Gizmag has an interesting article covering research being undertaken by Dr Erich Rome. The Multi-sensory Autonomous Cognitive Systems (MACS) project is aimed at giving intelligent agents environmental awareness through appearance-based recognition, instead of the traditional feature-based recognition.
As I confusingly noted in the heading, whilst a traditional recognition approach would search an environment for a chair, MACS would search that same environment for something to sit on - a chair, stool, couch etc. This approach is an affordances-based environmental understanding.
Gizmag's article, and its source, are certainly worth a read, especially for those interested in robot-environment interaction - beyond just not hitting anything!
source
