University of Cape Town's Surprising Discovery About Cheetahs Helps Build Better Robots in South Africa

Research from the University of Cape Town's African Robotics Unit reveals how cheetah biomechanics are revolutionizing robot design.

Key Points

✓ Cape Town researchers discovered cheetah tails use aerodynamic drag rather than weight for stability at high speeds. This finding led to lightweight robot tails that improve self-righting and jumping capabilities.
✓ The African Robotics Unit developed motion capture systems that track wild cheetahs from long distances using LiDAR and telescopic lenses. The technology captures skeletal movement data impossible to replicate in laboratory settings.
✓ Three robots emerged from the research: Dima (wheeled with aerodynamic tail), Baleka (Africa's first bipedal robot), and Kemba (pneumatic-powered quadruped). Each tests different aspects of cheetah-inspired movement mechanics.
✓ The ARU has trained 38 Masters students and two PhD graduates through its research programmes. Two patents and a spinoff company, Acinotech, commercialise affordable biomechanics equipment for African markets.

The cheetah's tail weighs almost nothing—just 2% of the animal's body mass. When Amir Patel's team at the University of Cape Town discovered this, they had to rethink their entire approach to building robots that could match the big cat's legendary agility.

"We'd assumed the tail worked through inertia, like a counterweight," explains Amir Patel, Director of the African Robotics Unit at the University of Cape Town. "But something that light can't generate enough inertial force. We started looking at the fur."

The fur, it turned out, creates drag. At 120 kilometres per hour, even lightweight fur becomes an aerodynamic rudder, stabilizing the cheetah's body through turns that would send most animals tumbling. The discovery has already changed how Professor Patel's team builds robots—their lightweight tail design improved a quadruped robot's ability to self-right and jump forward, work they've now patented.

This finding emerged from a decade of studying cheetahs in ways no one had attempted before. The team's proximity to these animals—an accident of geography that most robotics labs would consider irrelevant—has become their greatest research asset.

Cheetah in motion — Cheetah. Photo by Sammy Wong

Measuring What Moves Too Fast to See

Traditional motion capture fails when your subject accelerates from zero to 100 km/h in three seconds. The biomechanics equipment designed for human athletes in controlled labs becomes useless when confronted with a hunting cheetah. Standard systems require markers attached to the body, controlled lighting, and subjects that stay within a confined space. None of these conditions exist on the African savanna.

The African Robotics Unit built their own system from scratch. They combined LiDAR with telescopic lenses to create long-range motion capture that works from distances where the cheetahs don't even know they're being studied. The setup tracks the full skeleton in motion—every joint, every subtle adjustment—as the animal hunts across open terrain.

"I had this crazy idea to measure wild animals from far distances—the whole body, the kinematics of the skeleton," Professor Patel explains. The system now captures data no laboratory could replicate. While most biomechanics research relies on treadmills and controlled environments, the Cape Town team watches cheetahs make split-second decisions at full sprint, adapting to uneven ground, changing direction to track prey, recovering from near-falls.

The motion capture breakthrough has applications beyond cheetah research. Sports scientists have expressed interest in using similar systems to study athletes in actual competition rather than laboratory simulations. Wildlife veterinarians see potential for monitoring animal health without capture or sedation. The technology could track everything from a limping elephant to the gait changes that signal disease in antelope populations.

Three Robots, Three Problems

The Cape Town team has built three robots based on their cheetah research, each addressing a different challenge in robotic movement. Dima rolls on wheels but uses a tail for high-speed turns, testing whether aerodynamic control beats inertial systems in practical applications. The robot's tail adjusts its angle to create variable drag, allowing it to corner at speeds that would typically require much heavier stabilisation systems.

Baleka became Africa's first bipedal robot, engineered to jump higher than a human by applying cheetah-derived principles to two-legged locomotion. While a cheetah runs on four legs, the explosive push-off mechanics translate surprisingly well to bipedal jumping. Baleka stores energy in spring-loaded joints, then releases it in coordinated bursts that mimic the cheetah's acceleration pattern.

Kemba takes the most radical approach, using pneumatic pistons instead of traditional motors. "Cheetahs don't try to finely control their feet," Professor Patel observed from the motion capture footage. "They just push off as hard as they can." Kemba does the same—explosive power over precise control. The robot's legs fire like pistons in an engine, each push calibrated for maximum force rather than careful placement.

