A 500Kg, self-contained biped robot, named Roboshift, has been conceived and tested to investigate issues associated with the control of industrial scale biped robots. This project represents the first credible attempt to build a heavy weight autonomous biped robot. The recent expansion in humanoid robot development has highlighted advances made in anthropomorphic biped technology. Current research into speech recognition, vision systems, laser topography, artificial intelligence and electroactive polymers will ultimately achieve an Android capable of human like actions and thought processes. Justification for this most demanding and expensive research is based on philanthropic models that suggest these robots will attend to the bedridden, or replace humans in dangerous areas. However, the cost of a biped robot when compared to that of a wheeled or tracked vehicle restricts commercialisation for these applications. As well, the size and working capacity of current humanoid robots is not compatible with the heavy lifting requirements found in such environments. It is proposed that only biped robots of an industrial scale, possessing a capacity much greater than that of a human, will be of commercial value in the future. Typical applications may include the handling of materials in confined or uneven terrain, where a forklift or other commercially available materials handling equipment would be unsuitable. For example, field handling in military, mining or geological environments. Minimal research has been conducted into the realisation of such a device, which presents challenges in terms of the magnitude of dynamic forces produced and of the systems required to control the robot in real-time. Review of relevant literature reveals that little research has been completed in this field. Therefore, operational characteristics for an industrial scale biped robot are defined. The design then details the structure and integration of mechanical, hydraulic, and electrical systems. Roboshift is powered by an internal combustion engine and is the first biped robot with a capacity for extended operation. Modelling was conducted to determine joint trajectories, power requirements, hydraulic flow parameters and dynamic characteristics. The robot is controlled by a distributed, hierarchical system comprising sixteen microprocessors, a control computer acting as the midbrain and a communications computer acting as the central nervous system. Sensors measure attitude and heading (vestibular system) as well as ground reaction forces and joint angles (propreoception). The control strategy is based on feed forward trajectories generated by inverse kinematic analysis. Corrections to trajectories are made in real time by higher level routines running on the main control computer. Joint position is achieved by local feedback control. Software for the robot was written in the C language. Experimental results are presented detailing the performance of the robot in comparison to theoretical analysis. After construction and testing of actuators and sensors, calibration software was tested successfully. Once calibrated, the robot was lowered to the ground where the active balance software was able to control the robot in the frontal and sagittal planes. Frontal sway software was tested with mixed success as natural oscillation of the structure, which was not detectable by the control system, led to erroneous force data. Detailed dynamic modelling was then completed to determine the causes of oscillation in the robot. The modelling led to the formulation of a control strategy where non-collocated sensors are used to measure link strain as a feedback to a modified proportional controller. The project has demonstrated that an industrial scale biped incorporating an internal combustion engine and hydraulic power system is feasible.. Analysis presented proposes that as the height of a biped robot increases, the expected elastic deformation of the structure increases as the cube of the height, making control extremely challenging. A strategy for the control of heavy-weight robots is suggested It is also proposed that technology incorporated in current humanoid robots can not be scaled to control industrial bipeds.
| Identifer | oai:union.ndltd.org:ADTP/279666 |
| Date | January 2005 |
| Creators | Cronin, Joe, Mechanical & Manufacturing Engineering, Faculty of Engineering, UNSW |
| Publisher | Awarded by:University of New South Wales. School of Mechanical and Manufacturing Engineering |
| Source Sets | Australiasian Digital Theses Program |
| Language | English |
| Detected Language | English |
| Rights | Copyright Joe Cronin, http://unsworks.unsw.edu.au/copyright |
Page generated in 0.002 seconds