Maintaining driver attentiveness in shared-control autonomous driving

TL;DR

The paper tackles driver attentiveness in shared-control ADS (Level 2–4) by embedding a MAP E loop (monitor, analyze, plan, execute) within ALKS, using biometrics and vehicle data to predict driver takeover with a DNN (DeepTake) and planning via a formally verified controller. It formalizes the attentiveness problem with multi-objective goals (nuisance, progress, risk) over a horizon $T$, and proposes a parametric CTMC design space whose controllers are synthesized to Pareto-optimality using probabilistic model checking and EvoChecker. Core contributions include the Safe-SCAD architecture, the integration of DeepTake with robust verification, and a scalable synthesis workflow that yields Pareto fronts and actionable controller configurations. The work aims to reduce minimum-risk maneuvers and improve journey progress while maintaining safety in Level 3/4 ADS, with plans to extend to Level 2 and professional testing, and to validate via driving simulators and larger design spaces.

Abstract

We present a work-in-progress approach to improving driver attentiveness in cars provided with automated driving systems. The approach is based on a control loop that monitors the driver's biometrics (eye movement, heart rate, etc.) and the state of the car; analyses the driver's attentiveness level using a deep neural network; plans driver alerts and changes in the speed of the car using a formally verified controller; and executes this plan using actuators ranging from acoustic and visual to haptic devices. The paper presents (i) the self-adaptive system formed by this monitor-analyse-plan-execute (MAPE) control loop, the car and the monitored driver, and (ii) the use of probabilistic model checking to synthesise the controller for the planning step of the MAPE loop.

Figures (6)

• Figure 1: Safe-SCAD driver attentiveness management approach for ALKS with $n=3$ attentiveness levels (attentive, semi-attentive, and inattentive)
• Figure 2: Safe-SCAD analysis MAPE stage
• Figure 3: The driver attention level (A=attentive, S=semi-attentive, I=inattentive) depends on the predicted driver takeover time and quality, on the speed of the vehicle, and on the verified robustness of the DeepTake input region that the sensor data belongs to. The diagram applies to a positive takeover intention (i.e., driver responsive to a control-transition demand); when DeepTake predicts a negative intention, the driver state is deemed inattentive.
• Figure 4: Safe-SCAD controller design space for the driver attentiveness management problem with three driver attentiveness levels ($L=\{\mathsf{'A'},\mathsf{'S'},\mathsf{'I'}\}$, where $\mathsf{A}$=attentive, $\mathsf{S}$=semi-attentive and $\mathsf{I}$=inattentive), two alerts ($A=\{0,1\}^2$) and two speed levels ($V=\{0,1\}$). In the initial state (indicated by an incoming arrow) the driver is attentive, the controller is inactive, the alerts $a=(0,0)$ are inactive, and the car drives at nominal speed level $v=0$; for brevity, this combination of alert activations and speed level is denoted $a\_v=000$ in the diagram.
• Figure 5: Fragment of EvoChecker-encoded controller design space for $m=2$ independent alerts and $q=2$ speed levels