In the period of reference (April 2012-May 2012) the following sub-tasks have been completed, according to the TESBE workplan:

T1.4: Experimental assessment of the developed control approaches

T1.5: Validation on elementary handling tasks

T3.4: Validation of the new gripper

In sub-task T1.4, after having implemented multi-DOF force control, as defined in the previous reporting period, it was decided to test its performances for the case of the tracking of the operator foot movements, being more demanding for the large variability of the mechanical impedance of the human leg.

It was assessed firstly the performances of the Feed-Forward (FF) contribution and secondly the improvements achievable adding the Full State (FS) Feedback contribution.

The test showed a good capability of the FF contribution to produce movements of the foot substantially aligned with the interaction force exerted at the foot, demonstrating to have achieved a good decoupling between the different DOFs of the legs. 

The adding of the FS contribution has allowed to increase the tracking performances in terms of accuracy and bandwidth, but practical limit exists to the increase of the gains of the FS contribution, due to the non consistency between the variable sensed by the accelerometer (absolute acceleration of the foot) and the variable that can be controlled by the actuators (velocity of the foot relative to the backpack).

In sub-task 1.5, the validation of the controller has been performed evaluating its tracking performances on elementary motion tasks.

In particular three specific  tests have been carried on, the first evaluating the tracking performances in case of foot motions, the second in case of hand motions and the third in case of trunk motions.

All the tests showed a significant reduction of the resistance forces achieved by the new controller with respect to those achievable with the original controller of the BE, being in the range of 5÷7 times less.

Being the intensity of these forces well under the target value of 20N in every operating conditions,  the movements of the subjects operating the BE resulted faster and more natural than those achievable with the original controller.

In sub-task T3.4, after having integrated the gripper (hardware and software components), a number of validation tests have been performed to assess the effective capability of the device to grasp different object shapes and to make the operator aware of the stability of the force closure.

For the grasping test, cylindrical and conical objects having different dimensions have been used. The device showed an excellent capability to automatic adapt the posture of its fingers to the shape of the objects, exerting grasping forces on the finger contact surfaces and on the palm surface sufficiently high to guarantee stable grasps that could withstand external forces and moments acting on the objects in the range of those exerted by the Body Extender.

Other tests have been performed to assess the  suitability of the haptic feedback provided by the device, to make the operator aware of the stability conditions of the grasp. The test demonstrated that the combination of the displacement and  force felt by the operator at the trigger of the gripper is sufficient to drive him to the achievement of a stable grasp, having at the same time the possibility to modulate the grasping force.

In the period of reference (February 2012-March 2012) the following sub-tasks have been continued/completed, according to the TESBE workplan:

T1.4: Experimental assessment of the developed control approaches

T2.2: Implementation and test of the collaborative control

T2.4: Assessment of the combination of the haptic cues and the collaborative control

T3.3: Detailed development of the gripper

T3.4: Validation of the new gripper

In subtask 1.4, the extension to the multi-DOF case of the force control developed in the previous reporting period has been initiated.

In accordance with the structure of new force control developed for the 1 DoF case, the vector of the required motor torques (control law) have been expressed as the sum of a feed-forward contribution and a full state feedback contribution. These contributions have been defined taking into account the dynamics of the robots with flexible joints.

On the basis of the assumption that the full state feedback could produce enough damping action to reduce at minimum the difference between the joint and position accelerations, the feed forward contribution has been evaluated as in the case of rigid manipulators.

To complete the definition of the feed-forward contribution, an optimal load distribution between the two legs when both the feet are in contact with the ground has been defined, with the aim of reducing the internal forces and the amount of the torques transferred to the soil by each foot.

In subtasks T2.2 and T2.4 the validation of the Collaborative Control (CC) involving human subjects has been completed, after having achieved a satisfactory accuracy of the evaluation of the BE Center of Gravity (CG) performed by the CC.

