DeepAI AI Chat
Log In Sign Up

Sampling-free obstacle gradients and reactive planning in Neural Radiance Fields (NeRF)

by   Michael Pantic, et al.
ETH Zurich

This work investigates the use of Neural implicit representations, specifically Neural Radiance Fields (NeRF), for geometrical queries and motion planning. We show that by adding the capacity to infer occupancy in a radius to a pre-trained NeRF, we are effectively learning an approximation to a Euclidean Signed Distance Field (ESDF). Using backward differentiation of the augmented network, we obtain an obstacle gradient that is integrated into an obstacle avoidance policy based on the Riemannian Motion Policies (RMP) framework. Thus, our findings allow for very fast sampling-free obstacle avoidance planning in the implicit representation.


page 1

page 2

page 3


Learning Obstacle Representations for Neural Motion Planning

Motion planning and obstacle avoidance is a key challenge in robotics ap...

Obstacle avoidance using raycasting and Riemannian Motion Policies at kHz rates for MAVs

In this paper, we present a novel method for using Riemannian Motion Pol...

Informed Circular Fields for Global Reactive Obstacle Avoidance of Robotic Manipulators

In this paper a global reactive motion planning framework for robotic ma...

Parallelised Diffeomorphic Sampling-based Motion Planning

We propose Parallelised Diffeomorphic Sampling-based Motion Planning (PD...

Log-GPIS-MOP: A Unified Representation for Mapping, Odometry and Planning

Whereas dedicated scene representations are required for each different ...

Creating Star Worlds – Modelling Concave Obstacles for Reactive Motion Planning

Motion planning methods like navigation functions and harmonic potential...

Reactive Motion Generation on Learned Riemannian Manifolds

In recent decades, advancements in motion learning have enabled robots t...