Swarm Intelligence

Optimization of traffic flow

Many downhill skiers have experienced that time spent queuing for the lift is longer than the time spent in the ski slope.  The goal of the Swarm Intelligence Project at Mid Sweden University is to create a framework that will enable ski resorts to minimize the queuing in the lift systems. Carrying a personal digital assistant or a cellular phone (PDA) the skier will get hints of which lift is less occupied and hence maximize the experience of downhill skiing. The framework will equalize the load of all lifts in the ski area and minimize the time spent queuing. The framework can be generalized and applied in other domains such as traffic in cities.

A branch of artificial intelligence and machine learning, swarm intelligence (SI), focus on small systems with limited intelligence. An example is the foraging strategies of ants. Despite their limited intelligence they will eventually find the shortest path to their food by using pheromones for trail-laying and trail-following. The pheromones are used to communicate and the strength of the pheromones will decrease with time. Another example is termites creating huge termite stacks without any planning or central supervision. Many small units communicating will create a relatively intelligent system.

In downhill skiing we will copy the phenomena by equipping each skier with a PDA with limited intelligence. The PDA communicates in short range and contains a GPS. A GPS is a receiver that uses signals from several satellites to calculate current position. The PDAs will hence acquire knowledge of the current speed and direction of the skiers in the close environment and also knowledge of normal behaviour in the ski slope and the lift systems. Traffic congestion will show different behaviour than the normal pattern and the artifact may advice the person going downhill to select another slope and hence avoid queuing. The framework will use machine learning and hence not attached to any specific lift system. The quality of the advice will increase as time goes by.

The project is sponsored by CyberCom group and Skistar and is granted free access to the lift system in Åre for three years. During the first winter season the geography of the resort was measured by students equipped with GPS. Data collected were used to simulate downhill skiing in Åre and the first results indicated positive outcome using Swarm Intelligence.