Force/Torque Sensor-Free Online Payload Estimation for a Pneumatically Driven Robot

Robots with pneumatic drives have advantageous physical properties and are lightweight and cost-effective. During operation, a pneumatic robot handles objects of different payloads, which are often not known in advance. Determining the mass of this load is crucial for user assistance functions, trajectory generation and model-based control to increase both productivity and ease of use. In this work, a model-based approach for estimating the mass of a payload gripped by a pneumatically driven robot is presented. This approach does not require additional force or torque sensors, but uses the joint angle and pressure sensors already installed for control purposes. From these signals, the non-measurable effective driving torque is calculated model-based, taking into account the particularities of the pneumatic drives. As a basis for the estimation, the robot equations of motion are formulated such that they are linear in the dynamic parameters of the robot. An optimization problem is set up and solved with a recursive least squares method to estimate the payload online. This is extended by a logic to activate the estimation when a payload is picked. The algorithm is validated experimentally on a pneumatic robot performing a typical pick-and-place task.