Model Predictive Control (MPC) has emerged as a pivotal technology for optimizing control tasks in autonomous driving, particularly within Adaptive Cruise Control (ACC) systems. However, the manual tuning of MPC cost function weights and prediction horizons remains a significant challenge. In this paper, we introduce a novel framework that combines Deep Reinforcement Learning (DRL) with MFC to dynamically tune both the weight parameters and prediction horizon in real time. This approach, referred to as the Weights and Prediction Horizon Varying MPC (W-PH-MPC), overcomes traditional MPC limitations by utilizing proximal Policy optimisation and Deep Deterministic Policy Gradient (DDPG) algorithms to adjust control parameters. We evaluate the effectiveness of our approach through simulations in vehicle-tracking scenarios. Simulation results show that the adaptive MPC-RL controller achieves better tracking performance, without compromising power consumption, and lowers longitudinal jerk compared to a fixed-parameter MPC baseline, resulting in smoother and more efficient vehicle behavior.

Backtracking is generally used for interior-point methods (IPMs) to keep some optimization parameters within a defined boundary. The idea is to reduce the step size such that the optimality conditions are fulfilled. However, backtracking might impede progress toward the optimal point, requiring longer solving time and higher iteration numbers, which are undesirable in real-time applications. In this paper, we present a novel algorithm based on an interior-point method that, in addition to accepting infeasible start guess points, liberates the Lagrange multiplier associated with the inequality constraints from backtracking. Accordingly, a new strategy for updating the weighting factor of the barrier function is proposed. The performance of this new IPM is benchmarked with well-known optimization solvers for solving both linearly and quadratically constrained quadratic programming (QP) problems, which are formulated based on an application of model predictive control (MPC) in the automotive industry. Furthermore, the new IPM has been implemented on embedded hardware and validated in an experimental real vehicle. The testing results show that the new IPM takes considerably fewer iterations and computational time to solve QP problems than other tested solvers.

This paper presents a novel formulation for scheduling the charging and discharging of electrical vehicles (EVs) in microgrids, considering a conditional minimum energy limit. The idea of utilizing EVs as storage units through vehicle-to-grid technology helps to stabilize the microgrid, especially with the intermittent power production of renewable energy resources. However, discharging the EV battery to very low levels can have negative impacts on its health and may not be satisfactory for the EV owner. Optimization-based methods show promise for scheduling, but the minimum energy constraint can make the problem infeasible when the initial EV energy is below the minimum limit. In this paper, we discuss existing approaches and their drawbacks in handling this issue. To address these drawbacks, we propose a novel approach that models the battery as two separate energy reservoirs. A case study is presented to demonstrate the effectiveness of our proposed method. Noteworthy, this concept can be applied not only to EV batteries but also to stationary batteries within microgrids.