Explicit MPC motion cueing algorithm for real-time driving simulator

Abstract

The performance of the driving simulator depends on the efficiency of the motion cueing algorithm. An explicit model predictive control was established recently for the motion cueing algorithm. The complexity of the explicit solution increases manifold when the human vestibular model is considered. This paper focuses on the complexity reduction of explicit solution using low complexity contractive sets for the motion cueing algorithm. The low-complexity explicit controller is formulated for the efficient control of the motion cueing algorithm for the ULTIMATE driving simulator at Renault.

A Motion Cueing Algorithm Based on Model Predictive Control Using Terminal Conditions in Urban Driving Scenario

“…As the provided model [23] was not able to tilt the driver's head based on human vestibular model, the SBMP could easily hit the boundaries. An explicit MPC was employed to generate the MCA for a SBMP with 2 degree of freedom (2-DoF) by Fang and Kemeny [24]. They employed the multi-parametric toolbox of MATLAB to find the affine function.…”

“…The real-time application of the time-varying MPC is hard to be implemented due to the high computational load for updating the MPC states based on the present configuration of the end-effector. All the above-mentioned research studies on MCAs based on MPC [23][24][25][26][27][28] only considered when the vehicle is driving on the roads without any traffic jam or traffic lights, and without applying much starting and stopping signals. Using the urban driving scenarios which involve traffic lights, interaction with other drivers and pedestrians, roundabouts and puddles of the road can destabilise the previous models [23][24][25][26][27][28] and cause undesired platform fluctuations due to the absence of terminal conditions (weights and states).…”

“…All the above-mentioned research studies on MCAs based on MPC [23][24][25][26][27][28] only considered when the vehicle is driving on the roads without any traffic jam or traffic lights, and without applying much starting and stopping signals. Using the urban driving scenarios which involve traffic lights, interaction with other drivers and pedestrians, roundabouts and puddles of the road can destabilise the previous models [23][24][25][26][27][28] and cause undesired platform fluctuations due to the absence of terminal conditions (weights and states). Therefore, they [23][24][25][26][27][28] should be penalised conservatively to respect the workspace area of the SBMPs, but this hypothesis reduces the better usage of the workspace area and enhances the motion feeling error for the driver of the SBMP.…”

“…Using the urban driving scenarios which involve traffic lights, interaction with other drivers and pedestrians, roundabouts and puddles of the road can destabilise the previous models [23][24][25][26][27][28] and cause undesired platform fluctuations due to the absence of terminal conditions (weights and states). Therefore, they [23][24][25][26][27][28] should be penalised conservatively to respect the workspace area of the SBMPs, but this hypothesis reduces the better usage of the workspace area and enhances the motion feeling error for the driver of the SBMP. Also, the stability of their models [23][24][25][26][27][28] cannot be guaranteed as the end-effector cannot turn back to the natural position at the end of the prediction horizon without using the terminal states.…”

“…Therefore, they [23][24][25][26][27][28] should be penalised conservatively to respect the workspace area of the SBMPs, but this hypothesis reduces the better usage of the workspace area and enhances the motion feeling error for the driver of the SBMP. Also, the stability of their models [23][24][25][26][27][28] cannot be guaranteed as the end-effector cannot turn back to the natural position at the end of the prediction horizon without using the terminal states. Terminal conditions only implemented in the time-varying MCA based on MPC by Qazani et al [31,32], however the time-varying MPC cannot regenerated the real-time motion cues in heavy traffic jam as it has a high computational burden to extract the optimal input signal.…”