ABR adaptive control demo in MuJoCo: CPU vs. Loihi

This video compares the adaptive control network running on the CPU versus running on Loihi. In all tasks, we show the amount of elapsed real time. The first two tasks, moving to a target "without weight" and "with weight", show normal control without adaptation. Both the CPU and Loihi implementations are able to simulate these examples in real-time, with one simulated second taking approximately one real second. In the third task, we enable the adaptive controller, which is more computationally intensive. It is able to run in real-time on Loihi, but the CPU is now significantly slower than real-time. Note that for Loihi, the majority of the time is spent communicating with the board. The actual computation on the board is relatively fast, so we expect to be able to scale up to even larger networks while maintaining real-time simulation.

[santiagoarroyob]

I really like your approach, what can i do to work with you