Optimised Cabin Configuration – October 2021

Result from partner:

  • COV (James Brusey)

Result has been achieved in October 2021 : month 48 of the project.

Short introduction
A particularly innovative part of the DOMUS proposal is the idea that we can optimise the cabin configuration to further reduce the energy consumption of the climate control system while maintaining safety and comfort. The basis for this work is the development of a high-speed virtual environment that allows changes to the environment to be explored in detail. In particular, this virtual environment needs to support the configuration changes being explored. With the delivery of Result 2‘s final report (Report 2.3), DOMUS has demonstrated this methodology for optimising over 4 main features: heated seats, windshield heating, the smart vent, and radiant panels. Each feature has some trade-offs associated with it (such as additional power consumption).

This part of the project relates to the following DOMUS objectives:

  • Radical redesign of EV cabin with understanding of driver comfort perceptions, technological innovations (Results 2–5), and vehicle level performance optimisation.
  • Virtual development process for cabin redesign and multidimensional optimisation.
  • Measure impact of cabin redesign and advanced efficiency increasing interventions on vehicle efficiency, range, comfort, and safety via comparison to baseline.
  • Development and application of an approach to simplify the virtual assessment of comfort and efficiency and perform optimisation work.

The cabin optimisation approach makes use of Bayesian optimisation to help identify the set of features that perform the best together.

  1. The optimisation selected the smart vent and window heating; while removing the radiant panels, seat heating and new air modes. Note however that there were a wide set of possible configurations with indistinguishable results. On the other hand, there were clearly some configurations that performed much more poorly. The interaction between configurations is clearly indicated in the results and it is not useful to examine individual configurations on their own.
  2. Key threats to validity of the cabin optimisation approach are:
    • Inaccuracy in the simulator could affect the transfer of the optimisation results to the real world.
    • The optimisation approach is sensitive to slightly sub-optimal control performance and more extensive tuning of the optimised controller than used here might help to disambiguate the best performing configurations.

Graph: simulation of interaction between configurations