Human Factors for Limited-Ability Autonomous Driving Systems Jeremy Salinger, PhD GM Research and Development Jeremy.Salinger@gm.com Disclaimer A portion of this material is based upon work supported by the Federal Highway Administration under Agreement No. DTFH61-08-H-00006. Any opinions, findings, and conclusions or recommendations expressed do not necessarily reflect the view of the Federal Highway Administration.
Introduction Project Background Premise: The first autonomous systems introduced to the driving public will not be perfect Definition: Limited-Ability Autonomous Driving Systems (LAADS) can control vehicle speed and steering on public roads for substantial distances and times in some situations requires that the driver/operator intervene to assure a safe and comfortable trip Project Goals Investigate driver interactions with a Limited-Ability Autonomous Driving System (LAADS) Determine impact of a LAADS on driver visual attention to the driving task willingness to engage in secondary non-driving related tasks ability to respond to events Understand the factors that impact the effectiveness of alternative concepts of operation human-machine interfaces control transition strategies
Research Phases & Environments Research Phases I: Problem Identification Research II: Alternative Driver-Vehicle Interaction Concept Research III: Integrated Countermeasure Concept Research Study Environments Surveys Expert Panel Studies Driving Simulator Studies Track Studies GM Milford Proving Grounds Circular Track Radius: 1178m, 3865 ft. Circumference 4.6 Miles, 7.4 Km 5 Lanes Use or disclosure of data contained on this sheet is subject to the restrictions on the title page of this proposal.
Treatment Conditions Problem Identification ACC only (manual steering) Research Approach ACC and perfect Lane Centering (PADS) ACC and imperfect Lane Centering (LAADS) Limitation Events Used to Measure Impact of Different Systems on Driver Performance Poor lane markings system requests driver take control Lateral drift within lane with adjacent vehicle, Lateral drift within lane with no adjacent vehicle Excessive curve vehicle leaves lane Construction Lead vehicle hard braking
Behavior Sample
Head Turn Frequency by Duration Category during Simulator Study Frequency of Eccentric Head Turns by duration categories Denotes statistical significance compared to control group Use or disclosure of data contained on this sheet is subject to the restrictions on the title page of this proposal. 6
Riskier Tasks Tended to be Limited to LAADS Driving (those with relative risk values above 1) Percentage of Drivers Observed to Engage 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 67% 83% 92% 92% 92% 75% 42% 17% 58% 8% 83% 17% 67% 25% 58% 8% 8% 42% ACC Driving LAADS Driving 67% 50% 42% 25% 8% 25% 8% 50% 0% 0% 0% 0% 0% Relative risk values based on 100-Car Study (value of 1 is crash risk normally associated with typical driving)
Phase I Results Two targeted and complimentary experiments were performed: Simulator-based driving study with sixty-three subjects Test track study with twelve subjects. The studies quantify the difference in drivers behaviors between ACC and LAADS Increase the propensity of secondary task engagements, Increase riskier behaviors (those requiring extended off-road glances) such as reading, reaching for object in back seat, texting, Negatively impact the degree of visual attention drivers devote to the forward roadway (increases off-road glances).
Countermeasure Research Approach Phase II: Formative Studies High levels of interaction between researchers and subjects Phase III: Quantitative Countermeasure Performance Analysis Low levels of interaction between researchers and subjects Periods with and without secondary tasks
Track Study 4 Design
Percentage of Off-Road Glances Across Driving Mode for Short, Intermediate and Long Duration Time Bins 100% 90% 80% 94.59% 77.09% ACC Driving, with Secondary Task (n=3086) LAADS Driving with Countermeasures, and Secondary Task (n=2366) LAADS Driving No Countermeasures, and Secondary Task (n=1789) 70% 65.34% Percent of Sample 60% 50% 40% 30% 26.33% 20% 19.19% 10% 0% 5.41% 3.72% 0.00% 3 sec and Under 3.1 to 7.00 sec > 7 sec 8.33% Duration of Off-Road Glance
Relationship Between Glance Frequency and Mean Glance Duration for Off-Road Glances Under Each Driving Mode (n=26 Under Each Driving Mode) Total Number of Off-Road Glances (During Sample Period) 180 160 140 120 100 80 60 40 20 ACC Driving ACC Driving LAADS With Countermeasures LAADS No Countermeasures LAADS With Countermeasures LAADS with No Countermeasures 0 0 1 2 3 4 5 6 7 8 9 10 11 12 Mean Off-Road Glance Duration in Seconds
Conclusions Driver s engagement in secondary tasks is likely to increase when driving automation affords the opportunity LAADS systems should be designed to Clearly indicate the mode of operation Monitor driver s attention to traffic conditions and vehicle operation Encourage drivers to attend to forward roadway conditions HMI components that can improve driver attentiveness to the driving situation Means to engage driver in driving task when system is engaged Means to encourage visual attention to forward roadway Active alerts for system failures and limitations