Efficient Neural Network Implementations on Parallel Embedded Platforms Applied to Real-Time Torque-Vectoring Optimization Using Predictions for Multi-Motor Electric Vehicles

Abstract

The combination of machine learning and heterogeneous embedded platforms enables new potential for developing sophisticated control concepts which are applicable to the field of vehicle dynamics and ADAS. This interdisciplinary work provides enabler solutions -ultimately implementing fast predictions using neural networks (NNs) on field programmable gate arrays (FPGAs) and graphical processing units (GPUs)- while applying them to a challenging application: Torque Vectoring on a multi-electric-motor vehicle for enhanced vehicle dynamics. The foundation motivating this work is provided by discussing multiple domains of the technological context as well as the constraints related to the automotive field, which contrast with the attractiveness of exploiting the capabilities of new embedded platforms to apply advanced control algorithms for complex control problems. In this particular case we target enhanced vehicle dynamics on a multi-motor electric vehicle benefiting from the greater degrees of freedom and controllability offered by such powertrains. Considering the constraints of the application and the implications of the selected multivariable optimization challenge, we propose a NN to provide batch predictions for real-time optimization. This leads to the major contribution of this work: efficient NN implementations on two intrinsically parallel embedded platforms, a GPU and a FPGA, following an analysis of theoretical and practical implications of their different operating paradigms, in order to efficiently harness their computing potential while gaining insight into their peculiarities. The achieved results exceed the expectations and additionally provide a representative illustration of the strengths and weaknesses of each kind of platform. Consequently, having shown the applicability of the proposed solutions, this work contributes valuable enablers also for further developments following similar fundamental principles.

An Approach of Feed-Forward Neural Network Throughput-Optimized Implementation in FPGA

“…Sigmoid activation function has been implemented with LUTs, where the input corresponds to the NN type while the output is specified by the table. Some of the configurations achieve an impressive two mega-samples per second, which even overcomes the FPGA performance results in the original article [31]. The approach potentially could outperform GPU implementation with six mega-samples if a bigger FPGA chip is used while having the benefit of lower power consumption.…”

“…The summary of different topologies and performance metrics of previous FFNN design approaches is provided in Table 1. An interesting use-case in terms of potential application of our developed approach is presented in [31], where FFNNs enhance vehicle dynamics for a multi-motor electric vehicle. The authors train a predictive NN for the estimation of the future slip values of each wheel for a batch of possible torque-vectoring set points.…”

“…To validate the developed methodology, a previously published [31] use-case has been chosen. The publication brings forward an interesting torque-vectoring application where a controller is proposed to estimate realistically unmeasurable signals of the vehicle's future dynamics.…”

“…These signal values depend on the different control actions, therefore, it is necessary to process as many data samples as possible. For more information on the algorithm, the reader is redirected to [31]. This work provides not only a realistic use-case for the throughput-optimized implementation of the FFNN, but the coefficients and data sets are available as well.…”

Machine Learning and Embedded Computing in Advanced Driver Assistance Systems (ADAS)

“…To address challenges and issues in the challenging area of torque vectoring on multi-electricmotor vehicles for enhanced vehicle dynamics, a neural network is proposed for batch predictions for real-time optimization on a parallel embedded platform with a GPU and an FPGA. This work will help others who are conducting research in this technical area [19].…”