Zoo animals go wild for artificial intelligence

From using moon rovers that encourage predators to hunt and forage in packs, to applying state of the art algorithms to try and understand the facial expressions of Sumatran Orangutans, artificial intelligence and robotics are giving zookeepers and researchers new insights into animal welfare.

Developed by University of Adelaide researchers, with support from Zoos South Australia’s research team, these hi-tech approaches are being tested at the Adelaide Zoo and Monarto Safari Park, giving keepers real time, unprecedented access to animal behaviour in support of animal welfare.

For Monarto Safari Park’s African Painted Dogs, this unique approach is allowing them to engage in a new kind of naturalistic enrichment that mirrors the thrill of the chase.

“We are using a semi-autonomous rover – an advanced remotely controlled vehicle – to carry a food carcass through the Painted Dogs habitat, simulating the movement of live prey. This novel approach prompts the dogs to interact, communicate, and move together as they would in the wild,” said the University of Adelaide’s Dr Xin (Vernon)Yuan, Senior Research Associate at the School of Electrical and Mechanical Engineering.

“It’s a great example of how the ongoing development of AI, remote sensing and robotics can and will continue to create more opportunities for humans to understand, preserve and protect wildlife.”

The rover not only stimulates instinctive behaviours but strengthens essential pack bonds – a vital part of the species' social structure, as African Painted Dogs rely on cooperative hunting to bond, typically in groups of up to six individuals.

“We’ve seen a huge increase in social communication and bonding when the carcass is delivered this way,” said Rachel Hemming, Carnivore Supervisor at Monarto Safari Park.

“The most important relationship for African Painted Dogs is the relationship they have with the pack. Encouraging them to move, hunt, and feed together helps them stay cohesive and promotes natural behaviours.”

At Adelaide Zoo, the focus is on understanding the welfare states of Sumatran Orangutans. A camera system has been installed in their habitat to monitor their activities around the clock, with computer vision algorithms detecting and tracking body movements to build and analyse ‘skeletal’ models. This tool allows researchers to classify behaviours more accurately.

Facial expression tracking is also underway to gather more information about the emotional states of the orangutan. The collected data will be interpreted alongside expert welfare knowledge in the next phase of the project.

“Orangutans, like humans, exhibit distinct facial features and micro expressions that convey emotions such as curiosity, distress and relaxation. Our AI model recognises individuals with high accuracy and interprets facial expressions to estimate emotional states,” said Georgia Mason, a third year Computer Science student who is part of the project.

Students from Electrical and Electronic Engineering, Computer Sciences, Mathematics, Veterinary Bioscience, and Animal Behaviours are collaborating in the research team, collecting data, developing models and field-testing potential solutions.

“These innovative engineering solutions are already significantly advancing our ability to monitor and enrich the lives of zoo-housed animals and show great promise for the future,” said Mark Smith, Conservation Manager at Zoos SA.

Researchers are hoping to secure more funding to expand the study to other species and potentially develop tools like a mobile app that can be used by keepers to access live behavioural insights.