This chapter presents research activities of UTDrive: the smart vehicle project, at the Center for Robust Speech Systems, University of Texas at Dallas. The objectives of the UTDrive project are to collect and research rich multi-modal data recorded in actual car environments for analyzing and modeling driver behavior. The models of driver behavior under normal and distracted driving conditions can be used to create improved in-vehicle humanmachine interactive systems and reduce vehicle accidents on the road. The UTDrive corpus consists of audio, video, brake/gas pedal pressure, head distance, GPS information (e.g., position, velocity), and CAN-bus information (e.g., steering-wheel angle, brake position, throttle position, and vehicle speed). Here, we describe our in-vehicle data collection framework, data collection protocol, dialog and secondary task demands, data analysis, and preliminary experimental results. Finally, we discuss our proposed multi-layer data transcription procedure for in-vehicle data collection and future research directions.