Sentinel Guard: Rover + AI for Rapid Wildlife Identification

Problem. Rangers need timely, accurate species identification in mixed herds and low-visibility terrain. Manual patrols miss events and yield sparse data. We propose a low-cost, mobile system built on a robot that detects and classifies target savanna species and triggers alerts for rescue and conservation. Why it matters / Results. Faster, species-aware alerts can reduce poaching risk, speed response to injuries, and generate geotagged counts for population monitoring. We aim for ≥75% detection accuracy across five species (elephant, zebra, giraffe, lion, wildebeest) with <300 ms inference suitable for field use. Our rover will have four wheels and use distance and optical sensors, plus a camera mounted on top to help detect and identify animals. The main computer (Brain) will control movement and collect data from all sensors.

We will build a dataset using open wildlife images and pictures we collect ourselves. Then we will train an AI model (a type of neural network) using the Supercomputing Challenge resources. The model will be made smaller and faster so it can run efficiently on the rover.

When the rover sees an animal, it will send the image to the computer, identify the species, and show a confidence level. If it’s a strong match, the rover will record the time and place, save a short video clip, and send an alert.

We will test the system by checking how accurate and fast it is in different lighting and distances. We will also make sure the rover keeps a safe distance from wildlife and that data is only used for science and conservation.