Human-friendly Acceleration Control for Mobile Cart with

IEEJ Journal of Industry ApplicationsVol.9 No.6 pp.629–636 DOI: 10.1541/ieejjia.19005166

Paper

Human-friendly Acceleration Control for Mobile Cart with TorqueSensor Built in Driving Shaft

Ryohei Kitayoshi∗a)Student Member, Hiroshi Fujimoto∗∗ Senior Member

(Manuscript received May 9, 2019, revised Feb. 22, 2020)J-STAGE Advance published date : October 1, 2020

Recently, mobile carts have been used in the field of production and welfare. Accordingly, the usability of mobilecarts has become more important than ever before, and a human-friendly control technique for users is needed. In thispaper, we propose a novel acceleration control for the mobile cart and illustrate that the weight felt by users followsrapidly and precisely the command reference of weight set in the controller. The proposed method makes it possibleto control the wheel-side acceleration with a torque sensor built in the driving shaft. The torque sensor measures theoutput torque of the reduction gear. Moreover, two modes of application using the proposed method are shown. Oneis the power-assisted mode, and the other is the rehabilitation mode. To confirm the effectiveness of these modes,evaluation values of the usability are defined, and effectiveness is verified by simulations and experiments.

Keywords: mobile cart, human-friendly control, torque sensor, impedance control, acceleration control

1. Background

Recently, a lot of handcarts, mobile carts, and autonomouswheel robots have been used in the field of production andwelfare (1) (2). For example, YASKAWA Electric Corporationhas released the handcart put on a cooperative robot in Fig. 1.Accordingly, the usability of the devices for users has becomemore important than ever before. For improvement of usabil-ity, downsizing and lightening of actuators are needed, andsmall-sized flat reduction gears are often used. However, thereduction gears have some faults such as low rigidity, back-lash, and lost motion. It is known that these faults make itdifficult to control the driving shaft precisely (3) (4).

For solving this problem, two kinds of researches havebeen studied in recent years. One is research on control tech-nology by sensing or observing joint torque and realize highbackdrivability of actuators. Yamada et al. (5) realized pre-cise joint torque control with a load-side encoder for the two-inertia system with backlash. Kawai et al. (6) illustrated thatthe output torque of the reduction gear could be controlledaccurately by the torque sensor. Kuroki et al. (7) developed anovel torque sensor for precise measurement of joint torqueand control of human support robots. Research on estimationof output torque (8)–(10) has also been performed. The other isresearch on control of driving shaft of the mobile cart. Haraet al. (11) accomplished vibration suppression of the drivingshaft for mobile cart equipped with flexible objects. Akada etal. (12) demonstrated control of the mobile cart with a wheel-side encoder. However, there has been no study on control

a) Correspondence to: Ryohei Kitayoshi. E-mail: [email protected]∗ Corporate Research & Development Center YASKAWA Electric

Corporation12-1, Otemachi, Kokurakita-ku, Kitakyushu 803-8530, Japan

∗∗ The University of Tokyo5-1-5, Kashiwanoha, Kashiwa, Chiba 277-8561, Japan

Fig. 1. Example of handcart in the field of production

of acceleration of mobile cart by joint torque feedback mea-sured by the torque sensor.

Therefore, we produced the mobile cart equipped with thetorque sensor, which can measure the joint torque of the driv-ing shaft, and propose a novel acceleration controller for im-provement of usability of the mobile cart. Our proposed con-troller makes it possible to estimate the wheel-side velocityfrom the joint torque and to control wheel-side accelerationrapidly and precisely. Moreover, we propose two modes ofthe application using impedance control. One is the power-assisted mode, which means users can move the mobile cartwith less weight than actual weight. The other is the reha-bilitation mode, which means users can move the mobile cartwith more weight than actual weight. We confirm that bothof two modes are effective by experiment.

This paper consists of six sections, and the following isa summary of each section. In section 2, we explain anoverview of mechanism of the mobile cart and construct dy-namic model needed for controller design and simulation. Insection 3, we propose two acceleration control methods. Insection 4, we define three evaluation values for estimationof proposal control method. In section 5 and 6, we demon-strate validity of the proposal control method by simulation

c© 2020 The Institute of Electrical Engineers of Japan. 629

Human-friendly Acceleration Control for Mobile Cart（Ryohei Kitayoshi et al.）

Fig. 2. Overview of mobile cart

Fig. 3. Structure of actuator

and experiment.

2. Mobile Cart

In this section, we explain an overview of mechanism ofmobile cart and construct dynamic model needed for designof control system and simulation.2.1 Overview of Mechanism The mobile cart con-

sists of three parts: actuators, a control lever and load inFig. 2. The cart is driven by the two actuators and one passivewheel. The passive wheel has an encoder for measuring ve-locity and acceleration of the cart. The actuator in Fig. 3 hasdriving parts and measuring devices. The driving parts are aservo motor (YASKAWA Electric Corporation product Type:SGMMV-A2A2A) and a harmonic reduction gear (HarmonicDrive Systems product Type: CSF-14-30-2UH). The servomotor and reduction gear drive a wheel.

