As the human population keeps booming, our actions are pushing life on our shared planet towards mass extinction. Human activity is causing extinctions at the rate of centuries versus the millions of years that take for natural extinctions. Poaching is playing a huge role in this human lead devastation and the numbers for some species are plummeting to critical/unsustainable levels. According to the Lindbergh Foundation:
To combat poaching and to rebuild the populations of the endangered species, some countries have set up protected wildlife reserves and conservation agencies, tasked with defending these large reserves. However, these efforts have not been good enough to counter poaching. Poachers have gone high-tech, using night-vision goggles and GPS to kill elephants and rhinos for their tusks and horns. So when patrolling is to be used as the primary method of securing large reserves, it is necessary to create “Smart” patrols.
Thanks to AI, now there is a way to tackle this problem in a different way. AI is being used for getting ahead in the “Patroller vs Poacher” game and to take more preventative measures to poaching.
Let’s look at two programs created to tackle poachers the “Smart Way”:
- PAWS (Protection Assistant for Wildlife Security): An approach based on Game Theory and Decision Trees.
- Air Shepherd, a program created by the Lindbergh Foundation: An approach based on drones, Microsoft Azure and Deep Learning (Image Processing).
Reservations and lands where protected species live are often too large to be patrolled in their entirety all the time, and this gives an edge to the poachers.
PAWS uses machine learning for predicting where the poachers are going to strike next and then integrates these predictions with game-theoretic reasoning for designing patrolling routes. Information on the local area and data collected on previous patrolling and poaching activities are used as input to their model, which then calculates a randomized patrolling strategy (in the form of a set of patrol routes and their probabilities). PAWS then suggests patrol routes sampled from this strategy to the patrollers.
Both their laboratory experiments and real-world field tests, have shown that using ensembles of decision trees is the best ML approach. They consider decision trees to be superior also because they are a “white-box” solution – domain experts (e.g., conservationists, park rangers) can easily look at the learned model (which is just a set of logical rules) and determine whether the decision tree is making reasonable inferences about how poachers behave.
The interesting aspect of PAWS is the use of Computational Game Theory. They have looked at the problem in the form of a “Patroller vs Poacher” game. This helps them to get a better understanding of strategies poachers may adopt and to stay ahead in the game via AI simulations.
For the readers interested in details, PAWS currently utilizes the SUQR behavioral model described in the following paper: “Analyzing the Effectiveness of Adversary Modeling in Security Games“.
2. Air Shepherd
Air Shepherd is a drone program created to help save African elephants and rhinos from illegal poaching and imminent extinction.
Air Shepherd uses surveillance drones that fly silently in the air and have infrared sensors, tailored for the anti-poaching mission, to locate poachers at night. These drones radio their location to rangers in real time. In this way, when potential poachers are spotted, nearby rangers can come to the rescue before it’s too late. You might ask, “Why at night?” Rangers own the daylight. But the tables turn at night. Almost all poaching of big animals happens overnight. Poachers gather intelligence about their prey’s whereabouts, and then – under cover of dark – they move in for the kill.
The step by step process:
Air Shepherd is actively incubating new technologies to stay ahead of poachers. Some of their investment areas include: laser and radar-based sensors, big data analytics, advanced battery technologies, neural networks and object recognition in images:
Air Shepherd has been recently testing a tool called SPOT for automatically detecting poachers in long wave thermal infrared UAV videos using Microsoft Azure.
Here is how their Deep Learning workflow in AirSim-W – a simulation environment designed specifically for the domain of wildlife conservation – looks like:
You can find the full details in this paper: AirSim-W: A Simulation Environment for Wildlife Conservation with UAVs
An example of their success story in their own words: “Flying in one area where as many as 19 rhinos were killed each month, there have been no deaths – for six months.”
Poachers also keep throwing new challenges to these organizations by adopting new strategies. For example, from the latest trends:
We have seen two different approaches to counter the problem of poaching. They are not perfect solutions and both initiatives have many open issues to address e.g., problems due to lack of connectivity for sending data in real-time. Regardless of the shortcomings, helping endangered species is a great use case of AI. And the main reasons why I decided to do this write-up:
- Awareness regarding these issues is important – nature is not separate from us.
- Very interesting and creative use cases of AI and Microsoft Azure (AI for Good)
I’ll end this post by leaving you with this thought:
“There is still time. And time means hope – but not without action.”
Thanks for reading! If you want to be notified when I write something new, you can sign up to the low volume mailing list.
References and Further Reading
PAWS: Protection Assistant for Wildlife Security
Machine Learning For Wildlife Conservation With UAVs