The team modelled Kemba's pneumatic knees using Simscape Multibody and Simulink, developing controllers that mimic the cheetah's burst-force approach to movement. Where most quadruped robots prioritise balance and foot placement, Kemba prioritises raw speed and recovery. If it stumbles, it doesn't try to prevent the fall—it uses the momentum to spring back up, just like a cheetah recovering from a missed turn.

Kemba legged robot — Kemba, a legged robot which combines powerful robust pneumatic actuators at the knees, with finely controlled electric motors at the hips, for dynamic and agile manoeuvres. Source: uct.ac.za

Patents and Force Plates That Actually Get Used

The research has generated two patents with immediate commercial applications. The first covers their GPS-based motion tracking system, which brings clinical biomechanics capabilities to field conditions. Instead of requiring subjects to visit expensive motion capture laboratories, the system can assess movement patterns anywhere—from a rural football field to a mountainside trail.

The second patent protects a 3D force plate that costs a fraction of conventional systems. Traditional force plates, which measure the ground reaction forces during movement, can cost over $50,000. The Cape Town version achieves similar results for under $5,000, designed specifically to make biomechanics research feasible for institutions operating on limited budgets.

Acinotech, a University of Cape Town spinoff, now commercialises these technologies. The force plate alone could transform how sports science and rehabilitation work across Africa, where a single piece of imported equipment often exceeds an entire department's annual budget. Physiotherapists in Johannesburg have already started using the system to assess patients recovering from knee surgeries. Track coaches in Kenya want to use it to analyse their runners' push-off mechanics.

The trajectory optimisation work has attracted international attention. Carnegie Mellon University researchers now collaborate with the Cape Town team on algorithms that predict optimal movement patterns. One joint paper, "Contact-Implicit Trajectory Optimization Using Orthogonal Collocation," has become Professor Patel's most-cited publication. The algorithm helps robots plan movements that account for unpredictable contact with the ground—essential for navigating rough terrain.

Building the Next Generation

The African Robotics Unit has trained thirty-eight Masters students and two PhD graduates through its various research projects. Both PhD graduates now teach at South African universities, bringing their expertise to Johannesburg and Durban. Each represents a reversal of the typical pattern where African engineering talent leaves for laboratories in Europe or North America.

The students work on problems that don't exist in MIT's laboratories or Stanford's engineering departments. They can walk outside and observe actual cheetahs. They understand the terrain these animals navigate because they've walked it themselves. This proximity to the source material creates insights that can't be replicated through YouTube videos or journal papers.

The project directly confronts what academics call "helicopter research"—the practice of foreign scientists dropping into Africa to collect data, then publishing papers that list local researchers as minor contributors, if at all. At the African Robotics Unit, African scientists lead every aspect, from conceptualisation to publication. When international collaborators join, they join as partners, not as senior authors swooping in to claim credit.

The Optimisation Puzzle

The team's next challenge involves cracking the cheetah's decision-making code. They can measure every movement, calculate every force, but still don't fully understand what the animal prioritises moment to moment during a hunt.

"Is the cat trying to conserve energy when running, increase manoeuvrability, or intercept prey in the shortest time?" Professor Patel asks. The answer changes everything about how robots should move. If energy conservation dominates, robots should glide between bursts of activity. If manoeuvrability matters most, they should maintain readiness for instant direction changes. If time-to-intercept drives decisions, the optimal strategy involves calculated risks and occasional spectacular failures.

The team now uses inverse reinforcement learning—essentially running their equations backward—to understand what objective function evolution has programmed into the cheetah's neural circuits. They feed the motion capture data into machine learning algorithms that try thousands of different optimisation strategies until they find one that produces cheetah-like movement patterns.

Their recently acquired Unitree Go-1 quadrupedal robot will test these theories in hardware. The next version of Kemba will add an active spine and tail, bringing it closer to true cheetah biomechanics. Each iteration teaches them something new about how evolution solved the problem of high-speed pursuit across unpredictable terrain.

"There is an advantage to being geographically isolated," Professor Patel reflects. "I have a bit of an underdog mentality."

The African Robotics Unit continues to expand its research, with plans to study other African animals—from the spring-loaded legs of springboks to the climbing mechanics of leopards. Each species offers solutions to problems roboticists struggle with in laboratories thousands of kilometres from where these animals evolved their remarkable capabilities.