Assistance systems tested: How reliable are lane keeping assistants?

Lane assist systems are divided into two categories. The first: Since July 2024, new cars have been required to be equipped with lane keeping assist systems, which steer the car back into the lane when the driver steers at or over a lane marking. The second category includes comfort assistants that independently guide the car in the lane. Many manufacturers refer to these as steering assist; Nio calls the function Lane Centering, and Volvo calls it Pilot Assist. Tesla has a special feature: the infotainment system lists "Steering Assist (Beta)" – a beta refers to the testing phase of software that is not yet fully developed. Tesla also lists the system in the settings menu under the heading "Autopilot Functions."

The lane guidance systems of the test vehicles Tesla Model Y , Volvo EX30 , Nio ET5 , and Mercedes E-Class are classified as Level 2 assisted driving systems, meaning they explicitly do not offer true autopilot functions. While Level 2 systems can brake, accelerate, and steer independently, the driver is always responsible. If the driver takes their hands off the steering wheel, the vehicle warns them visually, acoustically, or haptically. Depending on the model, the emergency assistant intervenes shortly afterwards and stops the car.

We typically test the quality and differences between steering assistants subjectively, but in this series of tests in collaboration with MdynamiX AG, we also do so objectively. The engineers install measuring equipment in the cars, calibrate it, and record the data during our test drives. The equipment records, for example, all vehicle movements, the speed traveled, and position coordinates via satellite technology (GPS, Glonass, Galileo, Beidou) at 1,000 Hz. The MdynamiX experts then overlay this data onto map material specially created for assistance system tests, which includes precise scans of the roads, including all lane markings. At the end, MdynamiX can evaluate, with centimeter precision, how the cars moved within the markings.

Our first test drive takes us along a highway, with the cruise control set to 130 km/h. In a straight line, the lane guidance of all systems works very well for the most part – although the orientation in the lane differs slightly between the test cars, subjectively they all drive precisely in a straight line. In the typically wide highway bends, the picture hardly changes: The steering inputs are at least clean, and in the case of the Mercedes, even extra-smooth. In any case, we editors rated the E-Class a corner better than its competitors after driving through each of the test sections several times. Objectively, the cars position themselves slightly differently here and there in the curves – but when driving in the far left lane, they only move slightly towards the guardrail and not towards the other cars in the lane. The Volvo has one serious drawback, irrelevant for our test: its cruise control system allows speeds of up to 150 km/h, but with Pilot Assist active, it cannot exceed the recommended speed. This is often simply a few km/h too low to keep up with the flow of motorway traffic.

The data for these diagrams were measured on a very winding federal highway at 120 km/h. The lane markings are approximately three meters apart.

It wasn't long ago that the lane guidance systems in Nio models were just steered in a jumble, even on highways—it's astonishing how much progress the Chinese brand has made in such a short time. However, not everything runs smoothly; for example, you often have to steer over resistance when changing lanes, even with the turn signal on.

On the highway, the steering assistant provides excellent guidance. And thanks to the sedan's excellent ergonomics, your hands automatically rest on the steering wheel with your elbows propped up—creating a kind of autopilot feeling in the Benz. The lane keeping assistant sometimes seems to understand when you intentionally cross lane markings, but often doesn't react when it definitely should.

The optional assistance packages ("Enhanced Autopilot," "Full Autonomous Driving Potential") include a lane-changing function that cannot be switched off. It works like this: Activate the indicator, wait for the lane change announcement, and then confirm it with a steering input. The Y then changes lanes, but often aborts the process. Manual lane changes deactivate the steering assist.

The lane guidance quality on the highway is adequate, but the maximum set speed of 130 km/h (81 mph) is less so, as congested cars typically move several km/h (6 mph) faster at that speed. Furthermore, it sometimes takes a very long time for the system to take over steering after activation. The "Please turn the steering wheel" prompts are also annoying, even when you're holding it with both hands.

For the next test, we switch to a four-lane federal highway, where the speed limit of 120 km/h is enough to pose a serious challenge to the cars at times. The test track starts with a comparatively tight downhill section. If we hadn't countersteered at the last moment, the Volvo would have crashed into the guardrail. Admittedly, in at least two corners on the test track, 120 km/h is quite a lot to hand over the steering to software – that would be unpleasant even if it worked perfectly. And yet, in such situations, you can expect warning messages that never appeared in the Volvo. The EX30 also stumbles further along the route, which then leads through more relaxed bends that are nevertheless sharper than those on most motorways.

The Mercedes repeatedly approaches the lane markings on the same stretch of road, sometimes even crossing them. In these situations, it warns and steers back into the lane – while this may not look like good performance, at least it doesn't radically undermine confidence. It stays in its lane much better on large sections of the test route, and handles the uphill return leg with confidence – unlike the Volvo, which struggles there in most of the test runs. The Nio manages lane tracking well at times on the main road, but depending on the test run, in some areas it drives directly next to the solid lane markings on the outer edge. These lines are very close to the guardrails, which is why the feeling of safety suffers particularly. Consequently, the comfort factor of the function is also significantly reduced.

The Model Y stays in its lane most reliably. The diagrams on page 65 show a section of the test track where the Tesla stays within 30 centimeters of the lane center in curves – its competitors sometimes deviate by half a meter or more. While no test car behaves the same across all test runs, the Tesla confirms its good performance most frequently. However, it falls flat because the steering assist beta doesn't allow manual lane changes, only semi-automatic lane changes, which takes time, rarely works well, and is therefore frustrating. In the other cars, the lane guides are inactive while signaling.

We conducted this and other assistance system tests with MdynamiX AG, a company that cooperates with the universities of Munich and Kempten. It specializes in evaluating such systems for customers in the automotive industry.

The test drives take place on public roads for which MdynamiX has precise maps, including data on the lane markings. To determine the distance from the cars to the markings, several points are first measured on each test vehicle. These include the exact position of the outer tire edges – this is necessary to determine the distance to the lane markings. All four cars carry two antennas (Wi-Fi, GPS) on the roof, and the interiors are equipped with special measuring technology that records all vehicle movements. The subsequent data analysis is carried out using the MXeval software developed and distributed by MdynamiX.

We also conducted a beta test with MdynamiX, ​​as they are currently developing a test procedure for the lane-keeping assistance systems in the first category – those designed to help when driving over lane markings. In the Model Y cockpit photo, you can therefore see the camera on the sun visor and the QR code-like sticker on the steering wheel. The collected data isn't ready for publication yet, but the behaviors we observed on the test routes are.

We tested the systems by waiting without steering input until the cars moved toward the dashed and solid lines. Alternatively, we approached gentle curves without steering at the entrance. The result? The Nio reacted most reliably, but often significantly too late. The Mercedes steered back into the lane by far the most smoothly, but most often didn't react at all. And Tesla and Volvo? Somewhere in between. In any case, our test confirms that the lane-keeping systems can help and do so in practice, but only with limited reliability.