This keynote speech discusses self-reconfigurable modular robots. It introduces the concept of self-reconfigurable systems made of homogeneous modules that can change shape and function through self-reconfiguration. The speech describes the modular transformer (M-TRAN) robot prototype, which uses a hybrid lattice-chain design of stackable modules connected by magnetic and shape memory alloy actuators. Demonstrations of M-TRAN include wall climbing, creeping carpet motion, and rhythmic locomotion generated by a central pattern generator network optimized with a genetic algorithm.