Safety-Critical Ergodic Exploration

Overview

Robots used for search and exploration need to cover a space thoroughly and never collide with anything along the way, and guaranteeing both at once is hard in tight, cluttered environments. This work generates search trajectories that are guaranteed to be safe by building discrete control barrier functions (DCBFs) directly into ergodic trajectory optimization.

How it works

• Ergodic trajectory optimization plans continuous exploratory trajectories whose time-averaged behavior matches a target information distribution, so the robot spends more time where there is more to find and is guaranteed to cover the space.
• Discrete control barrier functions encode collision avoidance as constraints that keep the robot inside the safe set at every time step of the plan.
• Combining the two produces trajectories that remain exploratory while being provably collision-free, for single robots and for teams of robots.

Experiments

We validated the approach in simulation and in real-world experiments on single- and multi-drone platforms, showing safe and effective exploration in cluttered environments.

My role

First author. [What you built: e.g. the DCBF formulation, codebase, drone experiments, multi-agent extension.]