A study of cognitive strategies for an autonomous search

• A mobile autonomous agent searching for an emitting source of unknown strength.
• Emitted substance transported by diffusion and advection.
• Three cognitive search strategies compared by simulations and using real data.
• Under persistent sensory cues, the distribution of search time is inverse Gaussian.
• Search time depends on the ratio between the search area and the sensing area.

Cognitive search is a collective term for search strategies based on information theoretic rewards required in sequential decision making under uncertainty. The paper presents a comparative study of cognitive search strategies for finding an emitting source of unknown strength using sparse sensing cues in the form of occasional non-zero sensor measurements. The study is cast in the context of an emitting source of particles transported by turbulent flow. The search algorithm, which sequentially estimates the source parameters and the reward function for motion control, has been implemented using the sequential Monte Carlo method. The distribution of the search time has been explained by the inverse Gaussian distribution.