A-EMS: An Adaptive Emergency Management System for Autonomous Agents in Unforeseen Situations

Reinforcement learning agents are unable to respond effectively when faced with novel, out-of-distribution events until they have undergone a significant period of additional training. For lifelong learning agents, which cannot be simply taken offline during this period, suboptimal actions may be taken that can result in unacceptable outcomes. This paper presents the Autonomous Emergency Management System (A-EMS) - an online, data-driven, emergency-response method that aims to provide autonomous agents the ability to react to unexpected situations that are very different from those it has been trained or designed to address. The proposed approach devises a customized response to the unforeseen situation sequentially, by selecting actions that minimize the rate of increase of the reconstruction error from a variational auto-encoder. This optimization is achieved online in a data-efficient manner (on the order of 30 to 80 data-points) using a modified Bayesian optimization procedure. The potential of A-EMS is demonstrated through emergency situations devised in a simulated 3D car-driving application.

Automated summary

Reinforcement learning agents often struggle to perform well in novel situations, but this paper presents an online, data-driven emergency response method called A-EMS that aims to provide autonomous agents the ability to react to unknown scenarios. By selecting actions that minimize the rate of increase of reconstruction error, the proposed approach sequentially designs a customized response to unforeseen situations and achieves this optimization in an online, data-efficient manner using a modified Bayesian optimization procedure.