The four-legged robot Morti, described in the journal Nature Machine Intelligence on Monday, optimizes its movements faster than an animal, learning to walk in about an hour, scientists say. In animals, muscle coordination networks located in the spinal cord help them take their first steps, but learning to precisely coordinate leg muscles and tendons takes some time, say researchers from the Max Planck Institute for Intelligent Systems (MPI -IS) in Germany. . Baby animals initially rely heavily on hard-wired spinal cord reflexes, studies have shown. The labrador-sized robot has animal-like reflexes and learns to walk from its mistakes using the complex mechanics of its legs and a learning algorithm, says Felix Ruppert, an official PhD student at MPI-IS. “We cannot easily investigate the spinal cord of a living animal. But we can model one in the robot,” study co-author Alexander Badri-Sprowitz said in a statement. In the quadruped robot, data from the leg sensors is continuously matched with target data from its modeled virtual spinal cord running as a program on its computer. It learns to walk by constantly comparing the sensor information sent and what is expected, running reflex loops and adjusting its motor control patterns, the researchers explain in the study. The robot has a central pattern generator (CPG) that works similar to the networks of neurons in the spinal cord of animals that produce periodic muscle contractions without input from the brain. These neural networks help create rhythmic tasks such as walking, closing eyes, or digestion in animals. When young animals walk on a perfectly flat surface, they say these nerve networks may be sufficient to control movement signals from the spinal cord, but a small bump on the ground can alter the walk. This is when the reflexes kick in and adjust movement patterns to prevent the animal from falling. But in newborn animals, researchers say these nerves aren’t regulated well enough at first, and the animals stumble, but soon learn how these reflexes control leg muscles and tendons. “We know that these CPGs are present in many animals. We know that reflexes are built-in. but how can we combine the two so that animals learn movements with reflexes and CPGs?” said Dr. Badri-Sprowitz. The scientists say Morti’s CPG – simulated on a small, lightweight computer, controls the movement of the robot’s legs – also learns in the same way. Sensor data from the robot’s legs is constantly compared to the expected fall predicted by this virtual spinal cord. “The data is fed back from the sensors to the virtual spinal cord where the sensor and CPG data are compared. If the sensor data does not match the expected data, the learning algorithm changes the walking behavior until the robot walks well and without stumbling,” explained Mr. Ruppert. “Changing the output of the CPG while keeping the reflexes active and watching the robot stumble is a key part of the learning process,” he added. The learning algorithm changes how far the legs swing back and forth, how fast the legs swing and how long a leg is on the ground when the robot dog stumbles, the study noted. “Our robot is practically ‘born’ knowing nothing about the anatomy of its legs or how they work. CPG is like a built-in automatic walking intelligence provided by nature that we have transferred to the robot,” added Mr. Ruppert. “This is fundamental research at the intersection of robotics and biology. The robotic model gives us answers to questions that biology alone cannot answer,” added Dr. Badri-Sprowitz.
