.jpg)
Accurate vehicle localization in environments where GNSS is unavailable remains one of the major technical challenges for intelligent mobility. Without the constant reference of satellites, inertial navigation systems accumulate a drift error that grows without bound, rapidly degrading the quality of the estimated position. Traditional solutions – such as expensive external sensors, dedicated infrastructure, or complex fusion of multiple sources – are often difficult to scale or integrate into commercial fleets. Faced with this scenario, an approach based on artificial intelligence and probabilistic models is redefining how vehicles can correct their speed without additional hardware.
The proposal consists of transforming the signals from the vehicle's internal sensors – such as the accelerometer and gyroscope – into a virtual speed correction, using a selective state model inspired by the Mamba architecture. This model efficiently captures the temporal dynamics of the vehicle's motion, while evidential learning based on a Normal-Inverse-Gamma distribution allows quantifying the uncertainty of each estimate. This correction, now accompanied by its confidence interval, is integrated as a virtual measurement into an error-state Kalman filter, reducing position drift to levels comparable to those of a dedicated external speed sensor. Results on real data demonstrate that inertial localization assisted by this correction achieves accuracy within 10% of that obtained with dedicated sensors, even during prolonged GNSS outages, and can run in real time at 40 Hz on embedded hardware.
This breakthrough opens the door to applications in autonomous vehicles, last-mile logistics, agricultural machinery, or mobile robots, where localization reliability is critical and the costs of additional sensors can be prohibitive. The key lies in an AI model for businesses that learns to interpret vehicle motion and provides an output with explicit uncertainty, allowing control systems to make informed decisions even under adverse conditions. For organizations looking to implement such solutions, having a technology partner that develops custom artificial intelligence is essential, as each fleet and each operating environment requires specific adaptations.
In this context, Q2BSTUDIO positions itself as a strategic ally for companies wishing to integrate intelligent localization technologies, signal processing, and embedded control. Through the development of custom applications and custom software, it is possible to build systems that capture sensor data, execute inference models in real time, and connect with cloud platforms for subsequent analysis. Our AWS and Azure cloud services enable the deployment of training and model update pipelines, while cybersecurity capabilities ensure the integrity of communications between the vehicle and the cloud. Additionally, business intelligence services with Power BI facilitate the visualization of localization performance metrics, helping operators make data-driven decisions.
The implementation of AI agents that continuously monitor the quality of speed correction and act on unexpected deviations is another area where expertise in AI for businesses makes a difference. The combination of evidential models with cloud and edge computing infrastructure allows not only correcting inertial drift but also progressively improving models with each new journey. In this way, organizations can scale their localization solutions without relying on expensive sensors, maintaining accuracy and operational safety.
In summary, uncertainty-aware speed correction represents an important step toward more autonomous, robust, and accessible vehicle localization. For companies wishing to adopt these capabilities, the path involves integrating custom software that leverages the latest advances in sequential models and evidential learning, along with an appropriate cloud and analytics platform. At Q2BSTUDIO, we offer precisely that cross-platform application and software development, combining artificial intelligence, cloud, and cybersecurity to transform intelligent mobility.
