The Problem With Beeps and Tugs
Drivers of semi-autonomous vehicles today typically receive information through a mix of beeps and flashing warning screens, according to the research team. When a driver disagrees with a decision the computer has made, the usual recourse is to wrestle the wheel away from it. The researchers say this combination tends to create confusion and raise stress, and it can push drivers to disable systems that might otherwise help them.
The finding frames a practical concern for automakers. A driver-assist feature that is turned off provides no safety benefit, and the team’s work starts from the premise that overwhelming or unexplained interventions are one reason that happens.
Two Zones, Two Directions
The design adds two haptic zones to the wheel, positioned at 10 o’clock and 2 o’clock, that can expand or contract to express what the automated system intends to do. A ring of white modules along the rear of the wheel, where a driver’s fingers naturally rest, inflates to deliver the signal.
“To improve driver communication with automation, we added two haptic zones onto a steering wheel at 10 o’clock and 2 o’clock that can expand or contract to express the intention of the automated system,” said Hannah Báez, a robotics Ph.D. student and first author of the study. “Drivers could agree or disagree with the automated system by squeezing the haptic zones, which allowed them to negotiate with the system before a given action to resolve any conflict.”
The exchange runs in both directions. The wheel tells the driver what the system plans to do, and the driver answers with a squeeze of their own.
Testing the Turn
To evaluate the concept, the team placed drivers in a simulator, a three-screen immersive display surrounding a steering wheel, seat and pedals. Participants approached an upcoming turn in scenarios where the driver and the automation might agree or disagree about which way to go. In both situations, the simulation could add a sudden obstacle.
As a driver neared the turn, the semi-autonomous system inflated the left or right haptic zone to indicate the direction it intended to take. A driver who disagreed could squeeze that zone to register an objection and negotiate a different direction. If an obstacle appeared in the driver’s planned path, the wheel pulsed on the side where the obstacle was located.
The researchers compared the two-way system against conditions with no information and against one-way displays of intention. The negotiation interface showed a clear improvement over both.
Smoother Paths, Less Fighting
“We found that drivers drove better when negotiating with the automated system through haptic feedback,” said Brent Gillespie, a U-M professor of robotics and the study’s senior author. “Drivers displayed smoother, more accurate driving paths, fought less with automation and used less braking with more confidence in their maneuvers.”
Participants also reported a lighter burden. The negotiation interface reduced driver workload, with significantly lower self-reported effort, frustration and physical demand.
Rebuilding Trust After a Mistake
Perhaps the most notable result concerned what happens when automation gets something wrong. After the simulated driving system made errors, such as missing obstacles or issuing false warnings, drivers using the negotiation interface regained their trust in the system much faster than those using traditional one-way communication.
The researchers attribute the difference to teamwork. With better communication, greater transparency in decision-making and the ability to adjust to feedback, the driver and the system function more as partners than as two parties taking turns behind the wheel.
The team also flagged a caution. Some drivers became overly trusting after successful negotiations, which the researchers say underscores the need for further study of trust in systems of this kind.
From One-Way Alerts to Dialogue
One-way information is the current norm for communication with autonomous vehicles. The Michigan work argues for a different model, in which driver and automation exchange intent, agreement and action in a two-way dialogue.
The study’s additional authors are Haochi Pan and Nadine Sarter of the University of Michigan, and Jean Costa and John Gideon of the Toyota Research Institute. The technology is covered by U.S. patent US12227190B2, applied for with the assistance of U-M Innovation Partnerships, and the team is seeking partners to bring it to market.
The study, titled “Haptic shared planning and control: enabling coordination of future actions in human-autonomous vehicle teams through a haptic negotiation interface,” carries the DOI 10.1177/00187208261483656.