| Research Interests |
| Evolving Neural Controllers I am interested in the study and development of intelligent, adaptive robots which can operate in complex and unknown environments. The conventional control algorithms are programmed to precisely execute pre-defined series of actions. Therefore, the environment must be well known. Creating intelligent robots consists of letting them evolve, self-organize, and adapt to their environment, just like all life forms on earth have done and keep doing. |
| Humanoid Robots Recently, humanoid robots have become very attractive for many robotic researchers. In my doctoral thesis, I introduced the use of Intelligent Algorithms for real time gait generation of humanoid robots. The humanoid robot gait during walking and going up-stairs was analyzed based on minimum consumed and minimum torque change cost functions. For real time implementation we considered teaching a Radial Basis Function Neural Network based on Genetic Algorithm results. I am now interested to see the effect of different parameters on human walking motion. Also, stability is an important issue related to humanoid robots. Zero Moment Point has been proven to give good results, but other stability criteria need to be considered as well. . |
| Analyzing human arm motion Up to date, the planar human arm motion is analyzed. I am interested to study the human arm motion in 3D space. The system with 3 degrees of freedom and 7 degrees of freedom is considered. The redundancy makes the system similar with human arm. |
| Application of Neural Networks on Control of Robot Manipulators Conventional manipulator control methods sufferers from the following difficulties: - Manipulators dynamics must be exactly known. - Uncertainties existing in real manipulators seriously devaluate the control performance. - The computational load is very high. In order to overcome these difficulties, the application of Neural Networks gives good results. Because of their universal approximation feature, they could be used as general controllers suitable for any manipulator. With their learning ability, neural networks can improve their performance and finally achieve satisfactory results through offline and/or online learning without requiring explicit knowledge of manipulator dynamics. Their parallel computation structure can help solve high-computational-load problems. |
 |
| Cyber Rodent Robot |
 |
| "Bonten-Maru I" |
 |
| Prismatic Joint Biped Robot |