Event
IDTD Trajectory Design Framework for System Identification
1 reports · 1 sources
Sangwoo Shin and co-authors propose the Information-Decoupled Trajectory Design framework (IDTD) for sim-to-real system identification. The method constructs exploration objectives using Schur complement scores derived from the Fisher information matrix, and after parameter normalization, selects the optimal trajectory segment per parameter and aggregates them logarithmically, in order to alleviate unreliable identification caused by parameter non-identifiability. The authors report that, in linear dynamical systems and simulation environments including the Go2 quadruped, G1 humanoid, and Crazyflie quadrotor, IDTD reduces parameter identification error by an average of 39.6% relative to the strongest prior active exploration baseline, and improves downstream policy transfer; when validated on a real K1 humanoid, the identified parameters accurately characterize the real system dynamics.
Report timeline
- IDTD: Information-Decoupled Trajectory Design for Sim-to-Real System Identification
The authors propose an information-decoupled trajectory design framework, IDTD, which constructs an exploration objective from a Schur complement score derived from the Fisher information matrix, then normalizes by parameter, selects the optimal trajectory segment for each parameter, and aggregates the scores logarithmically, to mitigate unreliable identification caused by indistinguishable parameters.
arXiv cs.RO ·