Mirroring reality
Next Limit - Stand: AV5010

Next Limit argues that its Anyverse is the answer to the twofold problem of training self-driving technology, a crucial step in ensuring the safety of AVs. Developing and training algorithms in the real world is time-consuming, in addition to being still insufficient. On the other hand, commonly used real-time simulators lack scene complexity and precision.

The brief of Anyverse is to provide the most high-fidelity data set generation possible using its unique spectral unbiased renderer. The technology makes it possible to generate image sequences that mirror reality and are physically correct, perfectly incorporating problematic environmental conditions, unpredictable human behavior, or peculiarities on the road. All data sets include automatically generated metadata with annotations that leave no margin for error.

By improving machine learning confidence levels, AV perception training can be sped up significantly. High-fidelity computer-generated data sets can help fill the gaps with challenging road scenarios featuring specific weather and lighting conditions that put driverless technology in potentially precarious - but actually quite common - situations.

Back to News