Drivers’ Manoeuvre Classification for Safe HRI

Abstract

Ever increasing autonomy of machines and the need to interact with them creates challenges to ensure safe operation. Recent technical and commercial interest in increasing autonomy of vehicles has led to the integration of more sensors and actuators inside the vehicle, making them more like robots. For interaction with semi-autonomous cars, the use of these sensors could help to create new safety mechanisms. This work explores the concept of using motion tracking (i.e skeletal tracking) data gathered from the driver whilst driving to learn to classify the manoeuvre being performed. A kernel-based classifier is trained with empirically selected features based on data gathered from a Kinect V2 sensor in a controlled environment. This method shows that skeletal tracking data can be used in a driving scenario to classify manoeuvres and sets a background for further work.

Drivers' Manoeuvre Prediction for Safe HRI

“…Initially, the idea of manoeuvre classification was revised. Models scored precision and recall scores higher than 95%, in contrast to 88% on previous work [10]. Results were validated for known and unknown test subjects.…”

“…The main contributions of this work are twofold: firstly, we replicate, improve and validate our previous work on drivers' manoeuvre classification using body posture data from a designed experiment [10] that includes drivers of different heights, seat preferences and levels of driving expertise. Secondly, we introduce a manoeuvre prediction scheme that can be learned with a data-driven approach from a general set of reduced manoeuvres, directly linked to HRI.…”

“…The trend must be kept for a certain amount of time-steps T s and theoretically diminish as the time-steps decrease. Figure 4 shows the statistical analysis (average, confidence interval, outliers) of the used driving features (see [10]) for one individual test subject and for all grouped test subjects inside a time-window r. For the case of a single individual, data dispersion is less at T s = 0, with a considerable number of outliers that increase as T s approaches -90. For the case of all test subjects, data are highly variable and show big amount of outliers for all time-steps.…”

A Systematic Literature Review About the Impact of Artificial Intelligence on Autonomous Vehicle Safety

“…The second largest number of papers (13) found encompasses studies related to Navigation and Control. They are mostly related to techniques necessary to ensure the proper autonomous navigation and control capabilities required by AVs, such as: remote-controlled semi-AV based on IoT [38]; adaptive pre-crash control [39]; safe trajectory selection [76]; AV following another car driven by a human pilot (Trailing) [40]; safe navigation [41]; heuristic optimization algorithm for unsigned intersection crossing [42]; vehicle coordination [43]; maneuver classification [44]; learning to navigate from demonstration [45]; AV movements optimization in intersection [46]; learning and simulation of the Human Level decisions involved in driving a racing car [47]; path tracking [48]; and fuzzy-logic control approach to manage low level vehicle actuators (steering throttle and brake) [49].…”

“…Considering some specificities of autonomous truck and its risks, at least a few more studies about the topic could be expected. [64], [67], [70], [61], [65], [69], [62], [58], [63], [71], [72], [74], [68], [ [39], [18], [32], [31], [33], [26], [28], [30], [55], [52], [29], [ [75], [41], [29], [20], [44], [35], Prediction of adc vanced driver assistance systems (ADAS) remaining useful life (RUL) for the prognosis of ADAS safety critical components Pedestrian Detection; How to "automate" manual annotation for images to train visual perception for AVs Road junction detection; [52], [27], [37], [30] Bayesian Artificial Intelligence…”