Unified Event Data Recorder for a Self-Driving Car

Warning Icon This event is in the past.

Date: February 25, 2020
Time: 11:30 a.m. - 12:20 p.m.
Location: Old Main #1163 | Map
4841 Cass
Detroit, MI 48201
Category: Seminar

Abstract

An Event Data Recorder (or EDRs) serves as a blackbox or an accident data recorder for automobiles to record information related to vehicle crashes or accidents. Almost all vehicles in the US manufactured from 2006 have been fitted with event data recorders. EDRs work by continuously recording data such as vehicle acceleration and speed at the time of impact, braking, airbag deployment, seat-belt status, steering angle etc. The function of an EDR is to capture such events prior to an accident, at the time of the accident, and a few seconds after the accident. Usually audio and video data are excluded from EDRs. This serves as a history to determine the cause of an accident either due to fault in the vehicle or external circumstance or driver error.

With the advent of compute engine SoCs that enable deep learning, object detection and tracking, and sensor fusion from sources such as camera, LiDAR, Radar, GPS, etc., it becomes imperative for an EDR to evolve in an autonomous self-driving system. The challenges faced range from not only capturing decisions, objects detected, position/navigation but also stitching a clear picture from sensors. All this has to be recorded synchronously in time along with system logs of various domain controllers such as Brake Control Unit, Steering Control unit, Vehicle Control Unit, Gateway, Digital Cockpit, and Autonomous control unit. The design challenges imposed by all this on an embedded real-time system will be explored in this talk. Standardization of tools to extract and interpret EDR data either from the cloud or the blackbox will be discussed as well.

Bio

Dr. Manish Kochhal is currently a technical lead in the Autonomous Drive group at NIO, Inc. He leads a team that develops platform software that enable autonomous drive assist features such as navigate on autopilot, automatic emergency braking, and event data recorders for an L2 self-driving car. These platform features have been deployed in the premium luxury SUV segment cars such as ES8/ES6 in China. In the past, Manish has worked in wireless companies such as Broadcomm, Qualcomm, Solpad, and Airvana. His work in these companies have scaled from chips that enable communications 3G/4G/WiFi/BT/BLE to devices such as smart phones, cable modems, IoT-enabled solar inverters that use these chips and finally to support VoIP/QoS on 3G EVDO base-stations and radio network controllers. Manish Kochhal graduated with a PhD in Wireless Sensor Networks (WSNs) and M.S. in Mobile Ad hoc Networks (MANETs) both under Dr. Loren Schwiebert 

Contact

LaNita Stewart
313-577-2478
LStewart@wayne.edu

Cost

Free

Audience

Academic staff, Alumni, Community, Current students, Parents, Prospective students, Staff