A Critical View of Vision-Based Long-Term Dynamics Prediction Under Environment Misalignment

TL;DR

This paper critically examines the limitations of vision-based long-term dynamics prediction models under environment misalignment, introduces challenging datasets, and proposes a mitigation strategy to improve robustness across domains and contexts.

Contribution

It introduces four new datasets for evaluating environment misalignment in dynamics prediction and proposes a mitigation approach that significantly reduces the impact of cross-domain challenges.

Findings

RPCIN performs well on standard conditions but struggles under environment misalignment.

The proposed mitigation strategy dramatically alleviates cross-domain prediction issues.

The datasets reveal specific weaknesses of current vision-based models in challenging environments.

Abstract

Dynamics prediction, which is the problem of predicting future states of scene objects based on current and prior states, is drawing increasing attention as an instance of learning physics. To solve this problem, Region Proposal Convolutional Interaction Network (RPCIN), a vision-based model, was proposed and achieved state-of-the-art performance in long-term prediction. RPCIN only takes raw images and simple object descriptions, such as the bounding box and segmentation mask of each object, as input. However, despite its success, the model's capability can be compromised under conditions of environment misalignment. In this paper, we investigate two challenging conditions for environment misalignment: Cross-Domain and Cross-Context by proposing four datasets that are designed for these challenges: SimB-Border, SimB-Split, BlenB-Border, and BlenB-Split. The datasets cover two domains and two contexts. Using RPCIN as a probe, experiments conducted on the combinations of the proposed datasets reveal potential weaknesses of the vision-based long-term dynamics prediction model. Furthermore, we propose a promising direction to mitigate the Cross-Domain challenge and provide concrete evidence supporting such a direction, which provides dramatic alleviation of the challenge on the proposed datasets.