Research \\ Automated Driving \\ Human-Machine Interaction

Human Factors in Automated Driving

The Design Ergonomics Research Group of Loughborough Design School (United Kingdom) is conducting this project as part of the university's efforts to remain one of the UK's leading research institutions (currently ranked 4th for the overall quality of research, with 93% of the school's research impact being classed as internationally excellent). Our research group is currently looking for study participants, please click here to jump to our call for participants.

Introduction

Driving automation is one of the key trends in technology, having recently surpassed its expectation peak in Gartner's Hype Cycle for Emerging Technologies (Gartner 2016). In the the buzz around enabling technologies such as machine vision that have been advanced at breakneck speed in recent years, one easily overlooks the challenges associated with the operation of suchlike vehicles. We will likely have to wait another decade before steering wheels are disappearing in vehicles (ERTRAC 2015) and face the same fate as floppy disks in computing - a degadration from a crucial part of the system to a metaphor. Until then, the human driver and the automated system will have to cooperate in one way or another. There will, inadvertently, be situations that cannot be handled by the system and the driver will be prompted to take over within seconds.

An excellent example and precursor for challenges to be expected regarding driving automation are those experienced in aviation, a domain that has a decades-long practical record of automation. Many will be surprised to read that - on average - pilots of commercial aircraft are in manual control for only about 2 minutes, during takeoff and landing. The vast majority of the flight is performed by the autopilot. But:

What happens if the autopilot unexpectedly fails?

Pilot and co-pilot need to identify the problem, decide and finally act upon appropriate measures. Several incidents and accidents, however, illustrate that this is not always successfully done. A recent example is the crash of the Air France flight AF447 from Rio de Janeiro, Brazil, to Paris, France, that resulted in the loss of life of all aboard. In this particular case, the autopilot disengaged as a result of malfunctioning (likely obstructed) Pitot tubes. The pilots reacted incorrectly and manoeuvred the aircraft into a stall - but is human error really the major culprit? In the report analysing the incident (BEA 2012), the responsible French government agency identified the human-computer interface as a contributing factor, specifically misleading or missing feedback regarding the system state.

Decades of research within the aviation domain have aimed at analysing the Human Factors issues associated with the interaction between a human operator and an (imperfect) automated system. One of the key conclusions of this research is the Out-of-the-Loop Performance Problem, a term coined by Mica Endsley in the 1990s. The term describes the problem that human operators who have been completely detached from the control loop of executing a particular task (e.g. driving) are slow to detect the problem that led to a disengagement of the automation and, as a consequence, are unlikely to perform appropriate actions. You might interject at this point and argue that driving a car and flying an aircraft are not exactly the same piece of cake. I agree. The problems are potentised in driving. While pilots go through a rigorous selection process and are repeatedly trained throughout their career, almost anyone can drive a vehicle and is - in most cases - only trained once. This requires an extremely easy and intuitive-to-use as well as potentially adaptive human-machine interface. Further, drivers will only have seconds to arrive at a particular decision while pilots often have minutes. The illustration below summarises the key challenges associated with a takeover process.


Our research project aims at developing a data-driven, user-focused interface for automated driving systems that require the driver to take over control in critical situations. The research has two main objectives: improve driving safety and enhance the user experience. An increase in driving safety without the consideration of user experience is fruitless. In aviation, where safety is the foremost criterium, pilots may only manually control the aircraft for a few minutese. This, however, does not mean that they can go to sleep or watch movies for the remaining time. Instead, they constantly monitor the system's behaviour, go through checklists, etc. Small, specifically designed tasks to keep the pilots engaged to some degree. Does this work with passenger cars? It might be the safest way, but does it also grant a sufficient user experience? Our project aims at finding a compromise that ensures safety without neglecting experience.



Call for Participants

The involvement of potential users is key to the successful development human-machine interfaces. We design experiments to collect and evaluate data which helps us to find out if our hypotheses are valid. The next study is scheduled for December 2017/January 2018, it involves:

Please contact A.Kunze@lboro.ac.uk if you have any questions. You are required to hold a valid driving license and must be at least 18 years old. The whole experiment will be performed in the Driving Lab of Loughborough Design School. Important: Please do not read any of our publications prior to the study as this might affect the results. Sign up below to indicate your interest!



Other Projects