External Measurement System for Robot Dynamics

Abstract

Robots require testing to verify modeling assumptions, confirm performance characteristics, and quantify their limits. One requirement for formal robot testing is an external measurement system (EMS), that measures the robot’s dynamics without relying on the robot’s modeling assumptions or using any component of the robot’s control system. Most robot dynamics measurement systems in literature are not external: they connect to the robot’s instruments, and/or rely on its modeling assumptions. In this work, I present a new, inertial sensor-based EMS, that consists of a set of inertial sensors, data acquisition hardware, and precisely specified calibration procedures. Static and dynamic calibration algorithms are tested for accelerometers, with static shown to be superior. Monte Carlo-based calibration algorithms are presented, to quantify calibration uncertainties, which can then be propagated to uncertainties in measurements. The EMS is tested on a Hopper robot, demonstrating its usefulness for dynamics measurement, and certain limitations for position measurement.