TeleAgent: Agent-Based Network Status Dialogue and Autonomous Recovery for Telepresence Human-Robot Interaction

Han HU, Qijia Wang, He Chen

Keywords: Human-Robot Interaction (HRI), Explainable AI (XAI) in Robotics, Telepresence, Agent-Based Systems, Network-Aware Robotics

TL;DR: We present TeleAgent, an agent that uses an onboard language model to translate network events into a clear dialogue for the user, improving resilience in robot telepresence over unreliable WiFi.

Abstract: The quality of experience for telepresence robots is fundamentally crippled by unreliable wireless networks, which cause abrupt connection freezes that leave remote operators confused, disoriented, and frustrated. To address this, we introduce TeleAgent, an agent-based framework powered by an onboard small language model (SLM) designed to facilitate a clear network status dialogue and provide robust autonomous recovery. The framework provides dual, context-aware strategies: 1) for predictable events like WiFi handovers, it initiates a proactive dialogue, explaining the disruption, deferring action until user confirmation, and providing timely feedback upon completion. 2) For unexpected catastrophic WiFi failures, it triggers an autonomous recovery mechanism where the robot takes over navigation, autonomously seeks a location with better signal quality, and maintains critical updates with the operator via a secondary cellular network. By using an SLM to interpret network logs and mediate these complex state transitions, TeleAgent transforms the telepresence experience, rebuilding the operator's sense of control and trust through transparent dialogue and proactive problem-solving.

Submission Number: 12