Brachiation Simulation

The idea behind this project was to compare brachiation mechanics in gibbons, humans and using a forward dynamic simulation. This was started as a student group project in Edinburgh and progressed reasonably well in terms of data collection. However we never managed to get the simulator to do more than a single swing which is not really enough for any analysis. I am sure we will be able to fix this - it was the first simulator we attempted and we were running on a single 1.2GHz Linux box at the time. One day this project will be dusted off and re-engineered to run on our Beowulf Cluster and we should get some interesting results. This might make a good PhD for someone! Click on the images on the left to see the movies.

Gibbon Brachiation

Gibbons are the masters of brachiation. This clip was taken at Apenheul near Amsterdam in the Netherlands and shows an example of high-speed richochetal brachiation. Gibbons make this sort of thing look easy but look at the human clips for comparison.

Human Brachiation

		When you ask a human to brachiate you probably expect the sort of performance illustrated on the left. Here the subject is clearly trying to minimise the single arm support time and the result is an extremely uneven swing that ends up being energetically expensive and tiring.

		However we found that a few subjects produced much better brachiation using long, slow swings similar to those seen when gibbons are moving slowly through the canopy. The end result as seen on the left is much less tiring than the uneven swing shown earlier. This suggest that with practice humans could brachiate reasonably well and might well make a good comparison. Interstingly the strength of the subjects did not correlate with brachiating ability.

Brachiation Simulator

The simulator was based on Dynamechs and used custom written contact elements to simulate the pole and the hand. Several alternatives were tried to produce a suitable muscle activation pattern (neural net, finite state, rule-based) using genetic algorithms to find the solutions. The best result was still rather disappointing (left) but encouraging and illustrates that this approach is likely to be successful.