Share

ANN ARBOR, Mich., May 9, 2026 — A haptic steering wheel developed at the University of Michigan, with funding from the Toyota Research Institute, lets drivers and semi-autonomous systems negotiate over upcoming maneuvers using only touch. The wheel inflates small zones to show the car’s plan; the driver can squeeze those same zones to agree, object, or prompt a revised action. New peer-reviewed findings in Human Factors show the two-way interface improves driving accuracy and helps people trust automation again after a system error.
Most driver-assist systems today communicate in one direction. They beep, flash a warning, or vibrate briefly, and the driver is expected to interpret the signal while also keeping eyes on the road. When the driver disagrees with what the automation wants to do, the usual option is to physically fight the wheel. The Michigan team calls that model stressful and confusing. Worse, they say, it can lead people to switch off assistance features completely, which removes any safety benefit the system might offer.
The new wheel is designed to turn that one-way alert into a tactile conversation. Two haptic zones sit at the 10 and 2 o’clock positions, the spots where drivers most often rest their hands. A ring of soft white pads along the rear of the rim also inflates to deliver signals. The same physical surfaces work in both directions: the system uses them to send its intention, and the driver uses them to send a response.
“We placed two expandable haptic zones on the steering wheel at 10 and 2,” said Hannah Báez, a robotics Ph.D. candidate and first author of the study. “The zones can expand or contract to show what the automation intends to do, and the driver can squeeze them to agree or disagree before the vehicle acts.”
That negotiation process happens quickly. When the semi-automated system plans to turn right, the right haptic zone expands and the rear pads on the corresponding side inflate. If the driver wants to challenge the plan, squeezing that zone registers an objection, giving the automation a chance to adjust before the maneuver starts. When an obstacle appears in the driver’s intended path, the wheel pulses from the side where the hazard is located.
Tactile communication has a practical advantage: the information arrives exactly where the driver’s hands are already positioned. The driver does not need to glance at a screen or listen for a chime in a noisy cabin. In that sense, the interface shortens the distance between a signal and an action.
To test the concept, the research team built a three-screen immersive driving simulator with a real seat, steering wheel, and pedals. Participants approached an upcoming turn in scenarios where the driver and the automation sometimes agreed and sometimes disagreed about which direction to take. A sudden obstacle could also appear in either situation.
The two-way haptic interface was measured against two baselines: a condition with no haptic information and a condition with only a one-way display of automation intent. The negotiation design performed clearly better than both.
“Drivers displayed smoother, more accurate driving paths, fought less with automation, and used less braking with more confidence in their maneuvers,” said Brent Gillespie, a U-M professor of robotics and the study’s senior author.
The subjective results were just as positive. Participants reported lower mental workload, less frustration, and less physical demand when they could talk back to the system through touch. For automakers, that could mean fewer cases of drivers disabling assistance because they feel the car is not listening.
The study also examined what happens after automation fails. In simulations where the system missed an obstacle or issued false warnings, people using the negotiation interface regained trust in the automation much faster than those using conventional one-way alerts. The authors attribute the recovery to a more collaborative relationship: the system states its reasoning, the driver can question it, and the two sides settle conflicts before they become dangerous.
That teamwork model represents a departure from current practice. Instead of a machine issuing commands and a human reacting, the Michigan design treats the human as a partner who can influence the car’s next move. The researchers argue that transparency and responsiveness are essential for maintaining appropriate trust in partial automation.
There is also a warning in the data. After several successful negotiations, some drivers became overly trusting of the automation. That pattern aligns with a broader concern in self-driving research: an interface that feels intuitive and reliable can lull drivers into paying less attention. The authors say more work is needed to understand when overconfidence appears and how to keep trust calibrated.
Production vehicles already use haptic alerts for things like lane departures, usually through vibrating seats or steering wheels. But those warnings are one-way. The Michigan wheel is different because it uses the same haptic elements as both a display and an input channel, enabling give-and-take rather than passive notification.
The research team includes co-authors Haochi Pan and Nadine Sarter from the University of Michigan, along with Jean Costa and John Gideon from the Toyota Research Institute. The technology is covered by U.S. Patent No. US12227190B2, and the University of Michigan is seeking commercial partners to bring the wheel to market.
The team’s long-term goal is not just a more comfortable steering wheel. It is a model for coordination between humans and automated systems, especially in Level 2 and Level 3 vehicles where the driver remains responsible. At a moment when regulators and safety advocates are asking for clearer communication about automation limits, a steering wheel that can both listen and respond offers a promising answer. The study appears in Human Factors under the DOI 10.1177/00187208261483656.









