Road-Shoulder Scanning for Minimum Risk Maneuver In Semi-Automated Driving Using Multi-sensor Kalman Filter Aptiv

Format:

Book

Conference/Event

Author/Creator:

Zhang, Guoguang, author.

Contributor:

Khayyer, Pardis

Sugiarto, Tanto

Conference Name:

SAE WCX Digital Summit (2021-04-13 : Live Online, Pennsylvania, United States)

Language:

English

Physical Description:

1 online resource cm

Place of Publication:

Warrendale, PA SAE International 2021

Summary:

Based on SAE's classification, level 3 driving automation and beyond will be capable to free human drivers while semi-automated driving features are engaged. However, semi-autonomous vehicles may encounter scenarios that under which, the semi-autonomous driving features might not be available or safe. Such scenarios include sensor/actuator malfunction, hazardous road-condition, et cetera To comply with safety requirement, semi-automated vehicles should be able to perform Minimum Risk Maneuver (MRM) into road-shoulder whenever semi-autonomous mode isn't possible.Determining availability of the road shoulder is a key to the MRM, but is also a challenging problem given state of art perception signals. None of the perception signals are robust enough to correctly detect shoulder availability for all circumstances.In this paper, a multi-sensor Kalman Filter is utilized to fuse multiple perception signals from radar, vision, and drivable corridors. The outputs of these sensors are the respective estimation of the length of available shoulder. Sensor-selection logic is performed to identify and discard bad perception signal prior fused operation. The multi-sensor Kalman filter fuses sensor information and generates an estimated length of available road shoulder, which passes to decision algorithm, where the decision is made whether the shoulder is available or not.The proposed algorithm is implemented in dSpace MicroAutoBox on a standard sedan. The algorithm results of the MicroAutoBox are logged, as well as the front view camera. The logging and execution rate is 100Hz.To evaluate performance of the algorithm, 1 hour 20 min's driving is performed. Manual ground-truth is performed by visually observed recorded camera-data. According to road test data, the accuracy of detecting available road shoulder has been greatly improved from 67% to 86% compared with camera-only method

Notes:

Vendor supplied data

Publisher Number:

2021-01-0867

Access Restriction:

Restricted for use by site license