With respect to the previous reporting period the inaccuracy of the evaluation has been significantly reduced (form 25mm to 8mm), through a better calibration procedure of the offset parameters, the development of a more realistic model of the leg compliance and the implementation of an evaluation method more robust to possible residual inaccuracies in the estimation of the joint angular positions.

Then, the validation of the CC involving human subjects has been carried on with two types of tests, the first providing the physical interaction with the BE from the external (i.e. without wearing the device) and the other providing the wearing of the device.

The tests demonstrated the good ability of the CC to keep the Zero Moment Point always inside the support polygon, for any random desired motion of the subject. Furthermore the intervention of the CC is felt by the subject as natural and smooth, making them highly confident of the capability of the device to keep the equilibrium of the system.

In subtasks T3.3 and T3.4, the revision 3 of the gripper has been finalized, taking into account all the feedbacks emerged in the technical meetings with SSSA.

The overall device dimensions and weight have been optimized according to structural analysis.  

The construction of the device in now in progress and it will be completed for the beginning of May.

In the period of reference (December 2011-January 2012) the following sub-tasks have been continued/initiated, according to the TESBE workplan:

T1.4: Experimental assessment of the selected control approaches

T2.2: Implementation and test of the collaborative control

T3.3: Detailed development of the gripper

In sub-task T1.4 the activity has been focused on the definition of suitable criteria for the optimal dimensioning of the parameters of the new force controller, developed in the previous reporting period.

This is needed, since on the one hand, the parameters of the new controller are dimensioned for a given value of the mass of the handled load that, on the other hand, can vary in a specified range.

Even if the straightforward solution would be that of modifying the parameters at run time (non linear control, using an on line estimation of the handled load), to keep low the complexity of the new controller, it was decided to verify its robustness to the variation of the value of the handled load.

The numerical analysis showed a good robustness of the new controller for whatever value of the handled load, belonging to its specified range of variability, is taken as reference for the dimensioning of the controller.  In particular, the analysis demonstrated that the highest system robustness is achieved when the parameters are dimensioned for the minimal possible value specified for the handled load.

Once closed the analysis of the new controller, the activity has been focused on its extension to the multi-DoF case, firstly defining the general architecture of the controller and then selecting the type and location of the additional sensors (accelerometers) required for its implementation.

In sub-task T2.3, the activity has been focused on the definition of suitable compensations for the three source of errors affecting the evaluation of the position of the BE center of gravity, identified in the previous reporting period.

In particular the parameters of the linear transformation between the measured positions of the motors and the corresponding positions of the BE joints have been experimentally calibrated, while the mechanical backlash in the leg abduction-adduction movements has been compensated with a simple contribution depending on the sign of the developed torque.

Furthermore, the compliance of the mechanical structure of the leg has been compensated introducing equivalent compliances concentrated at the leg joints, whose  best fitting values has been preliminary determined for three distinct circular trajectories of the trunk.

Finally, the updated collaborative control has been preliminary tested on the real device, showing a good capability to prevent the overturning of the system for lateral and longitudinal motions of the trunk.

In sub-task T3.3, an updated version of the gripper design (revision 3) has been defined, aiming at minimizing the overall dimensions and weight of the device. Furthermore all the issues concerning the electronics and the Human Machine Interface have been addressed and integrated into the gripper design.

In the period of reference (October 2011-November 2011) the following sub-tasks have been continued, according to the TESBE workplan:

T1.3: Numerical assessment of the selected control approaches

T2.2: Implementation and test of the collaborative control

T3.3: Detailed development of the gripper

In subtask 1.3, the activity has been focused on the design and implementation of a more sophisticated primary motion control and related stabilizing contributions, aiming at substantially increase the system performances achieved with the simple primary motion control analyzed in the previous reporting period, consisting of a high gain velocity control closed on the motor velocity.

