next up previous
Next: Control Identification on Individual Up: Learning to Associate Observed Previous: INTRODUCTION

   
ADAPTIVE AND HAND-CRAFTED COMPONENTS

We divided learning to associate driver behavior with traffic controls into two tasks. The first task is to identify the control present on a road segment. The second is to validate identification by checking the consistency of controls at segment connections. We treated the identification task as a supervised learning problem. We solved it with a supervised neural network as described below in §2.1. We wrote rules that compare the controls on immediately connected segments and used them to validate identification as described in §2.2.



 

Seth Rogers
1998-11-20