When (or if) Amazon and other companies begin delivering items via drones, there will be little room for error. Their drones must be equipped to deal with any unexpected obstacles that come their way, whether that’s a construction crane or strong winds and rain.
MIT researchers revealed a set of algorithms today that could help drones avoid catastrophe with advance planning. The first step takes place before the drone even takes to the air, when it computes a few potential routes out of the many it could take, allowing it to determine different situations it might encounter along its journey. That has the added benefit of freeing up computing time while the drone is in the air so it can focus on staying aloft.
“Imagine a huge tree of possibilities, and a large chunk of leaves collapses to one leaf, and you end up with maybe 10 leaves instead of a million leaves,” researcher Ali-akbar Agha-mohammadi said in a release. “Then you can … let this run offline for say, half an hour, and map a large environment, and accurately predict the collision and failure probabilities on different routes.”
Then, while in the air, the drone tracks vital information like fuel levels, sensor health and the state of its cameras and other equipment in real time. It notices when it is running low on battery power, it can fly itself to a charging station instead of just falling out of the air or landing in a random location. In tests, drones equipped with the MIT monitoring system failed less often when faced with a variety of external challenges than drones that did not monitor their vitals.
The perfect delivery drone will be fail-proof. Sensors will prevent it from running into buildings or other obstacles and allow it to perfectly gauge where to drop a package. But even the best-planned drone route can run into unexpected hazards like strong winds that cause the aircraft to burn through a battery charge faster than expected. Proactive monitoring tools will without a doubt be necessary.