Innovative whole Vehicle-In the-Loop approach for Advanced Rider Assistance Systems calibration and verification: application to self-braking Adaptive Cruise Control

Abstract

On board motorcycles’ control logics are arising in number and complexity. It follows how on road testing, for correct response verification, becomes danger and time consuming. E.g. the mandatory ABS system, that shall be tested on both the high and the low road friction with installation of outriggers to prevent falling (UNECE Reg. No. 78). The testing complexity induces the acceptance criteria to be mostly limited to the subjective feeling of the tester. With the more challenging Cornering-ABS it is almost not possible to test due to limited availability of steering lanes with dedicated road frictions, where calibration and verification can be successfully accomplished. It is becoming harder and harder to objectify systems’ performances with reproducible and repeatable metric. This becomes even harder when dealing with several interacting control logics. The latest Advanced Rider Assistance System (ARAS)s, for example, such as the self-braking, radar-based Adaptive Cruise Control (ACC): there is a lack of real scenarios on which to execute calibration and verification tests and, even when available, the dangerousness in execution increases, and the subjective final assessment falters, bringing the rider’s psychophysical capabilities to the limits (N.Valsecchi, 2020). With multiple demanding control logics’ performances and state dependencies, the needed time for on-road calibration amplifies in duration and mileage; moreover, weather conditions and proving ground availability may frustrate the calibration and verification results. The possibility of doing most of the work in-door, in safe and repeatable conditions, despite adverse weather, already exists exploiting the HIL approach, that nowadays is becoming more and more popular. But still some limitations occur, when attempting to make multiple control logics working together, with real systems in the loop operating as on the real riding conditions. It may result more efficient to have the human in the loop able to ride the real vehicle, fully connected to a real time (RT) computer, reproducing road scenarios in the simulation environment. Human reactions and scenarios-dependent behavior can be realistically reproduced. Driven by the above motivations, the innovative idea to connect the whole motorbike in the RT simulations loop enables this investigation capability and allows to reduce the on-road riding risks to properly verify the behavior of all the involved systems operating together and include the real human response as well. The whole Vehicle-In the-Loop proposed in this work allows testing in full safe, manned or unmanned riding modes, through the exploitation of the automation suite, decoupling systems complexity and, finally, executing hard to replicate on-road scenarios otherwise. The final scope of the work is to stress the ARAS ACC implemented on a state-of-the art motorcycle, investigating the performances of the system when running the typical public road scenarios. To achieve the scope, the target motorbike has been fully connected to the Real Time PC, cheating the ECUs, now fed by simulated PWM wheels speed encoders and IMU signals computed in the virtual environment. The original radar has been by-passed and object-injection has been established in order to make the system believe the existence of traffic. The High-Fidelity vehicle model has been implemented, including proper tires models characterized on the target surfaces (D.Vivenzi, 2019)(E. Leo et Al. 2019). Parametrizable use cases have been implemented, enabling to test on-road realistic critical situations with moving traffic objects. More in details, a forward car driving at constant speed has been implemented as traffic object; the ego motorcycle, riding at higher speed, approaches the forward car. Different scenarios have been then analyzed. Scenario#1_the platooning: the logic capability to detect the forward vehicle, compute the time-to-collision and decelerate the ego vehicle till the platooning condition; repetitions in different ACC user-modes (e.g. very-short, medium, very long relative distance); scenario#2_the emergency brake: the real rider operates the throttle, reducing the safety relative distance; the self-braking logic activates only while releasing of the throttle to re-establish the safety distance, avoiding the front collision (if/when possible); emergency signs shall promptly alert the rider; scenario#3_the µ-drop: while self-braking to achieve the platooning condition, the road friction µ drops while braking, at a given speed, activating the ABS logic; the two logics interaction is observed and the implemented hierarchical criteria are chosen and verified (e.g. ACC promptly disconnection). Despite the scenario variability and complexity, both the repeatability and reproducibility in testing execution have been guaranteed, maintaining the same boundaries conditions (initial conditions, environmental conditions, road conditions, unevenness and tire-to-surface friction characteristics), allowing selective sensitivity execution and control logics parameters setting. Between the most relevant results, the possibility to tune the brake pressures, front and rear, in order to achieve the target vehicle deceleration and the relative distance and velocity with respect to the vehicle in front; the verification of the vehicle response against expectations; the interaction between ACC and ABS. Finally, the possibility to monitor how the logic behaves in case of functional faults (BS ISO 26262, ed. 2020). The separation of the effects allows to simplify and speed the analysis, confirming the effectiveness in performances improvement.