Robot dog learns to walk in just one hour by analyzing the animals’ movements

They are born with muscular coordination networks with rigid threads located in their spinal cord that they rely on to stand and move through neural reflexes.

Although a little more basic, motor control reflexes help a puppy avoid falling and injuring itself during their first attempt at walking. Then further and more precise muscle control is practiced until the nervous system is well adapted to the animal’s leg muscles and tendons so that it can keep up with adults.

Researchers at the Max Planck Institute for Intelligent Systems (MPI-IS) in Stuttgart, Germany, conducted a study (published in the journal Intelligence of natural machines) to find out how animals learn to walk and overcome obstacles. For this, they built a four-legged robot, the size of a medium-sized dog, which helped them understand the details.

German researchers have developed a robot dog to understand the movement learning process in four-legged puppies. Photo: Max Planck of Intelligent Systems (MPI-IS)

“As engineers and robot technicians, we seek the answer by building a robot that thinks like an animal and learns from its flaws,” Felix Ruppert, a former doctoral student in the Dynamic Locomotion Research Group at MPI-IS, said in a statement. “If an animal stumbles, is it a mistake? Not if it happens once. But if it stumbles often, it gives us a measure of how well the robot is doing.

A learning algorithm optimizes a robot dog’s virtual spinal cord

A Bayesian optimization algorithm drives machine learning: information measured by the foot sensor is combined with target data from the modeled virtual spinal cord running as a program on the robot’s computer. It learns to walk by continuously comparing information sent and received from the sensor while running loops reflexes and adaptation of their motor control patterns.

The learning algorithm adjusts the control parameters of a Central Pattern Generator (CPG). In humans and animals, these central pattern generators are neural networks in the spinal cord that produce periodic muscle contractions without brain input.

Networks of key pattern generators help generate rhythmic tasks such as walking, blinking, or digesting. In addition, reflexes are involuntary motor control actions triggered by coded neural pathways that connect sensors in the bone to the spinal cord.

When the young animal walks on a perfectly flat surface, CPGs may be sufficient to monitor movement signals from the spinal cord. However, a small bump on the ground changes course. Reflexes insert and adjust movement patterns to prevent the animal from falling.

These instantaneous changes in motion signals are reversible or “elastic”, and motion patterns return to their original configuration after the disturbance. But if the animal does not stop stumbling through many cycles of movement – despite active reflexes – then the movement patterns must be repaired and made “plastic”, ie. irreversible.

1658268729 344 A robot dog learns to walk in just one hour
Morti, the robot dog created to understand the movement learning process for four-legged animal puppies. Photo: Max Planck of Intelligent Systems (MPI-IS)

In the newborn animal, the CPGs are not yet close enough at first, and the animal stumbles, both on flat and uneven ground. However, the animal quickly learns how its CPGs and reflexes control the leg muscles and tendons.

The same goes for the robot dog called “Morti”, which also optimizes its movement patterns faster than an animal, in about an hour. Your CPG is simulated on a small, lightweight computer that controls the movements of your legs.

The virtual spinal cord is placed on the four-legged robot where the head would be. During the hour it takes the robot to walk smoothly, data from its foot sensor is continuously compared to the expected trip predicted by the CPG.

If this happens, the learning algorithm changes the distance and speed at which the legs swing back and forth, and measures how long they stay on the ground. Adjusted motion also affects how the robot can use its compatible leg mechanics. During the learning process, CPG sends appropriate motorized signals so that the robot begins to stumble less and optimizes its gait.

In this structure, the virtual spinal cord has no explicit knowledge of the design of the robot’s legs, its motors, and its springs. Since he knows nothing about the physics of the machine, he is missing a “model” robot.

“Our robot was practically ‘born’ without knowing anything about its autonomous legs or how they function,” says Ruppert. “CPG is like an integrated automatic gait intelligence that nature provides and that we transmit to the robot. The computer produces signals that control the leg motors, and the robot initially stumbles. Data goes up 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 gait behavior until the robot runs well without tripping.Changing CPG output, keeping the reflectors active and monitoring robot tripping is central to the learning process.

Low power consumption makes the mechanism more viable

While four-legged industrial robots from well-known manufacturers who have learned to operate using complex controllers are very power-intensive, Mortis’ computer only needs five watts to run.

“We cannot simply search the spinal cord of a living animal. But we can model one on the robot, ”says Alexander Badri-Spröwitz, co-author of the study and head of Dynamic Locomotion Research Group, and Max Planck. “We know that these CPGs are found in many animals. And we know that there are built-in reflexes, but how can we combine the two so that animals learn movements with reflexes and CPGs? This is basic research at the crossroads of robotics. and biology.The robot model gives us answers to questions that biology alone cannot answer.

Have you watched our new videos Youtube? Subscribe to our channel!

Leave a Comment