Design and development of a humanoid with articulated torso

Learning Dual Arm Coordinated Reachability Tasks in a Humanoid Robot with Articulated Torso

Abstract: Performing dual arm coordinated (reachability)tasks in humanoid robots require complex planning strategies and this complexity increases further, in case of humanoids with articulated torso. These complex strategies may not be suitable for online motion planning. This paper proposes a faster way to accomplish dual arm coordinated tasks using methodology based on Reinforcement Learning. The contribution of this paper is twofold. Firstly, we propose DiGrad (Differential Gradients), a new RL framework for multi-task learning in manipulators. Secondly, we show how this framework can be adopted to learn dual arm coordination in a 27 degrees of freedom (DOF)humanoid robot with articulated spine. The proposed framework and methodology are evaluated in various environments and simulation results are presented. A comparative study of DiGrad with its parent algorithm in different settings is also presented.

Learning Coordinated Tasks using Reinforcement Learning in Humanoids.

Abstract: With the advent of artificial intelligence and machine learning, humanoid robots are made to learn a variety of skills which humans possess. One of fundamental skills which humans use in day-to-day activities is performing tasks with coordination between both the hands. In case of humanoids, learning such skills require optimal motion planning which includes avoiding collisions with the surroundings. In this paper, we propose a framework to learn coordinated tasks in cluttered environments based on DiGrad - A multi-task reinforcement learning algorithm for continuous action-spaces. Further, we propose an algorithm to smooth the joint space trajectories obtained by the proposed framework in order to reduce the noise instilled during training. The proposed framework was tested on a 27 degrees of freedom (DoF) humanoid with articulated torso for performing coordinated object-reaching task with both the hands in four different environments with varying levels of difficulty. It is observed that the humanoid is able to plan collision free trajectory in real-time. Simulation results also reveal the usefulness of the articulated torso for performing tasks which require coordination between both the arms.

Human balance and posture control during standing and walking

Abstract: The common denominator in the assessment of human balance and posture is the inverted pendulum model. If we focus on appropriate versions of the model we can use it to identify the gravitational and acceleration perturbations and pinpoint the motor mechanisms that can defend against any perturbation. We saw that in quiet standing an ankle strategy applies only in the AP direction and that a separate hip load/unload strategy by the hip abd/adductors is the totally dominant defence in the ML direction when standing with feet side by side. In other standing positions (tandem, or intermediate) the two mechanisms still work separately, but their roles reverse. In the tandem position ML balance is an ankle mechanism (invertors/evertors) while in the AP direction a hip load/unloading mechanism dominates. During initiation and termination of gait these two separate mechanisms control the trajectory of the COP to ensure the desired acceleration and deceleration of the COM. During initiation the initial acceleration of the COM forward towards the stance limb is achieved by a posterior and lateral movement of the COP towards the swing limb. After this release phase there is a sudden loading of the stance limb which shifts the COP to the stance limb. The COM is now accelerated forward and laterally towards the future position of the swinging foot. Also ML shifts of the COP were controlled by the hip abductors/adductors and all AP shifts were under the control of the ankle plantar/dorsiflexors. During termination the trajectory of both COM and COP reverse. As the final weight-bearing on the stance foot takes place the COM is passing forward along the medial border of that foot. Hyperactivity of that foot's plantarflexors takes the COP forward and when the final foot begins to bear weight the COP moves rapidly across and suddenly stops at a position ahead of the future position of the COM. Then the plantarflexors of both feet release and allow the COP to move posteriorly and approach the COM and meet it as quiet stance is achieved. The inverted pendulum model permitted us to understand the separate roles of the two mechanisms during these critical unbalancing and rebalancing periods. During walking the inverted pendulum model explained the dynamics of the balance of HAT in both the AP and ML directions. Here the model includes the couple due to the acceleration of the weight-bearing hip as well as gravitational perturbations. The exclusive control of AP balance and posture are the hip extensors and flexors, while in the ML direction the dominant control is with the hip abductors with very minor adductor involvement. At the ankle the inverted pendulum model sees the COM passing forward along the medial border to the weight-bearing foot. The model predicts that during single support the body is falling forward and being accelerated medially towards the future position of the swing foot. The model predicts an insignificant role of the ankle invertors/evertors in the ML control. Rather, the future position of the swing foot is the critical variable or more specifically the lateral displacement from the COM at the start of single support. The position is actually under the control of the hip abd/adductors during the previous early swing phase. The critical importance of the hip abductors/adductors in balance during all phases of standing and walking is now evident. This separate mechanism is important from a neural control perspective and clinically it focuses major attention on therapy and potential problems with some surgical procedures. On the other hand the minuscule role of the ankle invertors/evertors is important to note. Except for the tandem standing position these muscles have negligible involvement in balance control.

On the stability of anthropomorphic systems

Abstract: This contribution treats definitions, dynamic aspects, and stability concepts of anthropomorphic systems. In addition to general conclusions about the new method of two-legged systems modelling, there are given some characteristic schemes of perturbed steady-gait regime stabilization.

Optimization-based locomotion planning, estimation, and control design for the atlas humanoid robot

Abstract: This paper describes a collection of optimization algorithms for achieving dynamic planning, control, and state estimation for a bipedal robot designed to operate reliably in complex environments. To make challenging locomotion tasks tractable, we describe several novel applications of convex, mixed-integer, and sparse nonlinear optimization to problems ranging from footstep placement to whole-body planning and control. We also present a state estimator formulation that, when combined with our walking controller, permits highly precise execution of extended walking plans over non-flat terrain. We describe our complete system integration and experiments carried out on Atlas, a full-size hydraulic humanoid robot built by Boston Dynamics, Inc.

The iCub humanoid robot: an open platform for research in embodied cognition

Abstract: We report about the iCub, a humanoid robot for research in embodied cognition. At 104 cm tall, the iCub has the size of a three and half year old child. It will be able to crawl on all fours and sit up to manipulate objects. Its hands have been designed to support sophisticate manipulation skills. The iCub is distributed as Open Source following the GPL/FDL licenses. The entire design is available for download from the project homepage and repository (http://www.robotcub.org). In the following, we will concentrate on the description of the hardware and software systems. The scientific objectives of the project and its philosophical underpinning are described extensively elsewhere [1].

Forces acting on a biped robot. Center of pressure-zero moment point

Abstract: In the area of biped robot research, much progress has been made in the past few years. However, some difficulties remain to be dealt with, particularly about the implementation of fast and dynamic walking gaits, in other words anthropomorphic gaits, especially on uneven terrain. In this perspective, both concepts of center of pressure (CoP) and zero moment point (ZMP) are obviously useful. In this paper, the two concepts are strictly defined, the CoP with respect to ground-feet contact forces, the ZMP with respect to gravity plus inertia forces. Then, the coincidence of CoP and ZMP is proven, and related control aspects are examined. Finally, a virtual CoP-ZMP is defined, allowing us to extend the concept when walking on uneven terrain. This paper is a theoretical study. Experimental results are presented in a companion paper, analyzing the evolution of the ground contact forces obtained from a human walker wearing robot feet as shoes.