Motion Planning Strategies for Humanoid Robots Using Computational Intelligent Techniques with Computer Vision Assistance

Kumar, Priyadarshi Biplab (2019) Motion Planning Strategies for Humanoid Robots Using Computational Intelligent Techniques with Computer Vision Assistance. PhD thesis.

[img]PDF (Restricted upto 10/07/2021)
Restricted to Repository staff only



The current investigation is devoted towards design, development and implementation of novel motion planning strategies for smooth and hassle-free movement of humanoid robots in complex environmental scenarios. The aim of the study is motion planning of humanoid robots from a definite start point to a goal point avoiding intermediate obstacles using artificial intelligent techniques with the help of various types of sensors. Here, NAO humanoid robots are used as one of the humanoid platforms, and a new humanoid robot NORI has been designed and fabricated under laboratory conditions. The kinematic analysis of NAO and NORI humanoid robots has been performed using DH parameter and multibody formulation approach to get an insight into the motion constraint parameters. The gait trajectory analysis has been studied for NORI humanoid robot with observation of stable movements along different directions of motion. Here, numerous standalone algorithms such as Regression Analysis, Artificial Potential Field, Fuzzy Logic Control, Genetic Algorithm, Glowworm Optimisation and DAYKUN-BIP Virtual Target Displacement strategy have been formulated for motion planning of single as well as multiple humanoid robots. To navigate multiple humanoid robots in a common platform, a Petri-Net control architecture has been added to the motion planning models. To negotiate with complicated obstacle settings, computer vision assistance tool has also been integrated with the motion planning schemes. The developed motion planning models have been implemented on simulation platforms, and the results obtained from the simulation platforms are validated in real-time platforms created under laboratory conditions. The outcomes from both simulation and real-time platforms are compared against each other with observation of close agreement and minimal percentage of errors. To improve the performance of standalone algorithms, hybridisations are also attempted between the individual techniques by two-step and three-step hybrid models, and substantial performance enhancements have been observed. The developed motion planning schemes are also evaluated against other standard navigational controllers, and significant performance improvements have been noticed. Finally, the application areas of the current research are highlighted with indication of future directions of extension.

Item Type:Thesis (PhD)
Uncontrolled Keywords:NAO; NORI; DH Parameters; RA; APF; FLC; GA; GO; DVTD; Motion Planning; V-REP; Hybridisation
Subjects:Engineering and Technology > Mechanical Engineering > Automobile Engineering
Engineering and Technology > Mechanical Engineering > Machine Design
Engineering and Technology > Mechanical Engineering > Computational Fluid Dynamics
Divisions: Engineering and Technology > Department of Mechanical Engineering
ID Code:10022
Deposited By:IR Staff BPCL
Deposited On:27 Jun 2019 16:33
Last Modified:27 Jun 2019 16:33
Supervisor(s):Parhi , Dayal R.

Repository Staff Only: item control page