Acquisition-State Telemetry monitors information gathering
A new architecture evaluates ongoing information acquisition relative to declared task requirements.
Autonomous systems monitor physical kinematics, environmental states, and task completion outcomes. Current approaches do not monitor if the live process of information gathering is functioning properly. Systems can remain active without resolving mission requirements.
Acquisition-State Telemetry (AST) is a model free architecture. It evaluates ongoing information acquisition from observable evidence relative to declared task requirements. A domain interface maps raw observations into task resolution coordinates and normalized comparative representations.
A fixed causal monitor processes these representations. This monitor emits two continuous real time signals. One signal quantifies prospective progress across unresolved requirements. The other quantifies informational friction generated when recurrent evidence yields negligible informational gain.
The monitor was evaluated across eight diverse domains. These include search and rescue dialogue, spatial grid exploration, hierarchical capability acquisition, continuous robotic manipulation, and probabilistic intent estimation. The friction channel distinguished unproductive stagnation from purposeful exploration.
In closed loop reinforcement learning experiments, penalizing informational friction allowed agents to sustain high mission success. This occurred under early termination constraints. Unpenalized policies and baselines with matched operational costs failed to complete required tasks.
Key facts
- Acquisition-State Telemetry (AST) evaluates information acquisition from observable evidence.
- The system uses a fixed causal monitor to emit two continuous real time signals.
- The friction channel distinguishes unproductive stagnation from purposeful exploration.
- Penalizing informational friction helped agents sustain high mission success in experiments.