Learning to Correspond Dynamical Systems
This work addresses the problem of modeling and simulating similar dynamical systems for researchers in robotics and control, but it appears incremental as it builds on existing latent space and encoder-decoder techniques.
The authors tackled the problem of learning correspondences between pairs of dynamical systems by developing a method that learns a shared latent state space and dynamics model from trajectory data, enabling simulation of one system to produce imagined motions of its counterpart and bisimulation synthesis. They demonstrated the approach on controlled bipedal walkers and a walker-pendulum pair, but no concrete performance numbers were provided.
Many dynamical systems exhibit similar structure, as often captured by hand-designed simplified models that can be used for analysis and control. We develop a method for learning to correspond pairs of dynamical systems via a learned latent dynamical system. Given trajectory data from two dynamical systems, we learn a shared latent state space and a shared latent dynamics model, along with an encoder-decoder pair for each of the original systems. With the learned correspondences in place, we can use a simulation of one system to produce an imagined motion of its counterpart. We can also simulate in the learned latent dynamics and synthesize the motions of both corresponding systems, as a form of bisimulation. We demonstrate the approach using pairs of controlled bipedal walkers, as well as by pairing a walker with a controlled pendulum.