4 ms·

The problems that are reported in the article seems simpler than I imagined. Is google really only gathering data via real world tests? If so then something lik

The problems that are reported in the article seems simpler than I imagined. Is google really only gathering data via real world tests? If so then something like NVIDIA simulator will completely crush waymo. https://blogs.nvidia.com/blog/2018/09/12/drive-constellation-open-simulation/ https://blogs.nvidia.com/blog/2018/09/12/drive-constellation...

YeGoblynQueenne 8y ago
You'd think that simulation is useful for training machine learning algorithms, but in practice, the differences between our best simulations and the real world are such that training on simulations is next to useless. Google AI blog explains why: Simulating many years of robotic interaction is quite feasible with modern parallel computing, physics simulation, and rendering technology. Moreover, the resulting data comes with automatically-generated annotations, which is particularly important for tasks where success is hard to infer automatically. The challenge with simulated training is that even the best available simulators do not perfectly capture reality. Models trained purely on synthetic data fail to generalize to the real world, as there is a discrepancy between simulated and real environments, in terms of both visual and physical properties. In fact, the more we increase the fidelity of our simulations, the more effort we have to expend in order to build them, both in terms of implementing complex physical phenomena and in terms of creating the content (e.g., objects, backgrounds) to populate these simulations. This difficulty is compounded by the fact that powerful optimization methods based on deep learning are exceptionally proficient at exploiting simulator flaws: the more powerful the machine learning algorithm, the more likely it is to discover how to "cheat" the simulator to succeed in ways that are infeasible in the real world. The question then becomes: how can a robot utilize simulation to enable it to perform useful tasks in the real world? The difficulty of transferring simulated experience into the real world is often called the "reality gap." The reality gap is a subtle but important discrepancy between reality and simulation that prevents simulated robotic experience from directly enabling effective real-world performance. Visual perception often constitutes the widest part of the reality gap: while simulated images continue to improve in fidelity, the peculiar and pathological regularities of synthetic pictures, and the wide, unpredictable diversity of real-world images, makes bridging the reality gap particularly difficult when the robot must use vision to perceive the world, as is the case for example in many manipulation tasks. https://ai.googleblog.com/2017/10/closing-simulation-to-reality-gap-for.html https://ai.googleblog.com/2017/10/closing-simulation-to-real...
- KaoruAoiShiho 8y ago
  The simulator isn't intended to fully replace real world training, the reality gap is a well known issue. However it's obvious that with *just real world training that you will never be able to get sufficient data and fail on easily predictable problems as described in the article. A lot of that post is spent on saying how hard it is to build a good simulation, but what is hard for some companies can be much easier for others. https://blogs.nvidia.com/blog/2018/12/03/physx-high-fidelity-open-source/ https://blogs.nvidia.com/blog/2018/12/03/physx-high-fidelity...

YeGoblynQueenne 8y ago

You'd think that simulation is useful for training machine learning algorithms, but in practice, the differences between our best simulations and the real world are such that training on simulations is next to useless. Google AI blog explains why: Simulating many years of robotic interaction is quite feasible with modern parallel computing, physics simulation, and rendering technology. Moreover, the resulting data comes with automatically-generated annotations, which is particularly important for tasks where success is hard to infer automatically. The challenge with simulated training is that even the best available simulators do not perfectly capture reality. Models trained purely on synthetic data fail to generalize to the real world, as there is a discrepancy between simulated and real environments, in terms of both visual and physical properties. In fact, the more we increase the fidelity of our simulations, the more effort we have to expend in order to build them, both in terms of implementing complex physical phenomena and in terms of creating the content (e.g., objects, backgrounds) to populate these simulations. This difficulty is compounded by the fact that powerful optimization methods based on deep learning are exceptionally proficient at exploiting simulator flaws: the more powerful the machine learning algorithm, the more likely it is to discover how to "cheat" the simulator to succeed in ways that are infeasible in the real world. The question then becomes: how can a robot utilize simulation to enable it to perform useful tasks in the real world? The difficulty of transferring simulated experience into the real world is often called the "reality gap." The reality gap is a subtle but important discrepancy between reality and simulation that prevents simulated robotic experience from directly enabling effective real-world performance. Visual perception often constitutes the widest part of the reality gap: while simulated images continue to improve in fidelity, the peculiar and pathological regularities of synthetic pictures, and the wide, unpredictable diversity of real-world images, makes bridging the reality gap particularly difficult when the robot must use vision to perceive the world, as is the case for example in many manipulation tasks. https://ai.googleblog.com/2017/10/closing-simulation-to-reality-gap-for.html https://ai.googleblog.com/2017/10/closing-simulation-to-real...

KaoruAoiShiho 8y ago
The simulator isn't intended to fully replace real world training, the reality gap is a well known issue. However it's obvious that with *just real world training that you will never be able to get sufficient data and fail on easily predictable problems as described in the article. A lot of that post is spent on saying how hard it is to build a good simulation, but what is hard for some companies can be much easier for others. https://blogs.nvidia.com/blog/2018/12/03/physx-high-fidelity-open-source/ https://blogs.nvidia.com/blog/2018/12/03/physx-high-fidelity...