Daniel M. LaChapelle

Lakshay Narula, Daniel M. LaChapelle, Matthew J. Murrian et al.

A public benchmark dataset collected in the dense urban center of the city of Austin, TX is introduced for evaluation of multi-sensor GNSS-based urban positioning. Existing public datasets on localization and/or odometry evaluation are based on sensors such as lidar, cameras, and radar. The role of GNSS in these datasets is typically limited to the generation of a reference trajectory in conjunction with a high-end inertial navigation system (INS). In contrast, the dataset introduced in this paper provides raw ADC output of wideband intermediate frequency (IF) GNSS data along with tightly synchronized raw measurements from inertial measurement units (IMUs) and a stereoscopic camera unit. This dataset will enable optimization of the full GNSS stack from signal tracking to state estimation, as well as sensor fusion with other automotive sensors. The dataset is available at http://radionavlab.ae.utexas.edu under Public Datasets. Efforts to collect and share similar datasets from a number of dense urban centers around the world are under way.

Daniel LaChapelle, Todd Humphreys, Lakshay Narula et al.

This paper offers a technique for estimating collision risk for automated ground vehicles engaged in cooperative sensing. The technique allows quantification of (i) risk reduced due to cooperation, and (ii) the increased accuracy of risk assessment due to cooperation. If either is significant, cooperation can be viewed as a desirable practice for meeting the stringent risk budget of increasingly automated vehicles; if not, then cooperation - with its various drawbacks - need not be pursued. Collision risk is evaluated over an ego vehicle's trajectory based on a dynamic probabilistic occupancy map and a loss function that maps collision-relevant state information to a cost metric. The risk evaluation framework is demonstrated using real data captured from two cooperating vehicles traversing an urban intersection.