John Lathrop
@jplathrop.bsky.social
Robotics PhD student at Caltech
In addition, we show a new convergence proof for MCTS with deterministic dynamics and rewards whose exploration law is constant in depth and problem-independent but matches the convergence rate of alternate proofs that use a depth- and gap-dependent exploration law
December 4, 2024 at 7:52 PM
In addition, we show a new convergence proof for MCTS with deterministic dynamics and rewards whose exploration law is constant in depth and problem-independent but matches the convergence rate of alternate proofs that use a depth- and gap-dependent exploration law
Our work uses the local controllability Gramian to define natural motions for dynamical systems, speeding the convergence of the search.
December 4, 2024 at 7:50 PM
Our work uses the local controllability Gramian to define natural motions for dynamical systems, speeding the convergence of the search.
One of my labmates has a recent paper on information seeking behavior in POMDPs with MCTS:
www.science.org/doi/abs/10.1...
They study the trade-off of information seeking and managing safety
www.science.org/doi/abs/10.1...
They study the trade-off of information seeking and managing safety
Online tree-based planning for active spacecraft fault estimation and collision avoidance
A tree-based method plans in belief space to actively gather information and infer spacecraft faults while maintaining safety.
www.science.org
November 26, 2024 at 4:07 PM
One of my labmates has a recent paper on information seeking behavior in POMDPs with MCTS:
www.science.org/doi/abs/10.1...
They study the trade-off of information seeking and managing safety
www.science.org/doi/abs/10.1...
They study the trade-off of information seeking and managing safety