The reach of a human being is more expansive than one can even imagine, but at the same, it barely covers anything more important than our tendency of improving at a consistent pace. This tendency, in particular, has brought the world some huge milestones, with technology emerging as quite a major member of the stated group. The reason why we hold technology in such a high regard is, by and large, predicated upon its skill-set, which guided us towards a reality that nobody could have ever imagined otherwise. Nevertheless, if we look beyond the surface for one hot second, it will become abundantly clear how the whole runner was also very much inspired from the way we applied those skills across a real world environment. The latter component, in fact, did a lot to give the creation a spectrum-wide presence, and as a result, initiated a full-blown tech revolution. Of course, the next thing this revolution did was to scale up the human experience through some outright unique avenues, but even after achieving a feat so notable, technology will somehow continue to bring forth the right goods. The same has turned more and more evident in recent times, and assuming one new automotive-themed development ends up with the desired impact, it will only put that trend on a higher pedestal moving forward.

Helm.ai, provider of next-generation AI software for autonomous driving and automation of robotics, has officially announced an assortment concerning DNN (Deep Neural Network)-based foundation models to facilitate better behavioral prediction and decision-making. As they are a part of the company’s AI software stack for high-end ADAS L2/L3 and L4 autonomous driving, Helm.ai has trained these DNN foundation models to make predictions about the behavior of vehicles and pedestrians, especially in complex urban scenarios. Not just that, it will also predict the path an autonomous vehicle would take in those situations, something which hints at an intuitive decision-making mechanism. But how did Helm.ai made such a value proposition possible from an actionable standpoint? Well, the company effectively leveraged its in surround view full scene semantic segmentation technology, along with a 3D detection system to allow training for intent prediction and path planning capabilities. Furthermore, the models were built up using the company’s proprietary Deep Teaching technology. The latter source of learning is expected to generate a relatively broader predictive capacity, and if all goes as per the plan, it will also do that in a scalable way. Anyway, given this extensive training setup, the process ensures that Helm.ai’s technology is eventually able to learn directly from real driving data. Once the initial bit of input is in, the technology moves on to using the company’s highly accurate and temporally stable perception system for capturing information about complex behaviors of vehicles, pedestrians, and the surrounding driving environment. A detail worth mentioning here would be how the particular methodology presents a promising prospect for learning more subtle yet important aspects of driving. We briefly touched on the method through which Helm.ai trained its DNN foundation models, but what we still haven’t covered is the fact that both intent and path prediction systems of the company are trained through various observed images and video sequences, sequences that represent the most likely outcomes of what will happen next. As a result, the whole setup delivers at your disposal a predicted path which is always consistent with the intent prediction.

“At Helm.ai we are pioneering a highly scalable AI approach that addresses high end ADAS L2/L3 mass production and large scale L4 deployments simultaneously in the same framework,” said Vladislav Voroninski, CEO of Helm.ai.

Another thing we now must go back to and elaborate on is the prospects brought by Helm.ai’s Deep Teaching technology. You see, the particular methodology should prove immensely useful when it comes to circumventing cumbersome physics-based simulators and hand-coded rules, which are insufficient to capture the full complexity of driving in the real world.

The new DNN foundation models are, of course, a part of Helm.ai’s wider push for conceiving an AI-first approach to autonomous driving, an approach designed to seamlessly scale right from high-end ADAS L2/L3 mass production programs to large scale L4 deployments. Aiding the company’s cause is its hardware-agnostic and vision-first platform. We say so because the stated platform is well-equipped to address the critical perception problem for vision, while simultaneously integrating sensor fusion between vision and radar/lidar as per the situational requirements.

“Perception is the critical first component of any self-driving stack. The more comprehensive and temporally stable a perception system is, the easier it is to build the downstream prediction capabilities, which is especially critical for complex urban environments. Leveraging our industry-validated surround-view urban perception system and Deep Teaching training technology, we trained DNN foundation models for intent prediction and path planning to learn directly from real driving data, allowing them to understand a wide variety of urban driving scenarios and the subtleties of human behavior without the need for traditional physics based simulators or hand-coded rules,” said Voroninski.