All research

MPC for navigation among moving obstacles

Autonomous Systems Group, Imperial College London · Aug 2026 – Present

In progress

Model predictive controlMobile robots
Model predictive control · Mobile robotsmedia/imperial-mpc-1

Abstract

An MPC framework for efficient navigation of mobile robots in cluttered environments assumes the clutter holds still. This work extends it to obstacles that move, under the supervision of Dr. Johannes Köhler in the Autonomous Systems Group, Department of Mechanical Engineering, Imperial College London.

The problem

Navigating dense clutter is already a hard planning problem, and the usual formulations buy their guarantees by treating the obstacle set as fixed. Once obstacles move, the free space a trajectory was certified against has changed by the time the robot reaches it, and a controller that replans against a stale picture inherits that error rather than detecting it.