Project: #329 Development of Safe, Profitable, and Fair Robotaxi Deployment Strategy Progress Report - Reporting Period Ending: Sept. 30, 2020 Principal Investigator: Ding Zhao Status: Active Start Date: July 1, 2020 End Date: June 30, 2021 Research Type: Advanced Grant Type: Research Grant Program: FAST Act - Mobility National (2016 - 2022) Grant Cycle: 2020 Mobility21 UTC Progress Report (Last Updated: Oct. 5, 2020, 6:27 a.m.) % Project Completed to Date: 10 % Grant Award Expended: 0 % Match Expended & Document: 0 USDOT Requirements Accomplishments We have used a data-driven method to identify the risk level of roads in the City of Pittsburgh based on the Argoverse dataset [1]. We first leverage a nonparametric method, Dirichlet Process Gaussian Process (DPGP), to identify scenarios in several representative road patches. Each scenario is represented by a 2D gaussian process (GP). The risk level is then defined based on the linear combination of complexity features, including (1) the number of clustered GPs and (2) the features of GPs, such as the kernel lengthscale that indicates the spatial changing rate of the scenario. We ran our algorithm on Amazon Web Services (AWS) to obtain the heatmap of 52 sections. We represented the results in a heatmap. We reviewed ride-sharing vehicle dispatch in literature that could provide guidelines for how to design dispatch algorithms for robotaxis. The main differences between the ride-sharing and robotaxis are identified: 1. The optimization for robotaxis involves vehicles with different functionalities. Therefore, there are more variables associated with vehicles to consider besides location and time in the case of ride-sharing. 2. Driver incentives are different since the robotaxis would most likely be owned by a company which has control over them, while drivers for ride-sharing are human individuals. 3. Since there is very limited real-world data for robotaxi at the current stage, we are not able to validate algorithms on empirical data. We will use simulation as an alternative. Impacts After obtaining the complexity heatmap for the City of Pittsburgh, we can incorporate the information of route complexity/risk into the optimization of robotaxi dispatch. With the knowledge of route complexity, we can reduce unnecessary sensors for specific driving zones that do not require sophisticated sensing capabilities of the vehicles. The operation cost of robotaxi companies will therefore be drastically reduced. Other We do not have output to report yet. Outcomes New Partners n/a Issues n/a