Leonardo Marquez Pedro

Learning how to grasp based on neural network retraining

Humans have an incredible capacity to manipulate objects using dextrous hands. A large number of studies indicate that robot learning by demonstration is a promising strategy to improve robotic manipulation and grasping performance. Concerning this subject we can ask: How does a robot learn how to grasp? This work presents a method that allows a robot to learn new grasps. The method is based on neural network retraining. With this approach we aim to enable a robot to learn new grasps through a supervisor. The proposed method can be applied for 2D and 3D cases. Extensive object databases were generated to evaluate the method performance in both 2D and 3D cases. A total of 8100 abstract shapes were generated for 2D cases and 11700 abstract shapes for 3D cases. Simulation results with a computational supervisor show that a robotic system can learn new grasps and improve its performance through the proposed HRH (Hopfield-RBF-Hopfield) grasp learning approach.

A robust manipulation strategy based on impedance control parameters changes and smooth trajectories

This paper proposes a new manipulation strategy that combines smooth trajectories with an impedance controller which parameters are adapted in time of execution. The proposal uses a single control law during all manipulation phases, i.e. when the robot is in contact with the object, and during regrasping when the robot moves freely in space. The manipulator impedance changes during the different manipulation phases and this is achieved modifying mass, stiffness and damping parameters of the controller. Controller stability is further enhanced by the adoption of smooth trajectories that are generated using a compact and computationally light Squeezed Screws Planner. Additionally, the approach shows good response to the occurrence of unplanned contact or collisions. Experimental results with an industrial manipulator unscrewing a jar cap, provides interesting data about performance and show the benefits of the adopted manipulation strategy.

Adjustable interaction control using genetic algorithm for enhanced coupled dynamics in tool-part contact

Impedance control is commonly implemented for robotic contact applications. Its performance is a function of the dynamic coupling between the environment and the robots impedance controller: inertia, stiffness and damping. An interaction task may be considered successfully accomplished when the elected performance criteria, such as rise time, overshoot and accommodation are achieved, all in addition to guaranteeing stability. The selection and online changes of impedance parameters for such purpose is considered challenging for practical applications, given the non modeled dynamics of the robot and its intrinsic impedance, uncertainties in the dynamics of environment and tool. This study proposes the usage of a multi-objective genetic algorithm to obtain enhanced impedance controller gains, with the purpose of providing impedance control adaptability to a robot in a real physical interaction application. In this context, a one-degree-of-freedom mechanical contact model is proposed to obtain the values of the objective function. A genetic algorithm is used to obtain the best gains, as this is a nonlinear task. Simulation and experimental results are presented and discussed validating the system performance, which has shown a convergence of less than 6 generations. Tendencies to convergence of controller parameters were observed and discussed.

A Novel Device for Grasping Assessment during Functional Tasks: Preliminary Results.

This paper presents a methodology and first results obtained in a study with a novel device that allows the analysis of grasping quality. Such a device is able to acquire motion information of upper limbs allowing kinetic of manipulation analysis as well. A pilot experiment was carried out with six groups of typically developing children aged between 5 and 10 years, with seven to eight children in each one. The device, designed to emulate a glass, has an optical system composed by one digital camera and a special convex mirror that together allow image acquisition of grasping hand posture when it is grasped and manipulated. It also carries an Inertial Measurement Unit that captures motion data as acceleration, orientation, and angular velocities. The novel instrumented object is used in our approach to evaluate functional tasks performance in quantitative terms. During tests, each child was invited to grasp the cylindrical part of the device that was placed on the top of a table, simulating the task of drinking a glass of water. In the sequence, the child was oriented to transport the device back to the starting position and release it. The task was repeated three times for each child. A grasping hand posture evaluation is presented as an example to evaluate grasping quality. Additionally, motion patterns obtained with the trials performed with the different groups are presented and discussed. This device is attractive due to its portable characteristics, the small size, and its ability to evaluate grasping form. The results may be also useful to analyze the evolution of the rehabilitation process through reach-to-grasping movement and the grasping images analysis.

Environment Exploration for Impedance Control Based Manipulators: Initial Experimental Study on Object Position Detection with Stable Contact

Robotic manipulation under unstructured environments still is an open challenge to the research community. Lots of effort have been addressed to visual perception and remarkable progress has been achieved in the area. However, the information obtained from vision systems is frequently incomplete and noisy, incapacitating robotic systems to complete their target tasks in some cases. Moreover, little attention has been given to the use of exploration by touch as an approach to enrich environment perception, mainly due to difficulties that arrive from stable physical interaction with the environment. This work presents the experimental results from an initial study on the capability of an impedance controlled manipulator perceiving its surrounding environment using touch as source of information. The experiments culminate on a repeatability test which indicates that such a method provides higher precision information in comparison with 3D vision systems. The experimental procedures consider different impedance parameters during establishment and loss of contact. Considering the obtained results, different surface approach heuristic are evaluated to determine the most appropriated to, firstly, establish a stable contact with the environment and secondly, as a consequence, to perform spatial measures.