動的時系列ワーピングによる指示条件付きナビゲーションの汎用評価
General Evaluation for Instruction Conditioned Navigation using Dynamic Time Warping
指示に従って経路を辿るナビゲーションエージェントの評価に動的時系列ワーピング(DTW)を導入し、参照経路との類似度を測るnDTWと成功経路に限定したSDTWを提案した論文。人間の類似度判断とよく一致し、強化学習の報酬としても性能を向上させる。
著者: Gabriel Ilharco, Vihan Jain, Alexander Ku, Eugene Ie, Jason Baldridge
分類: cs.RO, cs.AI, cs.CL
原文アブストラクト
In instruction conditioned navigation, agents interpret natural language and their surroundings to navigate through an environment. Datasets for studying this task typically contain pairs of these instructions and reference trajectories. Yet, most evaluation metrics used thus far fail to properly account for the latter, relying instead on insufficient similarity comparisons. We address fundamental flaws in previously used metrics and show how Dynamic Time Warping (DTW), a long known method of measuring similarity between two time series, can be used for evaluation of navigation agents. For such, we define the normalized Dynamic Time Warping (nDTW) metric, that softly penalizes deviations from the reference path, is naturally sensitive to the order of the nodes composing each path, is suited for both continuous and graph-based evaluations, and can be efficiently calculated. Further, we define SDTW, which constrains nDTW to only successful paths. We collect human similarity judgments for simulated paths and find nDTW correlates better with human rankings than all other metrics. We also demonstrate that using nDTW as a reward signal for Reinforcement Learning navigation agents improves their performance on both the Room-to-Room (R2R) and Room-for-Room (R4R) datasets. The R4R results in particular highlight the superiority of SDTW over previous success-constrained metrics.
