Zexin Li

New attack method targets deep reinforcement learning agents

Researchers have developed a new trajectory-level adversarial attack method designed to target deep reinforcement learning (DRL) agents. This approach optimizes a sequence of perturbations over a receding horizon, utili…