Wildlife strikes in airports are increasing year over year. Current precautionary methods involve a lot of human involvement and shooting of wildlife as a safety precautionary method. We felt a need for bringing autonomy to current methods in this field. The fact that the problem was very challenging inspired us more to work on this problem.
What it does
We developed an algorithm for using Multiple UAVs for surveillance of the area and animal dispersal on and near airport runways. Higher-level control of decentralized multiple UAVs will cover the entire airport area by dividing the area into segments. During surveillance, if there is any animal movement, the UAVs detect that by using computer vision techniques. This information is sent into the network. Three of UAVs nearest to the animal collaborate and form a pattern such that they scares away animals and makes them disperse in the direction away from the runway and towards a specific trap. This logic is developed from wolves hunting patterns.
How we built it
We divided the whole problem into two sub-problems. Detection of animals and Deterrence of animals post detection.
Detection: The area of inspection needs to be covered periodically during the patrolling. Therefore, the entire area is divided into segments and we followed the sweep algorithm for area coverage in which one UAV monitors each segment. The algorithm is designed such that each point is inspected once in time, T. During its surveillance operation, if a UAV camera detects any animal movement it tags it and follows it.
Deterrence: The second part of the problem is dispersing animals away from the runway. One method that is prevailing currently is shooting the animals that get on runways but this will not preserve wildlife. Therefore, in addressing this problem, we came up with a novel technique which is inspired by wolves hunting pattern. Cristina Muro in his paper on "Wolf-pack (Canis lupus) hunting strategies emerge from simple rules in computational simulations" mentioned how wolves form a pattern surrounding their prey animals for chasing them down. However, our technique is designed to disperse them away from the runway. UAVs form a shape around animals within a safe distance in such a manner that animals will have only one direction to escape. These UAVs use high-frequency sounds or flashlights to disperse animals away from the runway. We simulated this behavior using MATLAB in which results are satisfying.
Challenges we ran into
Initially, the problem we choose looked simple, but when we went into details we came across many problems. The main challenge is finding logic for dispersing animals without having the need to shot them. Later, while coding the developed algorithm. It is hard to incorporate all the mechanisms of UAV after further literature review we came across behavior-based robotics which can be used for this scenario.
Accomplishments that we're proud of
Developing an idea in this field, which is challenging, within two days. Coming up with a new algorithm for tackling a particular type of problem and checking its behavior in the simulated environment.
What we learned
We learned about different types of airports. Various regulating agencies and rules we need to follow. Surveillance algorithms that can be used for large area coverage. We researched animal behavior during hunting and escaping from a predator.
What's next for Bio-Inspired Multi-UAV for Wildlife Control on Runways
- Path Planning of the multi-agent system using adaptive A* algorithm for avoiding runways and guiding animals to traps
- Simulation of these scenarios in ROS and GAZEBO
- UAVs battery cycle implementations into the surveillance logic
- Developing an algorithm for failsafe cases
- Safety of the airspace while using autonomous vehicles
- Hardware testing of UAVs with the bioinspired logic
- Training of Personnel to understand and control the system