SBIR Phase II: Energy-Efficient Perception for Autonomous Road Vehicles

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase II project will be to allow more consumers to make use of assisted and autonomous driving systems in automobiles. Fully Autonomous Vehicles (AV) will reduce traffic collisions and enable humans to spend less time driving and more time on productive activities. Commercially deploying AVs requires a number of key technologies including sensing, perceptual systems, motion planning, and actuation. Our discussions with leaders and decisionmakers at automotive companies have shown that the development of robust, accurate, and energy-efficient perception systems is a major technical obstacle to creating mass-producible autonomous road vehicles. Of particular interest to automakers is scaling down the computational requirements of perceptual systems while preserving high levels of accuracy and robustness. This Small Business Innovation Research (SBIR) Phase II project will use deep learning to create perception systems that are (a) scalable across different computational platforms and (b) scalable across smaller or larger sensor sets. Specifically, the company will develop scalable systems from small compute platforms (used for Highly Automated Driving) to somewhat larger compute platforms (used for Fully Automated Driving). Further, the company will develop perceptual systems that scale from few sensors to many sensors. The goal is to "do more with less," advancing the pareto-optimal frontier of efficiency-accuracy and price-accuracy tradeoffs. The company has already engaged with automotive OEMs and suppliers to develop partnerships and to define metrics for success in this endeavor. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.