#10 Research meet-up: Wildlife protection analytics through software used by 1,000+ parks

Lily Xu of Harvard University presented PAWS, the Protection Assistant for Wildlife Security, which uses analytics to support patrol planning under uncertainty. The research has included field tests in Cambodia and integration with SMART, a protected-area management system used by more than 1,000 wildlife parks worldwide.

Why is wildlife patrol planning uncertain?

Wildlife poaching contributes to a multi-billion-dollar illegal wildlife trade and threatens species in protected areas.

Rangers need to decide where limited patrol resources should be deployed without knowing exactly where future poaching activity will occur.

PAWS was developed to support these decisions by learning from available information and using it to inform future patrol planning.

How does the approach work?

The research includes methods from multi-armed bandits and reinforcement learning.

A multi-armed bandit method balances using strategies that have worked before with testing alternatives that could provide better information. Reinforcement learning similarly uses feedback from previous decisions to improve future choices.

Xu's work also considers robust sequential decision-making, namely making a series of decisions while accounting for uncertainty in what may happen next.

Importantly, the research questions were shaped by challenges encountered in field settings rather than being developed only as theoretical problems.

How did the project move beyond a prototype?

During the meet-up, we discussed field tests in Cambodia and the integration of PAWS with SMART, software used to manage protected areas.

Because SMART is used by more than 1,000 wildlife parks worldwide, the integration provides a route for analytical methods developed through the research to become accessible within existing conservation-management systems.