Robotics & Perception

My work relates to giving machines the ability to perceive, reason, and act in the world.

Algorithm Tradeoffs

In robotics, there is no "best" algorithm, only the right set of tradeoffs for the constraint manifold.

Comparison between Algorithm A (Rapid, Approximate) and Algorithm B (Precise, Heavy).

Planner Comparison

Different planning algorithms suit different problem structures. Here's my practical guide:

| Algorithm | Best For | Tradeoff | |-----------|----------|----------| | A* | Grid/graph search, known maps | Optimal but exponential in high dimensions | | RRT* | High-dimensional spaces, manipulation | Asymptotically optimal, slow convergence | | PRM | Multi-query scenarios, static environments | Preprocessing cost, struggles with narrow passages | | MPC | Dynamic systems, real-time control | Computational cost, requires good model |

When I Use Each

A*: When the state space is discrete and reasonably sized. Good for 2D navigation on grids.
RRT*: Robot arm planning where the configuration space is high-dimensional. Willing to trade computation for solution quality.
PRM: When multiple queries will be made in the same environment. Preprocessing amortizes over queries.
MPC: When dynamics matter and we need to reason about the future trajectory. Good for smooth, constraint-aware motion.

Multi-Agent Planning

My thesis work focuses on coordinating multiple robots in shared spaces. The challenge: scaling from single-agent planning to dozens of agents without exponential blowup.

Key research questions: