Robot Learns to Walk Like a Stick Insect in Under an Hour, Using AI That Copies the Goal, Not the Steps

Researchers led by Japan's Tohoku University and Thailand's Vidyasirimedhi Institute of Science and Technology (VISTEC) have taught a hexapedal, or six-legged, robot to walk over different surfaces by having it learn from the walking pattern of a single stick insect. 'It's remarkable that a few steps from a single stick insect were enough to find a principle that works on a machine five times its size,' said Dai Owaki, an associate professor at Tohoku University. The findings were published in the journal Bioinspiration & Biomimetics.

Insects have tiny brains and nerve networks compared with humans, yet are capable of remarkably complex and adaptable movement, and researchers have long struggled to identify and computationally model the exact principles behind their leg coordination. Earlier attempts produced imperfect coordination rules that limited flexibility and were too specific to transfer from one robot body type to another. Owaki's team took a different approach: 'We never told the robot how to walk. We asked what the insect was trying to achieve, and let the robot chase the same thing entirely on its own,' Owaki said.

The team used a technique called adversarial inverse reinforcement learning, an imitative method that identifies the 'reward' of safe foot placement while adapting to changing conditions and identifying new rewards. The hexapedal robot needed only brief exposure to a stick insect walking on flat ground to learn how to walk over various terrains, and did so in under an hour. The reward network proved robust enough to work across robots with different body types without needing specific customization for each one.