Meet PAWS – the elephant’s new protector. This new machine-learning and game-theory system can predict where poachers are likely to strike.
Poachers kill around 27,000 African elephants each year. This represents around 8% of their total population and, if unchecked, could drive them to extinction within the next decade.
The problem already protected wildlife zones have is how to allocate their finite staff across a huge area.
Speaking to spectrum.ieee.org, Milind Tambe – the computer scientist at the University of Southern California, in Los Angeles behind the project explained that the problem is two-fold; “can you predict where poaching will happen? And can you [target] your patrols so that they’re unpredictable, so that the poachers don’t know the rangers are coming?”.
This is problem is what PAWS – Protection Assistant for Wildlife Security – aims to solve.
The machine-learning algorithm can process data from past patrols to predict where poaching is likely to occur now.
The system can use game-theory models to then generate new, unpredictable, routes that the rangers can patrol.
So far the system has been tested in the field in both Uganda and Malaysia – with positive results.
In the Uganda trial; the system highlighted two zones which were not routinely visited as being at risk. Rangers dispatched to inspect the areas discovered snares and other tell-tail signs of poaching.
Across the eight-month trial, patrols found high-probability areas flagged by the machine algorithm as having ten times as much poaching activity as the low-probability areas.
The team hope to expand into China and Cambodia this year; and integrate the tool into SMART – an existing tool that conservation agencies use to manage existing patrol data.