# Controlled Sensing for Sequential Multihypothesis Testing with Controlled Markovian Observations and Non-Uniform Control Cost

arXiv: https://arxiv.org/abs/1310.1844 · HTML: https://mlsanity.com/paper/1310.1844/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 7 Oct 2013.

## Built on by (most cited first, 5 of 5)

- [Deep Multi-Agent Reinforcement Learning for Decentralized Active Hypothesis Testing](https://mlsanity.com/paper/2309.08477/) (2023)
- [Active Sampling of Multiple Sources for Sequential Estimation](https://mlsanity.com/paper/2208.05406/) (2022)
- [Searching for a Hidden Markov Anomaly over Multiple Processes](https://mlsanity.com/paper/2506.17108/) (2025)
- [Revealing After Overwriting: An Exponential POMDP OPE Lower Bound under History-Dependent Logging](https://mlsanity.com/paper/2610.05063/) (2026)
- [CASE: Cost-Aware Stopping for Efficient Long-Video Agents](https://mlsanity.com/paper/2610.05400/) (2026)

## Abstract (arXiv)

A new model for controlled sensing for multihypothesis testing is proposed and studied in the sequential setting. This new model, termed em controlled Markovian observation model, exhibits a more complicated memory structure in the controlled observations than existing models. In addition, instead of penalizing just the delay until the final decision time as in standard sequential hypothesis testing problems, a much more general cost structure is considered which entails accumulating the total control cost with respect to an arbitrary control cost function. An asymptotically optimal test is proposed for this new model and is shown to satisfy an em optimality condition formulated in terms of decision making risk. It is shown that the optimal causal control policy for the controlled sensing problem is self-tuning, in the sense of maximizing an inherent "inferential" reward simultaneously under every hypothesis, with the maximal value being the best possible corresponding to the case where the true hypothesis is known at the outset. Another test is also proposed to meet em distinctly predefined constraints on the various decision risks em non-asymptotically, while retaining asymptotic optimality.