Reinforcement learning in linear embedding space unlocks generalizable control across soft robot configurations | Startup Networx | Startup Networx