An increase of the maximum achievable constant loop gain of the force control, of about 15 times with respect to that achieved with the previous motion control, has been achieved using the actuator as a force generator and controlling its force with suitable feed forward and full state feedback contributions.

An accelerometer has been used as addition sensor, due to the easiness of its integration into the existing BE mechanics and the quality of the state estimations that can be achieved.

The numerical evaluations and the experimental verifications have shown a good accordance with the theoretical predictions.

In subtask 2.2 the module of the collaborative control developed in the previous reporting period has been integrated into the existing control architecture of the Body Extender, in order to allow its experimental validation.

Before testing the system with a human operator, a specific test has been carried out, for evaluating the accuracy of the estimation of the projected position of the system center of gravity on the support plane.

The test has shown the existence of significant errors in the evaluation of the position of the system center of gravity, due to a series of causes that have been reduced or experimentally evaluated.

A compensated evaluation of the system center of gravity will be developed in the next reporting period, taking into account the experimentally quantified errors.

In subtask T3.3, the detailed design of the gripper has been updated in order to further reduce the weight and encumbrance of the device, host the control and drive electronics and increase the strength of the gripper body.

An evaluation prototype of the finger has been also constructed and tested, in order to verify the kinematics and the accuracy of the numerical estimations of the forces that can be exerted at the grasping surfaces.

In the next reporting period the detailed design of the device will be finalized and the construction of the mechanical parts will be initiated.

In the period of reference (August 2011-September 2011) the following sub-tasks have been continued/initiated, according to the TESBE workplan:

T1.3: Numerical assessment of the selected control approaches

T2.2: Implementation and test of the collaborative control

T3.3: Detailed development of the gripper

In sub-task 1.3, an in depth analysis, aiming at evaluating the effects of the variable parameters of the plant on the achievable performances of the system, has been carried on, using a simple 1 DOF model for the BE mechanics, spring-damper-mass models for impedances of the human limb and of the environment and a simplified controller, having a virtual damper as desired admittance, a high gain velocity control closed on the motor velocity as primary motion control and no stabilizing contributions.

The analysis evidenced that the specified high values of the human limb stiffness and of the environment mass, combined with the specified low values of environment stiffness and of the human limb mass, produce a shape of the Open Loop Transfer Function (OLTF) that is very critical for the stability of the system. For these combinations of the variable parameters of the plant, the value of the max achievable DC gain of the OLTF, able to guarantee a specified minimal phase margin of 30 degrees, is very low, bringing to poor system performances, in terms of bandwidth and resistance forces, that are not compatible with the applications envisaged for the BE.

The analysis also allowed to conclude that the elements preventing the possibility to increase the value of the OLTF DC gain are the high sensitivity of the OLFT to the human limb stiffness and the presence of significant amplification at the resonance of the mechanical transmission.

Different controllers will be tested in the next period, able to reduce those negative elementes.

In sub-task 2.2, the results achieved for the simple 1DOF case has been extended to general n-DOFs system. This has been possible by making simplifying assumptions that allowed to evaluate the displacement of the ZMP with respect to projection of the center of gravity on the support plane as a linear function of the BE joint angular accelerations, as well as to define an equivalent height of the center of gravity with respect to the support plane that is function of the direction of the BE joint angular accelerations vector and not by its module.

The new collaborative control has been implemented and tested in a simulated environment, using a simplified 12 DOFs model of the BE (arms excluded). The simulation has shown that the new collaborative control works properly also in presence of relatively high operator intended velocities, guaranteeing that the ZMP always falls inside the support polygon.

In sub-task 3.3, the detailed design of the gripper has been continued, producing two successive revisions of the device design.

The first revision have been completed at the end of August 2011, while a second revision have been initiated after an internal technical meeting between PERCRO and TELROBOT, in which several elements for improving the design of the device have been identified and discussed.

At present the design of the new gripper is in the optimization phase, aiming at reducing the weight and the encumbrance of the device to make them comparable with the existing gripper developed by PERCRO.