Energy and Time Optimal Autopilot for Electric Vehicles Performing Ackerman Cornering


Ahiska K., ÖZGÖREN M. K. , LEBLEBİCİOĞLU M. K.

IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, 2022 (Journal Indexed in SCI) identifier identifier

  • Publication Type: Article / Article
  • Publication Date: 2022
  • Doi Number: 10.1109/tits.2022.3159747
  • Title of Journal : IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
  • Keywords: Autopilot, Optimal control, Wheels, Mathematical models, Electric vehicles, Vehicle dynamics, Torque, Mathematical modelling, cornering autopilot, optimal control, electric vehicle, Pareto-front analysis, TRACKING CONTROL, STEERING CONTROL, MODEL, CAR, DESIGN, LIMITS

Abstract

This paper studies energy and time optimality of electric vehicles during constant Ackerman steering along a quad-circle, and proposes an autopilot assimilating the optimal results. The energy and time optimal solutions satisfying the steering and battery limitations are generated and a Pareto-front analysis is carried out with multi-objective optimization using NSGA-II algorithm. In the autopilot design, the indicators for the energy and time optimality performances are merged in a vehicle status vector. At each control cycle of the optimal drives, the torque commands and the vehicle status vectors are stored and later clustered using the k-means algorithm. At each cluster centre, a pair of the vehicle status vector and the control command vector is acquired and these pairs designate the rules that generate the unique optimal control command associated with a particular vehicle status. A convex combination of these rules constitutes the autopilot design. The proposed autopilot is tested against the energy-time Pareto-front extracted, and it is observed that in terms of the performance measures, the optimality has been preserved. Additionally, for various weightings of time and energy optimality objectives, the performance of the autopilot is compared with that of optimal controllers with weighted objectives. The results verify the quality of the autopilot design using a decision-making process over a rule set inferred from the solutions acquired with optimal solutions.