Lifting Objects with Power-Assist: Weight-Perception-Based Force Control Concepts to Improve Maneuverability

We developed a 1-DOF power assist robot system to lift objects of different sizes by human subjects. We adopted a hypothesis that weight perception due to inertia might be different from that due to gravity when lifting an object with a power assist robot because the human feels a difference between the actual weight and the perceived weight of the object. We included this hypothesis in the robot dynamics. We then discussed the suitability of force control for the robot for lifting objects and developed several weight-perception-based force control strategies. These force control strategies may be compared to previously developed position control strategies, and the comparison results may help determine appropriate control for the robot to achieve desired maneuverability. The results, as a whole, may help develop human-friendly power assist devices to handle heavy objects in various industries.


Materials
We developed a 1-DOF power assist system for lifting objects as shown in Fig.1. A ball screw was actuated by a servomotor. A force sensor was tied to the ball nut and the object (rectangular thin aluminum box) was tied to the force sensor.
(a) (b) Fig.1 (a) 1-DOF power assist robot system for lifting objects. (b) Dynamics for lifting an object with the system.

Method 1:
As shown in Fig.1, the dynamic equation for the real system is as Eq.(1), where m is the actual mass of the object, x is displacement, x" is acceleration, f h is human force (load force), f a is actuator force, g is acceleration of gravity. However, the model system is as in Eq. (2). We adopted a hypothesis pertaining to weight perception that perception of weight due to inertia differs from perceived weight due to gravity when manipulating an object with a power assist robot. For this reason, we thought that the mass parameter for inertia force might be different from that for gravity force for the dynamics of manipulating an object with the power assist robot [3]. Hence, Eq.(2) becomes Eq.(3). We subtract Eq.(3) from Eq.(1) and get Eq.(4) and then get Eq.(5) from Eq.(4).

278
Mechatronics and Information Technology mx" =f h +f a -mg We see in Eq. (5) that there is no f h . Hence, no force sensor is required.But, acceleration sensor is required. It is possible to measure x from encoder & counter and to derive x". But it is noisy. Hence, acceleration sensor is necessary. We diagram the force control based on Eq.(5) as shown in Fig.2.

Method 2:
We multiply Eq.(1) by m 1 and Eq.(3) by m, and then subtract the latter from the former, and get Eq.(6). Hence, a force sensor is necessary to measure f h , but there is no need of acceleration sensor. We diagram the force control based on Eq.(6) as shown in Fig.3.

Experiment Methods
We may simulate the system shown in Fig.2 and Fig.3 separately using Matlab/Simulink (solver: ode4, Runge-Kutta; type: fixed-step; fundamental sample time: 0.001s) for different values of m, m 1 and m 2 and may subjectively evaluate maneuverability for the system. Then, we may compare the maneuverabilty obtained for force control to that previously obtained for position control in [3]. The comparison results may help determine appropriate control strategies for the robot to achieve desired maneuverability.

Conclusions and Future Works
We developed a 1-DOF power assist robot system for lifting objects. We included human's weight perception in the dynamics and derived force control schemes following two methods. The novelty in these force control schemes is that these schemes include human characteristics such as weight perception. We will simulate the system shown in Fig.2 and Fig.3 separately using Matlab/Simulink for different values of m, m 1 and m 2 and will subjectively evaluate maneuverability for lifting objects with the system. Then, we will compare the maneuverabilty obtained for force control to that previously obtained for position control in [3]. The comparison results may help determine appropriate control strategies for the robot system to achieve desired maneuverability. We will also compare maneuverability between horizontal and vertical manipulation of objects with the power assist robot system [6]- [9].