EpEx – Episodic Memory Reconstruction for UAS Behavior Explanation

Intro

Accelint is advancing transparency and trust in unmanned aerial system (UAS) autonomy with EpEx (Episodic Memory Reconstruction for UAS Behavior Explanation). Developed under contract with the Air Force Research Laboratory's Human Performance Wing, EpEx reconstructs and visualizes the decision-making process of autonomous UAS, enabling operators to better understand autonomous behavior and supporting easier testing and fielding of autonomous systems.

As UAS autonomy increases, operators often cannot see why an aircraft acted in a particular way during missions. The primary challenge is the difference between information available to the UAS at the time of a decision and information available to the operator during after-action review. This gap in visibility complicates testing, validation, and operational deployment of autonomous systems, particularly as the Air Force seeks to expand the range of missions over which autonomous systems can operate.

Solution

Accelint is developing EpEx to integrate and correlate large volumes of internal UAS records with external observations, identifying and characterizing critical UAS decisions. The system builds on prior AFRL-funded work in adaptive human-machine interfaces and computational models of human episodic memory, incorporating new algorithms for episodic memory analysis and visualization. Key capabilities include:

• Episodic memory analysis and visualization, translating machine actions into clear, human-readable explanations.

• Integration of multi-source mission data, correlating sensor inputs, internal system states, and environmental conditions.

• Adaptive user interfaces using Accelint’s Weaver display manager, delivering tailored information aligned with operator workload and task relevance.

‍

EpEx enables operators to better understand UAS decision-making when the UAS is operating in autonomous modes, supporting the Air Force's objectives for testing, fielding, and expanding the operational range of autonomous systems.