Self-Supervised JEPA-based World Models for LiDAR Occupancy Completion and Forecasting

TL;DR

This paper introduces AD-LiST-JEPA, a self-supervised world model for autonomous driving that predicts future LiDAR data and improves occupancy completion and forecasting without requiring labeled data.

Contribution

It presents a novel JEPA-based framework for learning scalable, self-supervised world models specifically for LiDAR data in autonomous driving.

Findings

Pretrained encoder improves occupancy completion accuracy.

JEPA-based model effectively captures spatiotemporal dynamics.

Self-supervised learning reduces dependence on labeled datasets.

Abstract

Autonomous driving, as an agent operating in the physical world, requires the fundamental capability to build \textit{world models} that capture how the environment evolves spatiotemporally in order to support long-term planning. At the same time, scalability demands learning such models in a self-supervised manner; \textit{joint-embedding predictive architecture (JEPA)} enables learning world models via leveraging large volumes of unlabeled data without relying on expensive human annotations. In this paper, we propose \textbf{AD-LiST-JEPA}, a self-supervised world model for autonomous driving that predicts future spatiotemporal evolution from LiDAR data using a JEPA framework. We evaluate the quality of the learned representations through a downstream LiDAR-based occupancy completion and forecasting (OCF) task, which jointly assesses perception and prediction. Proof of concept experiments show better OCF performance with pretrained encoder after JEPA-based world model learning.