In these experiments the researchers are examining algorithms for robots to ‘recover’ from damage. In this case, they shortned a leg of the hexapod, and use a T-Resilience Algorithm to calculate a new walking pattern/gait.
Their initial gait was able to acheive a top speed of 26cm/s. When they shortened one leg to half it’s original length, the performance dropped to 8cm/s. At this point the robot begins to calculate a new gait. It runs 40 simulations, then trys the best of these simulations in real life. The hexapod will run this experiment 25 times to determine the best new gait. In only 20 minutes the hexapod is able to use a new gait that gets 18cm/s – three times the performance of the original gait under damaged conditions!
Now this is no normal PhantomX Hexapod. This model has a custom chassis, an IR camera, MX-28s, an on on-board computer, and a custom built ‘damaged’ leg.
Researchers Sylvain Koos, Antoine Cully and Jean-Baptiste Mouret have done a great job with this experiement as well as documenting it. More information about the paper can be found here and here. A PDF of the paper can be found here.