent years is their ability for generalization. On the other hand, the theoretical guarantees of model-based control methods are usually based on accuracy of the models or their structural assump...
more
by Jason
(Panel B)
by Jason
0
Vote
Given that Vision-based localization is a big part of real-world autonomy and it is susceptible to drift or wrong associations, how does one talk about safety ? Especially, when the confidence reported itself can be wrong.
by Abhishek Goudar
0
Vote
I would be interested in your opinion on what we can acctually call a "safe" system. Is it enough for our system not to actively cause accidents? or does it need to be able to actively participate in preventing accidents? And if so. to what degree does the system need that capability?
0
Vote
[Panel C] How inexpensive would a (physical) safe robotics platform need to be to be useful to researchers?
by Jacob
In order to validate safety in the real world, it is necessary to run algorithms on a real robot,...
more
0
Vote
Reinforcement learning algorithm is typically trial-and-error. How can you provide safe assurance (without error) while also train an agent effectively?
by Zikang
0
Vote