Placeholder image
Placeholder image


Research


While advancing robots towards full autonomy, it is important to minimize deleterious effects on human and infrastructure. To achieve this, I have been developing data-efficient robotic mapping techniques that capture uncertainty in dynamic environments. By modeling the nonlinear spatiotemporal relationships, these techniques can characterize the uncertainty in long-term and short-term patterns of occupancy, speed, and directions. Since these maps represent uncertainty, they can then be used for robust decision-making.
As part of the SAIL-Toyota Center For AI Research lead by Mykel Kochenderfer, Mac Schwager, and Marco Pavone, I work on safe interactions of autonomous systems.

Placeholder image



Uncertainty quantification in perception


Uncertainity in Occupancy

Temporal variations of spatial processes exhibit highly nonlinear patterns and modeling them is vital in many disciplines. For instance, robots operating in dynamic environments demand richer information for safer and robust path planning. In order to model these spatiotemporal phenomena, I develop and utilize theory in reproducing kernel Hilbert space (RKHS) and deep learning. I am mainly interested in modeling and propagating the uncertainty of dynamic environments and therefore I frequently use Bayesian modeling techniques.

  • Scalable Gaussian process with stochastic variational inference
  • Bayesian Hilbert maps (BHM)
  • Under construction


Uncertainity in Directions

We model the uncertainty of directions using a mixture of von Mises distibutions.

  • Directional grid maps
  • TBC

Safe Decision-Making under Uncertainty


State uncertainty

We use partially observable Markov decision processes (POMDPs) for planning.

  • Under construction

DA


Under Construction

Under Construction

  • Computer Vision
  • Human Factors in Human-Computer Engineering
  • Image Processing
  • Kernel Methods
  • TBC


---