English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Robust reinforcement learning (RL) seeks to train policies that can performwell under environment perturbations or adversarial attacks. Existingapproaches typically assume that the space of possible perturbations remainsthe same across timesteps. However, in many settings, the space of possibleperturbations at a given timestep depends on past perturbations. We formallyintroduce temporally-coupled perturbations, presenting a novel challenge forexisting robust RL methods. To tackle this challenge, we propose GRAD, a novelgame-theoretic approach that treats the temporally-coupled robust RL problem asa partially-observable two-player zero-sum game. By finding an approximateequilibrium in this game, GRAD ensures the agent's robustness againsttemporally-coupled perturbations. Empirical experiments on a variety ofcontinuous control tasks demonstrate that our proposed approach exhibitssignificant robustness advantages compared to baselines against both standardand temporally-coupled attacks, in both state and action spaces.

Game-Theoretic Robust Reinforcement Learning Handles Temporally-Coupled Perturbations