For all their sophisticated mechanics, most robots are not usually very clever.

Put something in their way and they will bump into it - and they will keep bumping into the object until you remove it.

But scientists are developing machines that can learn from experience and evolve behaviour.

Take Robo-monkey. It is an example of the new generation of machines that incorporate artificial intelligence, and is featured in Wednesday's edition of the BBC Science programme Tomorrow's World.

Unique problems

Robo-monkey is the product of 15 years of study by a team of researchers led by Dr Toshio Fukuda at Nagoya University, Japan.

The robot will swing like a gibbon from branch to branch - or rung to rung on a horizontal ladder set up in the laboratory. The research team chose to study swinging, or brachiation, because it has a set of unique problems that need to be solved.

Robo-monkey has 14 motors controlling a fully-articulated body and a computer brain to direct all its movements.

The human instructors have told it some equations for swinging and the distance between the rungs - it must do the rest.

Trial and error

The first thing the robot does is kick its legs to initiate the crucial pendulum motion that will allow it to reach out for the first hand hold. The metal creature's progress down the track is a case of trial and error. When it fails to grasp a rung, Robo-monkey must work out why and try again.

Coloured balls on the limbs combined with stereo vision tell the robot where its arms are in space. And the data is updated 60 times a second to provide the realtime information needed to complete different tasks involved in swinging successfully.

Robo-monkey looks impressive.

Swinging itself may not be a skill that is much use to 21st Century robots but the knowledge being built from this experiment will certainly lead to machines that are more interactive and self-reliant.

Tomorrow's World is broadcast on BBC One on Wednesday's at 1930 GMT. And yes we do know that gibbons are apes and not monkeys.