Learning Deep Generative Spatial Models for Mobile Robots

TL;DR

This paper introduces a probabilistic deep generative model for mobile robot environments that unifies low-level geometry and high-level semantics, enabling versatile spatial understanding and task execution.

Contribution

It presents a novel universal model based on Sum-Product Networks that jointly captures spatial geometry and semantics for mobile robots, surpassing traditional task-specific models.

Findings

Outperforms state-of-the-art models like GANs and SVMs in experiments

Capable of semantic classification, uncertainty estimation, and data generation

Successfully models environment at multiple abstraction levels

Abstract

We propose a new probabilistic framework that allows mobile robots to autonomously learn deep, generative models of their environments that span multiple levels of abstraction. Unlike traditional approaches that combine engineered models for low-level features, geometry, and semantics, our approach leverages recent advances in Sum-Product Networks (SPNs) and deep learning to learn a single, universal model of the robot's spatial environment. Our model is fully probabilistic and generative, and represents a joint distribution over spatial information ranging from low-level geometry to semantic interpretations. Once learned, it is capable of solving a wide range of tasks: from semantic classification of places, uncertainty estimation, and novelty detection, to generation of place appearances based on semantic information and prediction of missing data in partial observations. Experiments on laser-range data from a mobile robot show that the proposed universal model obtains performance superior to state-of-the-art models fine-tuned to one specific task, such as Generative Adversarial Networks (GANs) or SVMs.