Nonlinear fuzzy observer for payload estimation in physical human-robot interaction

Robotics have developed at a very high rate in the last few years. Industries are becoming capable of automating repetitive procedures without pause, and the production speed and quality are improving everyday. The main drawback of robotics in industry is the inability to deal with slight changes in an environment or in certain object parameters. This problem is very much present at Koninklijke Luchtvaart Maatschappij (KLM) Cargo, therefore they want to automate parts of their product handling cycle. KLM Cargo handles 540 000 tons of cargo every year and their processing hub is located at Schiphol airport. A number of automation opportunities exist at this location, such as the use of Automatic Guided Vehicle (AGV), smart planning algorithms and palletising/depalletising robots. This last point is addressed in this thesis. The automation of the palletising/depalletising procedure is highly complex for a number of reasons: large variety of packages to be handled, restrictions for orientation of packages, hazardous materials, and more. The planning of a palletising order can handle the orientation problem and the placement of hazardous cargo, but the variety in size, weight and shape of packages makes fully automated palletising/depalletising difficult. The process of handling the packages is performed by manual labour and with the help of forklift trucks. An ideal solution to compensate the arduous labour is to use physical Human-Robot Interaction (pHRI). This concept allows a robot and human to handle cargo together to place it at a location. The current academic developments in human-robot collaborative object manipulation require known object dynamics and dimensions. This information is not known at KLM cargo, and therefore, the framework of human-robot collaborative object manipulation must be extended to allow for unknown object dynamics. One approach is currently being developed by using the Slotine and Li method of adaptive control. Unfortunately, this approach has a big drawback with regards to persistence of excitation. A new method utilizes fuzzy Takagi-Sugeno (TS) system with the sector nonlinearity approach where payload information is included in the systems state vector. This approach allows an observer to estimate the payload parameters without the persistence of excitation limitation. This thesis shows the motivation for using the fuzzy approach and develop it for a two-link manipulator and compare its performance to the classical approach. While the classical approach works under persistence of excitation, a one link manipulator is shown to work without this condition. Two fuzzy observers were tested, the fuzzy Luenberger observer and a sliding mode observer. Unfortunately, for the two link manipulator a feasible observer gain is not found. The approach was tested again, but this time only to estimate the mass of the payload. Unfortunately, this did not yield a feasible observer either. In order to find out the reason for this infeasibility and create an approach that will obtain a feasible solution requires more analysis of the observability of the fuzzy system.