Tokyo Tech 6-Legged Controller Gets Closer to Nature

A new controller from the Tokyo Institute of Technology is using just a few simple parameters as input to generate diverse, lifelike movements in a six-legged robot. This semiautonomous robot surprised its handlers by executing gaits that it had not been explicitly designed to produce, but which are more commonly observed in real-world insects. The ability to generate novel behaviors without first learning them from a computer simulation brings insect-based robots much closer to scuttling across the ground without human assistance.

Central Pattern Generator Has Biological Basis Across Many Organisms

The key to the controller’s versatility lies in its two-layered circuit system, which allows for the generation of an overall gait pattern as well as local adjustments in each of the controller’s legs. The top level of the controller acts as a central pattern generator (CPG), setting the rhythm of the robot’s gait.

CPGs are observed in almost all living things that are capable of walking, swimming or flying. They manifest as a series of nerve impulses that are sent out from the lower levels of the nervous system—without ever passing through the brain. In one gruesome study, cats were shown to have intact CPGs even after their spines were severed.

The lower level of the controller created at Tokyo Tech consists of six local pattern generators (LPGs) that take the output from the CPG to calculate and execute individual leg trajectories.

In the video above, the activity in the controller is gradually amplified until a gait pattern emerges. The robot starts thrashing uncontrollably when the LPGs are disabled and fail to fine-tune the gait set by the CPG. Similar movement patterns are observed in insects that have been poisoned with a neurotoxin. (Video courtesy of the Tokyo Institute of Technology and IEEE.)

An Analog Gait in a Digital World

Circuit of the two-level controller (top) that makes the insect-like movements of the new robot (bottom) possible. (Image courtesy of IEEE and the Tokyo Institute of Technology.)

Field-programmable analog arrays (FPAAs) installed in the CPG portion of the controller can be configured digitally, but perform real-time adjustments on analog circuits. This means that while the robot’s CPG can have its parameters set by an easily written (and decidedly nonbiological) digital program on a computer, the small gait and posture adjustments necessary for the controller to adapt to novel terrain are executed via gradual analog oscillations, with the attendant decreases in processing time and energy consumption.

Previous implementations of robot gait and posture, such as the cockroach-inspired hexapod that can climb up walls, relied on digital circuits to make these minute adjustments. In this case, the controller’s circuits were programmed by a computer that had learned from real-world data and simulations before transferring a gait implementation to the robot. Instead of a digital computer learning for the hexapod, and writing a program for it to implement, this analog robot is taking its first steps on its own and adjusting to environmental feedback in real time.

Top: Architecture of the controller, showing the central pattern generator, which controls the overall gait pattern, and at both sides, the six local pattern generators, which control individual leg trajectories. Bottom: Representative still-images of the robot walking in an ant-like posture (left) and a cockroach-like posture (right). (image courtesy of IEEE Access and Tokyo Institute of Technology.)

Check out how the pairing of digital and analog signals is being used to generate synthetic biological circuits .