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

Deep Reinforcement Learning (DRL) has received a lot of attention from theresearch community in recent years. As the technology moves away from gameplaying to practical contexts, such as autonomous vehicles and robotics, it iscrucial to evaluate the quality of DRL agents. In this paper, we propose asearch-based approach to test such agents. Our approach, implemented in a toolcalled Indago, trains a classifier on failure and non-failure environment(i.e., pass) configurations resulting from the DRL training process. Theclassifier is used at testing time as a surrogate model for the DRL agentexecution in the environment, predicting the extent to which a givenenvironment configuration induces a failure of the DRL agent under test. Thefailure prediction acts as a fitness function, guiding the generation towardsfailure environment configurations, while saving computation time by deferringthe execution of the DRL agent in the environment to those configurations thatare more likely to expose failures. Experimental results show that oursearch-based approach finds 50% more failures of the DRL agent thanstate-of-the-art techniques. Moreover, such failures are, on average, 78% morediverse; similarly, the behaviors of the DRL agent induced by failureconfigurations are 74% more diverse.

Testing of Deep Reinforcement Learning Agents with Surrogate Models