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

The increasing adoption of Reinforcement Learning in safety-critical systemsdomains such as autonomous vehicles, health, and aviation raises the need forensuring their safety. Existing safety mechanisms such as adversarial training,adversarial detection, and robust learning are not always adapted to alldisturbances in which the agent is deployed. Those disturbances include movingadversaries whose behavior can be unpredictable by the agent, and as a matterof fact harmful to its learning. Ensuring the safety of critical systems alsorequires methods that give formal guarantees on the behaviour of the agentevolving in a perturbed environment. It is therefore necessary to propose newsolutions adapted to the learning challenges faced by the agent. In this paper,first we generate adversarial agents that exhibit flaws in the agent's policyby presenting moving adversaries. Secondly, We use reward shaping and amodified Q-learning algorithm as defense mechanisms to improve the agent'spolicy when facing adversarial perturbations. Finally, probabilistic modelchecking is employed to evaluate the effectiveness of both mechanisms. We haveconducted experiments on a discrete grid world with a single agent facingnon-learning and learning adversaries. Our results show a diminution in thenumber of collisions between the agent and the adversaries. Probabilistic modelchecking provides lower and upper probabilistic bounds regarding the agent'ssafety in the adversarial environment.

On Assessing The Safety of Reinforcement Learning algorithms Using Formal Methods