AV-CONNECT, INC. — Department of Energy SBIR Phase II: 12c

Description

Consumer surveys show that the biggest obstacles to the mass adoption of electric vehicles EVs by consumers are range anxiety, charging time, inadequate charging infrastructure & cost of vehicle. The company has developed high precision predictions of vehicle energy consumption and optimal charge management. The approach is to combine machine learning algorithms and predictive behavior modeling to process EV sensor telemetry, contextualized with maps, traffic and weather forecasts, and deliver high precision realtime predictions of journey charge consumption and charge depletion trajectory on the journey route while providing optimal charging recommendations based on driver preferences for journey time, time of arrival at destination, charging time and minimum battery charge during and end of trip. In Phase I, the company has developed and validated a learningbased cloud platform for EVs with observed energy savings of 1520%. The platform consists of: proprietary vehicle performance models capturing the energy consumption and travel time on a pervehicle, perdriver, perroadsegment basis; novel learning algorithms which estimate model parameters from contextual data; and modelbased algorithms. The algorithms predict and optimize vehicle performance on road networks delivering high precision charge consumption prediction, optimization of charging stops, and route optimization. During our Phase I technical development and customer interactions we have identified two critical elements required by automakers OEMs: 1 prediction accuracy: changes of time varying parameters such as weight, air drag and temperature gradients have a major effect on charge consumption, and 2 adoption: driver engagement cannot be taken for granted and the platform needs to deliver high precision prediction even when driver destination is unknown. In Phase II, we propose to build on the work of Phase I and develop models and learning algorithms to address both issues by improving prediction accuracy in parametervarying scenarios and regardless of driver engagement by learning of destination. The company’s Contextual Intelligence platform will deliver numerous cloud services to EVs from highprecision charge consumption prediction, intelligent charge management, ecorouting, ecocruise control among others. This would enable a more rapid adoption of EVs and improve their energy performance, resulting in dramatic reductions in fuel emissions and GHG through the replacement of gasoline vehicles.