Stanford · Robotics·2021AI & Autonomous Systems

Dijkstra & ROS for Perception-Based Path Planning

Planning paths directly over ROS perception data — trading efficiency for a simpler, more legible approach.

ROS offers many ways to process perception data. Running Dijkstra's algorithm along a desired path within that data wasn't necessarily the most computationally efficient choice, but it presented a simpler, more legible alternative — a trade-off between optimality and clarity that I still think about in system design.

Next project

Perception-to-Control for Occlusion Prediction

A perception-to-control framework that reasons about what might emerge from occluded regions.