A study by the IIT-CNR, in collaboration with MIT and University of Pisa, proposes the TLM protocol to make vehicles truly independent through communication with the road.
Are the self-driving taxis traveling on U.S. roads truly independent of human control and safe? It would appear not, according to admissions from the companies operating them. This was revealed by a safety inquiry into autonomous vehicles launched by the U.S. Congress.
On February 4th, both Tesla and Waymo — the Alphabet group company that manages fleets of “robotaxis” in numerous U.S. cities — were called to testify before the commission. The commission’s stated goal is to shed light on the transparency of these technologies and the safety risks associated with communication latency and the use of remote personnel to manage road emergencies.
While the news was covered by the italian newspaper Corriere della Sera and commented on by numerous international outlets, the hearing featured significant testimony from Waymo’s Chief Security Officer, Mauricio Peña. He stated that his fleet’s vehicles are not yet fully autonomous; in situations of uncertainty—such as confusing signage or unmarked construction sites—the system relies on “remote pilots”, primarily located in the Philippines, who intervene via software to resolve situations by interpreting the scenario from a distance.
The company has denied the hypothesis of actual “remote driving”, emphasizing that operators act only as consultants for the on-board software. Nevertheless, Waymo’s operational model has sparked major controversy and crucial doubts regarding cybersecurity and data protection, alongside legitimate questions about signal latency and human error.
The problem is well-known within the automotive and cybersecurity scientific community, and a solution may come directly from research. Researchers Ilaria Matteucci and Marco de Vincenzi from IIT-CNR, along with Chiara Bodei (University of Pisa) and the MIT Auto-ID Lab team led by Sanjay Sarma and Stephen S. Ho, have developed TLM (Time-Logic-Map) for this purpose. The study was presented last October at the VTC (Vehicle Technology Conference) Fall 2025 in Chengdu.
TLM is a spatial messaging language that allows infrastructure (intersections, construction sites, traffic lights) to “explain” to the vehicle what to do. The system transmits all data necessary for navigating a specific urban space to the vehicle via messaging, based on three logical layers: Map provides precise 3D geometry of the road segment, useful even where GPS signals are absent (such as “urban canyons”)
Logic encodes behavioral rules (who has the right of way, which lanes are active) for that specific road stretch; Time synchronizes the vehicle with real-time traffic light cycles.
“The ongoing parliamentary inquiry in the United States highlights a systemic criticality: we cannot have widespread, safe, and efficient autonomous driving if every vehicle requires a remote human supervisor,” states Ilaria Matteucci. “With TLM, the intelligence needed to resolve an impasse does not reside in a call center on the other side of the world, but in the intersection itself. If the road communicates its logic in a standardized and secure way, the robotaxi can operate in total safety even in low visibility or unforeseen contexts, drastically reducing the need for human intervention.“
“Roads were created for human beings, and autonomous vehicles still have to interpret signals designed for human vision through complex sensors and algorithms,” notes Prof. Chiara Bodei (University of Pisa). “Time-Logic-Map, instead, introduces a spatial language for machines that directly communicates the geometry, logic, and timing of the road, lightening the perceptual load in the most complex contexts”.
“Unlike current systems that download heavy maps via the cloud, TLM transmits lightweight, local messages via V2X (Vehicle-to-Everything),” comments Marco De Vincenzi. “This approach not only protects privacy—as there is no video streaming to remote operators—but also eliminates dangers related to internet connection ‘lag,’ since the decision happens instantaneously between the infrastructure and the car.“
The study involves the creation of a digital twin to simulate the behavior of autonomous cars, followed by real-world testing—first in a controlled environment and then on busy public roads.
For further information: ResearchGate