Scientists have developed a system to learn how humans drive.

Scientists at the Centre for Computational Intelligence at Nanyang University have been working on an intelligent system that learns how to drive from example. They hope to begin paving the way towards developing systems that can drive cars like humans.

Driving skills are a product of the intense cerebellum processing of visual, auditory and even olfactory input signals which are factored to approximate rules and spatial awareness. This is then relayed as motor neuron responses which we have learnt to associate with the controls of a vehicle. In other words, we sense what’s going on around us and adapt how we drive accordingly.

With this complexity, creating a rule-set for driving is almost impossible as it requires constant decision making based on information gathered from the surroundings, which can change rapidly on the road. Inspired by the way human brains learn from being exposed to a situation, the computer system will analyse driving simulations and create its own rule set as they go along.

A driving simulator provides scenarios whilst a data collection feature analyses the responses to these scenarios. A driving model is created from neuro-fuzzy rule-based subsystem collecting the data in an ‘intelligent system’.

Tests are carried out on both a simulator and microprocessor controlled model car. Driving skills successfully learnt so far are parallel parking, U-turns and motorway lane-following and lane changing with more advanced tactical driving skills on the cards.

Automatic cars is often depicted in futuristic films however before the practicality would only work unless all cars were automatically controlled, which would be unfeasible. This system however provides a way robotic driving could exist side by side with human drivers. The implications of this system reach farther than just automobiles with the simulated learning techniques being used in other advanced human actions such as medical procedures and maybe even playing in a band.