Issue link: https://iconnect007.uberflip.com/i/1172746
30 SMT007 MAGAZINE I OCTOBER 2019 that we can operate a burger alarm system for $15–20 dollars per month, which is well within the range of any park in the world. We think it's a great solution, and we're trying to scale it. We just returned from a trip to China with Intel to visit the tinyGO factory—a subsidiary of WeiBu, an original design manufacturer (ODM) that specializes in building devices with their VPUs, including the Myriad™ 2 and X—and build the next 1,000 Trail- Guard AI devices. In many places, but par- ticularly Africa, the parks are vast, and the wildlife is hard to protect with lim- ited personnel. The Seren- geti is the same size as the state of Maryland in the U.S., which is about 17,000 square kilometers, but there are only about 150 rangers on the force to protect it. Imagine 150 guards try- ing to protect the wildlife of the state of Maryland; it's impossible. How- ever, what if those 150 guards could act more like 1,500 guards? TrailGuard becomes a "force multiplier and an intelligence multiplier" by reorganizing the protected area system from rangers at guard posts to rang- ers who are involved in "event-triggered rapid response." We are turning them into rapid- response teams based on TrailGuard alerts; they don't have to conduct random searches. The images come into a program, and there are several different groups. The late Paul Allen's company Vulcan has created a domain aware- ness system for park management that Trail- Guard feeds right into and pops up as an alert; the rangers can see the image of the intruder on the screen. Similarly, ESRI—the world's leading manu- facturer of GIS software—also has a protected area management domain awareness system, and we integrate with both. We have an even simpler system for parks that don't have the budget to afford those. Now, we have the capa- bility that when there's an intruder, an alert is sent in under two minutes to the headquar- ters—either via GSM or LoRa—and somebody at the other end can see that there is an image and decide whether to send out a ranger team, which works well. After talking with a lot of protected area manag- ers, wildlife enforcement people, and military and ex-military we found out that no matter how big a park is, or what kind of vegetation or terrain it has, the basic problem is that there are essentially 10 major access routes into a park that poach- ers use; how they enter is not random. It could be influenced by guard posts, roads, villages, and most importantly, where the animals they're after con- centrate. If you have that knowledge, then maybe you can create a program where it predicts the most logical entry points, and you can put your TrailGuard devices at those "chokepoints." Again, experts we talked to said that there should be no more than 10 major chokepoints per park, which experience about 80% of the poaching traffic. If you can inter- dict those poachers, that will have a tremen- dous effect on wildlife. So, our hypothesis is if we can identify those 10 major access routes to any park and have 10 satellite modems put there—assuming that there's no GSM—and have our TrailGuard units in a star network around the satellite modems communicating with it, then we'll effectively and cheaply cover those access points and be able to provide that information for rangers to take action. I went a little off-topic, but I am so excited about this that it is hard to contain myself Three TrailGuard AI units. The left version shows how the unit is camouflaged using local detritus to make it even more cryptic when deployed in the field.