Energy Efficient Computing for Automated Vehicles (EECAV)


Automated vehicles (AV) hold great promise for improving transportation safety, as well as reducing congestion and emissions. In order to make AVs commercially viable, a reliable and high-performance vehicle-based computing platform that meets ever-increasing computational demands will be key. Meeting energy efficiency and computational performance goals requires research and development (R&D) targeting computation and energy efficiency within the size, weight, power, and thermal constraints of a vehicle.

Sandia, in collaboration with industrial and academic partners, has developed a roadmap outline, Energy Efficient Computing R&D Roadmap Outline for Automated Vehicles. The Roadmap Outline identifies in a technically unbiased way the R&D challenges that must be overcome for the realization of highly automated driving in retail vehicles with low power consumption and high computational performance. The Roadmap Outline has two purposes: to provide guidance on future public and private funding in this arena and stimulate a more detailed R&D Roadmap as an important next step in developing safe and reliable AV technology.

Browse reports, presentations, and other materials related to this activity below. All materials are available for direct download.

Energy Efficient Computing R&D Roadmap Outline for Automated Vehicles  

Download the Roadmap Outline (PDF)

The Energy Efficient Computing R&D Roadmap Outline for Automated Vehicles  identifies areas of R&D necessary to attain the high computational performance with low power consumption that will be required to achieve automated driving in retail vehicles. The identified areas for R&D assume that all of the computational capacity resides on the vehicle and that the “all-in” total electrical power involved in AV computation would be 300 W for a commercially credible vehicle. Within this AV technology development landscape, four technical areas were identified for R&D investment:

  • I. Chips: Materials, Device, and Circuits;
  • II. Chips: Architecture, Safety, and Security;
  • III. Algorithms and Data Management; and
  • IV. Sensors Data Interface.

The Roadmap Outline identifies specific R&D problems within each technical area, with assessments given for the problems’ timing, impact, and relation to one another. Since many R&D needs for AVs are interrelated, the Roadmap Outline recommends that R&D activities within these four technical areas be co-designed and conducted in a holistic way to successfully meet the stiff technical challenge of developing energy efficient computing that enables highly automated vehicles.

Energy Efficient Computing for Automated Vehicles (EECAV) Workshop

Draft technical content for the Roadmap Outline was presented at an online EECAV workshop held with the AV and related technical communities on May 11 and May 12, 2021. The workshop presented the work of the Roadmap Team, discussed the general AV R&D challenges, and collected feedback on the R&D problems within the technical areas identified by the Roadmap Team. A total of 50 participants joined the invitation-only workshop, with experts representing automotive Original Equipment Manufacturers (OEMs), semiconductor companies, academics, and national laboratories. The sessions in the first day covered the motivation and purpose of the EECAV road mapping activities, an overview of the current status of the semiconductor and autonomous vehicle technology and projections, and an overview of the EECAV Roadmap Outline, including relevant scope, timeline and assumptions. The sessions in the second day of the workshop collected feedback from the workshop participants on several R&D problems in each of four technical areas. The feedback from the workshop is summarized in this report.

The summary of workshop feedback, agenda for the EECAV Workshop, and the workshop presentations can be downloaded via the links below.

Workshop Summary Report (PDF)

Workshop Agenda (PDF)

Workshop Presentations

Additional Documents

Chris Moen

Lennie Klebanoff