The National Research Council describes Rocky, a knee-high experimental robot that learns to navigate obstacles using reinforcement learning and an array of sensors.
What Rocky can do today
In an October 28 research feature, the National Research Council of Canada described Rocky, a knee-high, four-legged experimental robot being trained to move across different surfaces. NRC scientist Abhishek Naik uses reinforcement learning, including simulated environments, to teach the machine to adjust its movement when encountering slopes and uneven terrain. The robot has been shown walking outdoors and in a facility that simulates lunar terrain. Those demonstrations are research trials, not evidence that the robot is already independently operating in emergency situations.
How the technology works
Rocky's equipment includes accelerometer and gyroscope sensors, a camera, LiDAR and sensors that detect forces acting on its feet. The sensors give the control system information about the robot's movement and surroundings. During training, an AI approach called reinforcement learning rewards behaviours that help the robot balance and progress across obstacles. The NRC notes that the machine still uses remote guidance in its current work, while greater autonomy remains the research goal.
Potential applications and limitations
The research team is investigating whether adaptable robots could eventually help with disaster recovery, mining, agricultural tasks or exploration in hazardous environments. These are potential applications, not deployed products or guaranteed capabilities. Moving from controlled tests to dependable outdoor operations requires safety testing, reliability across unfamiliar surfaces and meaningful assessment of how the machine behaves when conditions change.
Sources & transparency
National Research Council Canada — Rocky the robot, October 28, 2026Original CDN News coverage. Sources are linked for reference.
Request a correction