<div>For heavy-duty vehicles equipped with automated mechanical transmission (AMT),
the control of automatic clutch torque is crucial during the start-up process.
However, the difficulty of controlling clutch torque is exacerbated by
differences in driver’s starting intentions, changes in vehicle mass, and road
gradient. Therefore, this article proposes the clutch starting torque
optimization strategy based on intelligent recognition of driver’s starting
intention, vehicle mass, and road gradient. First, an intelligent recognition
strategy is proposed based on the combination of data-driven and onboard
transmission control unit (TCU) algorithms, which improves the accuracy of
recognizing the driver’s intention to start as well as the vehicle mass and road
gradient. Based on the vehicle’s historical state data information, the
predictive model is trained offline using a long–short-term memory (LSTM)
network to obtain predicted parameter identification results, which are then
used to calibrate the computed values of the onboard TCU algorithm. Second, the
clutch torque optimization strategy is designed based on the driver’s starting
intention, while considering the effects of road gradient and vehicle mass on
the clutch starting resistance torque. The weight coefficients of the objective
performance function are adjusted according to the driver’s starting intention,
and the Pontryagin’s minimum principle (PMP) is used to solve the clutch target
torque. Finally, offline data training and real-vehicle testing are performed.
The results show that the optimization strategy can effectively reduce the
friction work and the degree of impact during the starting process, minimize the
clutch slipping time, and improve the smoothness of vehicle starting and driving
comfort.</div>