APF-PF: Probabilistic Depth Perception for 3D Reactive Obstacle Avoidance

TL;DR

This paper introduces a probabilistic depth perception framework using particle filters combined with artificial potential fields for real-time 3D obstacle avoidance in UAVs, effectively handling partial observability and noisy data.

Contribution

It presents a novel integration of particle filtering with APF for obstacle avoidance, enabling robust, real-time 3D navigation under uncertain perception conditions.

Findings

Demonstrated robustness in obstacle avoidance with noisy, partial data.

Validated real-time performance onboard a quadrotor UAV.

Showed improved navigation reliability in complex environments.

Abstract

This paper proposes a framework for 3D obstacle avoidance in the presence of partial observability of environment obstacles. The method focuses on the utility of the Artificial Potential Function (APF) controller in a practical setting where noisy and incomplete information about the proximity is inevitable. We propose a Particle Filter (PF) approach to estimate potential obstacle locations in an input depth image stream. The probable candidates are then used to generate an action that maneuvers the robot towards the negative gradient of potential at each time instant. Rigorous experimental validation on a quadrotor UAV highlights the robustness and reliability of the method when robot's sensitivity to incorrect perception information can be concerning. The proposed perception and control stack is run onboard the UAV, demonstrating the computational feasibility for real-time applications and agile robots.