Enhanced canine odor generalization through concept training in automated olfactometers
Abstract
Abstract: Enhanced canine odor generalization through concept training in automated olfactometers Offeror: Clive D. L. Wynne, PhD Presently, the trained explosives-detection dog (EDD) is the best tool available for the detection of home-made explosives (HMEs), Improvised Explosive Devices (IEDs) and other explosives, but suffers from 1) low training success rates, 2) lengthy training that demands intensive human interaction, and 3) difficulties getting dogs to generalize to novel explosive mixtures. This project will address these issues through the integration of automated olfactometers and modern techniques in animal learning, comparative cognition and animal psychophysics into the training of EDDs. Novel to this proposal is (1) more efficient training of dogs in olfactometers and (2) more effectively training using novel methods. Two specific objectives will be pursued: 1. To demonstrate that automated training in olfactometers can optimize detector dog training 2. To demonstrate the effectiveness of a form of training derived from comparative cognition: Òconcept formation,Ó instead of the standard elemental target training. An olfactometer is any device for the delivery of odors. In this pilot project we will develop a mobile olfactometer that will can automatically deliver a range of different odors, detect dogsÕ responses, and deliver appropriate reinforcers. The adoption of automated olfactometers permits our second innovation: the development of Òconcept formation trainingÓ instead of the standard ÒelementalÓ or Òpure targetÓ training currently in widespread use. Elemental target training involves hiding odoriferous samples in different locations around a training arena manually. Animals trained in this way exhibit search patterns for highly specific stimuli but fail to generalize even to different intensities of the stimulus they were trained on. They also ignore closely related odorants or the same odorant in the presence of contaminants or distractors. To train an animal to generalize detection of a target odorant in widely varying and novel mixtures a different method, concept formation, is better suited. In concept formation training, the animal is exposed to a large and varied set of both target and non-target stimuli. Animals trained in this way show broader generalization to diverse target stimuli even in the presence of distractors. In this pilot project, dogs will be trained to alert two different explosive precursors/components (Ammonium Nitrate and Hydrogen Peroxide) using concept formation training and pure target training. Throughout the study, we will assess whether concept formation training or pure target training leads to enhanced detection of ammonium nitrate and hydrogen peroxide in odor mixtures the dogs have never previously been exposed to. We expect concept formation training will lead to a far superior detection of the explosive precursors in novel mixtures.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- May 22, 2016
- Source ID
- N000141512347
Entities
People
- Clive D.L. Wynne
Organizations
- Arizona State University
- Office of Naval Research
- United States Navy