Isaac Gym offers a high performance learning platform to train policies for wide variety of robotics tasks directly on GPU. Both physics simulation and the neural network policy training reside on GPU and communicate by directly passing data from physics buffers to PyTorch tensors without ever going through any CPU bottlenecks. This leads to blazing fast training times for complex robotics tasks on a single GPU with 2-3 orders of magnitude improvements compared to conventional RL training that uses a CPU based simulator and GPU for neural networks.
P. Kapahnke, P. Liedtke, S. Nesbigall, S. Warwas, and M. Klusch. The Semantic Web -- ISWC 2010: 9th International Semantic Web Conference, Shanghai, China, Revised Selected Papers, Part II, volume 6497 of Lecture Notes in Computer Science, Springer, Berlin, (2010)
M. Gaham, B. Bouzouia, and N. Achour. Service Orientation in Holonic and Multi-Agent Manufacturing, volume 594 of Studies in Computational Intelligence, Springer, Cham, (2015)
N. Galaske, T. Wagner, D. Strang, R. Anderl, and R. Bruder. Gestaltung der Arbeitswelt der Zukunft: 60. Frühjahrskongress der Gesellschaft für Arbeitswissenschaft, München, 25, page 360-362. Dortmund, GfA-Press, (2014)
J. Lin, S. Sedigh, and A. Miller. Proceedings of the 8th IEEE International Symposium on Dependable, Autonomic and Secure Computing (DASC '09), Chengdu, China, page 690-695. (2009)
L. Liu, F. Felgner, and G. Frey. Proceedings of the 9th International Conference on Modeling, Optimization and Simulation (MOSIM'12), Bordeaux, France, (2012)
P. Fritzson. Proceedings of the 7th International Wireless Communications and Mobile Computing Conference (IWCMC 2011), Istanbul, Turkey, page 1648-1653. (2011)