For biological systems, the prevailing theory suggests that legged animals coordinate joint actuation through antagonistic pairs of muscles controlled by spinal sensorimotor circuits. This 'myotatic unit' concept is mimicked in nowadays robots through the control of joint extension and flexion, through one actuator per joint. Joint control is typically achieved through complex, optimized algorithms based on internal robot models and rapid sensory feedback loops. Phase transitions between stance and swing are controlled based on fast sensory feedback and communication. As a results, robots can smoothly transition through the gait cycle and learn to react to unforeseen perturbations []. However, the robustness and agility of legged robots remain limited. Paradoxically, animals vastly outperform current robots despite considerably slower sensing and information transfer rates.

Previous evidence suggests the potential for embodied, intrinsic mechanics and interjoint mechanical coupling in vertebrates’ legs to simplify control. Multiarticular muscle-tendon coupling can facilitate energy transfer between joints and improve efficiency by allowing muscles to work closer to optimal length and velocity. So far, the role of multiarticular mechanisms in the control of animal locomotion remained poorly understood. A challenge for demonstrating the role of embodied, intrinsic mechanics in animal locomotion is that both active neural and intrinsic mechanical control occur simultaneously.

In the BirdBot [] studies, we tested the hypothesis that an avian-inspired linkage mechanism can replace most of the neural circuitry required to control leg trajectory and transitions between stance and swing phases. We developed a multijoint linkage mechanism fully integrated into a bipedal robot’s legs that achieves consistent interjoint coordination and rapid, automatic phase transitions between stance and swing. The leg design was inspired by the muscle-tendon units of large ratite birds. A multiarticular spring network guides the leg trajectory and provides a rapid transition between stance and swing using a mechanism reminiscent of a self-engaging and disengaging clutch.

At BirdBot robot, mechanical leg joint coupling engages and disengages the robot's legs. This functionality can be extended, to individually engage toe joints, i.e., for grasping and perching, or to adhere to rough terrain [] during legged locomotion.