Kavraki Lab

Physical and Biological Computing

Academic Background
I am a research programmer working in the Physical and Biological Computing Group at Rice University. I received my MS in Computer Science from the University of Nevada, Reno in 2011, and my BS in Computer and Information Engineering from the University of Nevada, Reno in 2009.

Sampling-based planners are usually thought of as offline planners which try to find a trajectory through some hard narrow passage. Typical problems include the need for possibly unbounded amounts of memory and the generation of trajectories that include a lot of redundant motions. Those problems are more pronounced if a robot has realistic motion constraints (e.g., bounded acceleration) and physical effects such are gravity and friction are modeled. This work describes a sampling-based kinodynamic motion planner that solves hard problems using replanning.

Falsification of LTL Safety Specifications using Sampling-based Algorithms

The contribution of this research is the development of a novel approach for falsifying LTL safety specifications for hybrid systems using decomposition-based planning algorithms. The approach extends the HyDICE approach so that complex specifications expressed using Linear Temporal Logic (LTL) can be falsified efficiently.