This paper describes an ant colony optimization approach adopted to decide on road-borders to automatically guide a vehicle developed for the DARPA Grand Challenge 2004, available from: < http://www.darpa.mil/grandchallenge >. Due to the complexity of off-road trails and different natural boundaries of the trails, lane markers detection schemes cannot work. Hence border detection is based on ant colony optimization technique. Two borders at two sides of the road (as seen by a camera fixed on the vehicle) are tracked by two agent colonies: agents' moves are inspired by the behaviors of biological ants when trying to find the shortest path from nest to food. Reinforcement learning is done by pheromone updating based on some heuristic function and by changing the heuristic balancing parameters with the experience over the last tracked results. Shadow removal has also been introduced to increase robustness.Results on different off-road environments, as provided in the DARPA Grand Challenge 2004, have been shown in the form of correct detections, false positives and false negatives and their dependence on number of ant-agents and balancing edge-exploitation and pheromone-exploitation.
An agent based evolutionary approach to path detection for off-road vehicle guidance / Broggi, Alberto; Cattani, Stefano. - In: PATTERN RECOGNITION LETTERS. - ISSN 0167-8655. - 27:(2006), pp. 1164-1173. [10.1016/j.patrec.2005.07.014]
An agent based evolutionary approach to path detection for off-road vehicle guidance
BROGGI, Alberto;CATTANI, Stefano
2006-01-01
Abstract
This paper describes an ant colony optimization approach adopted to decide on road-borders to automatically guide a vehicle developed for the DARPA Grand Challenge 2004, available from: < http://www.darpa.mil/grandchallenge >. Due to the complexity of off-road trails and different natural boundaries of the trails, lane markers detection schemes cannot work. Hence border detection is based on ant colony optimization technique. Two borders at two sides of the road (as seen by a camera fixed on the vehicle) are tracked by two agent colonies: agents' moves are inspired by the behaviors of biological ants when trying to find the shortest path from nest to food. Reinforcement learning is done by pheromone updating based on some heuristic function and by changing the heuristic balancing parameters with the experience over the last tracked results. Shadow removal has also been introduced to increase robustness.Results on different off-road environments, as provided in the DARPA Grand Challenge 2004, have been shown in the form of correct detections, false positives and false negatives and their dependence on number of ant-agents and balancing edge-exploitation and pheromone-exploitation.File | Dimensione | Formato | |
---|---|---|---|
prl-ants.pdf
non disponibili
Tipologia:
Documento in Post-print
Licenza:
NON PUBBLICO - Accesso privato/ristretto
Dimensione
777.5 kB
Formato
Adobe PDF
|
777.5 kB | Adobe PDF | Visualizza/Apri Richiedi una copia |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.