LOCOMORPH explored a fundamental idea: that the physical body of a robot is not just a passive carrier for sensors and actuators, but an active participant in the control of movement. This concept — morphological computation — proposes that by carefully designing the shape, material properties, and mechanical compliance of a robot's body, you can reduce the computational burden on its controller and produce more robust, adaptive locomotion.
Traditional approaches to robot locomotion treat the body as a rigid structure and rely entirely on software to handle the complexity of walking, running, and climbing. This works in predictable environments but fails when conditions change. A legged robot on flat concrete behaves very differently on sand, gravel, or ice — and reprogramming the controller for every new surface is impractical.
Animals solve this problem elegantly. A cat landing on an uneven surface doesn't recalculate joint trajectories in real time. The compliance of its tendons, the springiness of its pads, and the passive dynamics of its legs absorb most of the variation. The nervous system only needs to provide high-level commands; the body handles the details.
LOCOMORPH investigated this principle across multiple levels:
The project used several custom-built robot platforms to test these ideas. These included modular legged robots whose limb geometry could be reconfigured, compliant walking machines with tunable joint stiffness, and simulation environments that could model the interaction between soft body structures and complex terrain at high fidelity.
The LocoKit platform, developed by the NextStep Robotics Collective, was a particularly important tool — an open-source modular robotics kit that allowed researchers to rapidly prototype and test different body configurations without building each robot from scratch.
LOCOMORPH produced several significant results that have since become foundational in the field: