On November 13, 2026, NTT Mobility announced that it will begin public-road trials in February 2027 of a self-driving bus that does not depend on pre-built high-definition (HD) maps. The bus will be equipped with an autonomous driving system developed by Israel's Imagry and tested in a trial field that includes public roads around the NTT Musashino R&D Center. The trials will start at Level 2, with a driver on board, to verify the system's suitability for Japan's road environment and its safety.
Introducing self-driving services to a new area or route requires not only building HD maps but also adjusting and verifying the system to match local road conditions and traffic rules. NTT Mobility aims to determine whether reducing reliance on HD maps can lighten that burden. Imagry's system recognizes its surroundings using camera images and is offered as an end-to-end (E2E) system that integrates everything from perception to driving planning and vehicle control.
Public-road trials begin in February 2027 with a driver on board
The trials will take place in the "Co-Creation Hub," which includes public roads around the NTT Musashino R&D Center. According to the press release, Imagry's autonomous driving system will be installed in a bus made by eVersum. The vehicle is scheduled for delivery on December 7, 2026, and system adjustments and on-site testing will proceed from that month.
Public-road driving tests will start in February 2027, and the plan is to begin driving evaluations and compile issues from April. The trials will check how the system handles Japan's left-hand traffic, traffic signals, and traffic regulations, while evaluating driving quality and safety. They will start at Level 2 with a driver on board, and NTT Mobility will work with transport operators to identify issues for commercial operation.
E2E approach combines AI with traffic rules
According to NTT Mobility, Imagry's system does not rely on LiDAR or pre-built HD maps. It recognizes its surroundings in real time from camera images and reconstructs the driving environment. Based on that information and route instructions, it decides its driving path and vehicle movements. While driving, it combines perception and path planning by trained AI with traffic rules such as stopping, right-of-way, and speed limits, which are set as explicit conditions. Ride comfort is also taken into account.
The company defines E2E as "a system that integrates everything from perception to driving planning and vehicle control." However, not all internal processing is consolidated into a single AI model. Functions for perception, decision-making, and routing are modularized, and speed and vehicle control are likewise separated by function. This structure allows individual functions to be improved separately while the system as a whole remains integrated. For example, the company explains that it could improve the detection of Japan-specific traffic signals on its own, then verify the related functions and the overall system's behavior.
Deployment and verification work remains even without HD maps
The goal of this trial is to reduce the burden of building HD maps in advance. Even if reliance on map building is reduced, however, confirming suitability for Japan's road environment and verifying safe commercial operation are still necessary. Based on the trial plan NTT Mobility announced on November 13, the work required for deployment and the evaluation content of this trial can be summarized as follows.
| Deployment task | Approach / what the trial addresses | Evaluation stage at announcement |
|---|---|---|
| Pre-building HD maps | Reducing reliance on high-definition maps and expanding to new routes | Verifying whether deployment burden can be reduced |
| Adapting to Japan's road environment | Handling left-hand traffic, traffic signals, and traffic regulations; responding to street parking, pedestrians, and cyclists | Driving quality and safety to be evaluated in public-road trials |
| Preparing for commercial operation | Integration with remote monitoring and operations management, compliance with Japanese laws, optimization of operating costs | Plan to identify issues with transport operators and address them |
Note: This classifies NTT Mobility's announced domestic trial plan by deployment task. No measured figures have been released on how much map-building cost or time can be reduced.
If pre-building HD maps can be skipped, the preparation needed to expand to new routes may be reduced. However, whether the bus can drive safely on local roads and work with existing operations management systems must still be confirmed separately. NTT Mobility itself cites adjusting to road environments and verifying safety as factors that make introduction to new areas and routes time-consuming and costly.
Night and bad-weather performance and commercial readiness are key issues
The public-road trials will evaluate recognition performance at night and in bad weather, in addition to responses to street parking, pedestrians, and cyclists. The company also plans to examine whether AI decisions can be explained and verified, and how dependence on training data affects performance. Toward commercial operation, alongside safety evaluation, it will work with transport operators to sort out and address issues related to integration with remote monitoring and operations management, compliance with Japanese laws, and optimization of operating costs.
At this point, no quantitative evaluation results on driving performance or recognition accuracy at night and in bad weather have been released, and the timing of commercial operation in Japan has not been disclosed. How much time and cost reduced reliance on HD maps can actually save when introducing new routes is also a question for future verification. In the public-road trials starting in February 2027, attention will focus on how far they can confirm not only safety in Japan's road environment but also more efficient deployment preparation and compatibility with existing operations management systems.