Runbot - robots that learn to walk, run better with time

Walking robots, so far only seen in science-fiction movies, could become a reality with engineers building a biped machine that adjusts its gait to climb slopes without falling over.

The highly sophisticated robot – named ‘RunBot’ – developed by engineers at the University of Gottingen in Germany, learns to walk like a human toddler.

On its first two or three attempts at climbing a ramp, the machine, consisting of a pair of legs and a trunk, loses its balance and topples backwards, but it quickly learns to correct itself and stay upright by taking shorter steps and leaning forwards. On reaching another flat surface, the robot adjusts its gait again.

Robots that walk upright in this way present a good deal of great technical challenges, but such abilities will be essential if bipedal domestic robots that can help around the house, carry shopping, or care for the elderly are to become a reality.

RunBot already holds the record for robot speedwalking and it can bowl along at more than three leg-lengths per second, slightly slower than the fastest walking human.

The German team of engineers upgraded the machine by giving it an infrared eye and an “accelerometer sensor” which acts like the balance mechanism in the human inner ear.

Both the infrared eye and accelerometer sensor are linked to an artificial ‘brain’ or neural network which decides how the robot should respond when it spots a slope ahead and feels its balance slipping.

The basic walking steps of RunBot are controlled by reflex information received by peripheral sensors on the joints and feet of the robot, as well as the accelerometer which monitors the pitch of the machine.

These sensors pass data on to local neural loops – the equivalent of local circuits – which analyse the information and make adjustments to the gait of the robot in real time.

Information from sensors is constantly created by the interaction of the robot with the terrain so that RunBot can adjust its step if there is a change in the environment.

As the robot takes each step, control circuits ensure that the joints are not overstretched and that the next step begins.

But, if the robot encounters an obstacle, or a dramatic change in the terrain, such as a slope, then the higher level functions of the robot – the learning circuitries – are used.

Of the robot’s total weight, 70% is concentrated in its trunk. RunBot run up slopes without falling by shortening its stride and leaning forward by just the right amount.

Since the ‘brain’ learns from experience, the robot’s performance improves over time.