Abstract
Infotaxis introduced by Vergassola [1] serves as a robust search strategy for passive sensing in environments with sporadic cues and partial information. Infotaxis chooses the location to measure that maximizes the expected rate of information gain to localize on source cues. This contrasts with the Maximum A Posteriori (MAP) search strategy that chooses the most probable location in the search space. This research investigates the effect of varying beamwidth on active sensing Infotaxis, and compares the performance with MAP in 1D searches. The search process iterates three steps; choosing the search location to measure, actively scanning the location using a Bernoulli process, and performing Bayesian update on the state vector. For wider beam widths, the probability of detection decreases moving away from the main response axis. This modification of the measurement model revises the measurement and Bayesian update blocks in the search process. Monte Carlo simulations estimate the average search lengths across varying detection probabilities for a desired false alarm probability. Robotic experiments validate the simulation results. Robotic experiments include results comparing performances conditioned on missed first opportunity to detect the target. Simulations show Infotaxis outperforming MAP for high and low detection probabilities for medium beam widths. Infotaxis also recovers faster from missed first-detections compared to MAP for medium beam widths. Furthermore, Infotaxis exhibits information-maximizing behavior, confirming the strategy’s effectiveness relative to MAP.