Active Neural Localization

01/24/2018
by   Devendra Singh Chaplot, et al.
0

Localization is the problem of estimating the location of an autonomous agent from an observation and a map of the environment. Traditional methods of localization, which filter the belief based on the observations, are sub-optimal in the number of steps required, as they do not decide the actions taken by the agent. We propose "Active Neural Localizer", a fully differentiable neural network that learns to localize accurately and efficiently. The proposed model incorporates ideas of traditional filtering-based localization methods, by using a structured belief of the state with multiplicative interactions to propagate belief, and combines it with a policy model to localize accurately while minimizing the number of steps required for localization. Active Neural Localizer is trained end-to-end with reinforcement learning. We use a variety of simulation environments for our experiments which include random 2D mazes, random mazes in the Doom game engine and a photo-realistic environment in the Unreal game engine. The results on the 2D environments show the effectiveness of the learned policy in an idealistic setting while results on the 3D environments demonstrate the model's capability of learning the policy and perceptual model jointly from raw-pixel based RGB observations. We also show that a model trained on random textures in the Doom environment generalizes well to a photo-realistic office space environment in the Unreal engine.

READ FULL TEXT

page 5

page 7

page 9

page 14

research
11/20/2017

Teaching a Machine to Read Maps with Deep Reinforcement Learning

The ability to use a 2D map to navigate a complex 3D environment is quit...
research
05/19/2020

Differentiable Mapping Networks: Learning Structured Map Representations for Sparse Visual Localization

Mapping and localization, preferably from a small number of observations...
research
06/17/2019

Iterative Model-Based Reinforcement Learning Using Simulations in the Differentiable Neural Computer

We propose a lifelong learning architecture, the Neural Computer Agent (...
research
11/20/2020

Learning Synthetic to Real Transfer for Localization and Navigational Tasks

Autonomous navigation consists in an agent being able to navigate withou...
research
03/05/2019

Deep Active Localization

Active localization is the problem of generating robot actions that allo...
research
09/20/2022

Active Particle Filter Networks: Efficient Active Localization in Continuous Action Spaces and Large Maps

Accurate localization is a critical requirement for most robotic tasks. ...
research
04/10/2020

Learning to Visually Navigate in Photorealistic Environments Without any Supervision

Learning to navigate in a realistic setting where an agent must rely sol...

Please sign up or login with your details

Forgot password? Click here to reset