BIPED ROBOT TRAJECTORY PLANNING, SIMULATION AND CONTROL FOR NON-FLAT TERRAINS

Abstract

This thesis developed novel gait generation approaches for two-legged or biped robots in non-flat terrain environments like a staircase or deformable soil. Biped robots have plenty of benefits over wheeled, quadruped, or hexapod robots due to their ability to behave like human beings in tough and non-flat environments. An energy-Efficient control strategy has been proposed for the staircase environment. In the case of downstairs motion, a novel brachistochrone hip trajectory has been proposed for a biped robot. A more accurate fast terminal discrete-time sliding mode trajectory tracking controller has been developed with fuzzy-based impedance modulation to climb the staircase. Finally, a reinforcement learning based controller has been developed for biped robot motion on deformable terrain. Chapter wise description of the thesis is elaborated below. Chapter 1 gives a brief introduction to biped robots, their applications, and the challenges associated with biped robots. It elaborates on the objectives and contributions of this thesis. A brief literature review for biped robots trajectory planning, kinematics, dynamics, optimization, and control approaches has been elaborated in Chapter 2. It also gives preliminaries with one test case to better adapt the basic concepts related to robotics. In Chapter 3, for the task of climbing stairs by a 9-link biped model, an adaptive cycloid trajectory for the swing phase is planned as a function of the staircase rise/run ratio. Zero Moment Point criteria for satisfying stability constraints. It incorporates i Abstract inverse kinematics using an unsupervised artificial neural network with a knot shifting procedure for jerk minimization. It is followed by a novel technique for joint angles trajectory tracking control with energy optimization in Chapter 4. Dynamics for toe-foot biped model using Lagrange formulation along with contact modeling using the spring-damper system with Neural Network Temporal Quantized Lagrange Dynamics, which couples inverse kinematics neural network with dynamics, has been developed. Ant Colony Optimization (ACO) to tune the Proportional-Derivative controller and torso angle in order to minimize joint trajectory errors and total energy consumed has been proposed. Three cases with variable staircase dimensions have been taken, and a comparison is made to validate the effectiveness of the proposed work. Generated patterns have been simulated in © Matlab and MuJoCo. In chapter 5, an adaptive trajectory planning algorithm with brachistochrone hip trajectory is developed so that biped robots of varying link lengths and masses can climb down on varying staircase dimensions. Brachistochrone is the fastest descent trajectory for a particle moving only under the influence of gravity. In most situations, while climbing downstairs, the human hip also follows a brachistochrone trajectory for a more responsive motion. Zero Moment Point based COG trajectory is considered, and its stability is ensured. The cycloidal trajectory is considered for the ankle of the swing leg by taking proper collision constraints. Parameters of both cycloid and brachistochrone depend on the dimensions of staircase steps. The proposed algorithms have been implemented using MATLAB®. Chapter 6 contributes to riddling out the joint space trajectory tracking problem for a fully actuated 9-Link bipedal robot while climbing the staircase using a faster and more accurate fast terminal discrete-time sliding mode control. The second objective is to provide a more realistic and practical approach for contact modeling, ensuring a much more stable staircase walk. For contact modeling, we have proposed a novel fuzzy-based impedance modulation approach considering a virtual spring-damper model with variable stiffness and damping coefficients. To obtain a discrete-time model, Euler’s method for discretization is considered after obtaining Lagrange’s dynamics. Higher control accuracy of order Δt3, where Δt is a constant time step, can be achieved with fast terminal discrete-time sliding mode control in comparison to conventional discreteii Abstract time sliding mode control with controller accuracy of order Δt2. A brief comparison has been shown between the conventional discrete-time controller and our proposed controller with the same initial conditions and in the same disturbance environment to corroborate the robustness of the controllers. The delayed Estimation method is used to estimate the unknown disturbances. All simulations have been performed in MATLAB and Simulink to manifest the efficacy of our proposed approach. Chapter 7 deals with Deformable terrain, which is another challenge for biped robots as it has to deal with sinkage and maintain stability without falling. In this chapter, a Deep Deterministic Policy Gradient-based controller is developed for a flat-foot bipedal robot walking in a deformable ground surface environment. We have considered a 7-link biped robot for our proposed approach. For soft soil terrain modeling, Mesh has been considered to describe its geometry, where mesh parameters determine the softness of soil. All simulations have been performed on PyChrono, which can handle soft soil environments. Finally, chapter 8 concludes the thesis with a discussion on the future scope related to biped robots.