Direct Data-driven Predictive Control: A Computationally Efficient Alternative to DeePC for Eco-driving in Mixed Traffic Flows
Date:
Presented our work on Direct Data-driven Predictive Control (D3PC) for eco-driving in mixed traffic flows.
This work proposes a computationally efficient alternative to conventional Data-enabled Predictive Control (DeePC). By reformulating the data-driven prediction mechanism into an explicit direct data-driven model, D3PC avoids the high-dimensional latent decision variable used in DeePC and makes the online optimization complexity nearly invariant to the size of historical data.
The method was applied to eco-driving control of connected and automated vehicles in mixed traffic flows with heterogeneous human-driven vehicles. Simulation results show that D3PC is orders of magnitude faster than DeePC while achieving improved energy efficiency across diverse driving scenarios.