The measuring devices are a torque sensor and a load sideencoder. The torque sensor is built in driving wheel and capa-ble of measuring precisely torque generated by the reductiongear. The load side encoder measures rotation angle of thewheel. The control lever has a force sensor. The force sensormeasures operating force which is applied to the cart whenthe user pushes or pulls the cart.2.2 Strict Dynamic Model Dynamic model of the

cart consists of two-mass system and nonlinear system inFig. 4 and Table 1. This model is called as “strict dynamicmodel” hereafter. The two-mass system expresses dynamicbehavior of driving parts in the actuator. The motor andwheel are expressed as a rigid body and the reduction gearis expressed as elastic element because stiffness of the reduc-tion gear is lower than other driving parts.

The nonlinear system expresses that the cart is driven by

Fig. 4. Block diagram of strict dynamic model

Table 1. A list of symbols in strict dynamic model

Symbol Name UnitM Mass of Cart kgJM Motor Inertia kgm2

JL Wheel Inertia kgm2

r Radius of Wheel mωM Motor-side Angular Velocity rad/sωL Wheel-side Angular Velocity rad/sτM Motor Torque Nmτs Joint Torque NmV Car body Velocity m/sVL Wheel Velocity m/sFd Driving force NFo Operating force NKw Stiffness of Reduction Gear Nm/radμ Road Friction Coefficient -λ Slip Ratio -N Ratio of Reduction Gear -

Fig. 5. Magic formula

driving force which is nonlinear friction between the wheeland ground. The driving force Fd is determined by roadfriction coefficient μ and vertical force Fv shown in Eq. (1).Road friction coefficient is determined by magic formula (13)

which is nonlinear relation between road friction coefficientand slip ratio λ shown in Eqs. (2), (3). This nonlinear relationis shown in Fig. 5. Red zone of Fig. 5 is called cohesive zonewhere the nonlinear relation is considered to be linear.

Fd = μFv · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · (1)

630 IEEJ Journal IA, Vol.9, No.6, 2020


Fig. 6. Block diagram of approximate dynamic model

Table 2. Parameters of approximate dynamic model

Symbol Name Number UnitJM Motor Inertia 4.0 × 10−6 kgm2

Jall Total Inertia 1.8 × 10−1 kgm2

DM Motor Viscous Coefficient 4.3 × 10−5 Nms/radDL Wheel Viscous Coefficient 1.1 × 10−1 Nms/radKw Stiffness of Reduction Gear 1.9 × 103 Nm/radM Mass of Cart 40.3 ∼ 80.6 kgr Radius of Wheel 0.05 mN Reduction Ratio 31 -λ Slip Ratio 0.1 -

μ = D · sin(C · tan−1 φ

)· · · · · · · · · · · · · · · · · · · · · · · · · (2)

φ = B ·[(1 − E) · λ + E

B· tan−1(B · λ)

]· · · · · · · · · · · (3)

B = 26.66 C = 1.40 D = 0.18 E = 0.40

2.3 Approximate Dynamic Model The nonlinearsystem makes difficult to design controller and so we con-struct dynamic model with only linear two-mass systemshown in Fig. 6 by assuming that the cart moves in cohesivezone. This model is called as “approximate dynamic model”hereafter. Based on the assumption, mass of cart M is con-sidered to be part of wheel-side inertia and total of wheel-side inertia Jall is expressed in Eq. (4) on previous research (14).In next section, we design acceleration controller based onthe approximate dynamic model. Nominal slip ratio is set asλn = 0.10

Jall = JL + Mr2(1 − λn) · · · · · · · · · · · · · · · · · · · · · · · · · (4)

3. Control System

In this section, we propose the human-friendly accelerationcontrol system for the mobile cart. The control system shownin Fig. 7 is composed of three parts: impedance controller,acceleration controller, and wheel-side velocity estimator.

The impedance controller generates acceleration referenceaccording to operating force and realizes the two modes bychanging value of virtual mass. One is power-assisted modewhich means user can move the mobile cart with less weightthan actual weight. The other is rehabilitation mode whichmeans user can move the mobile cart with more weight thanactual weight.

The acceleration controller calculates torque reference tothe servo drive based on the acceleration reference, wheel-side angular velocity and motor-side angular velocity.

The wheel-side velocity estimator calculates wheel-sidevelocity from joint torque measured by torque sensor andmotor-side angular velocity measured by motor encoder.

Fig. 7. Block diagram of Control system

Fig. 8. Block diagram of Impedance control

Table 3. Virtual mass and actual massMode Virtual Mass[kg]: Mv Actual Mass [kg]: M Ratio : Mv/M

Power-assisted 40.3 80.6 0.5Rehabilitation 80.6 40.3 2.0

3.1 Impedance Control The Impedance control gen-erates wheel-side angular acceleration reference ALre f fromoperating force Fo in Eq. (5) and Fig. 8. Mv expresses a vir-tual mass. Values of the virtual mass is set depending on themode in Table 3.

ALre f =Fo

Mvr· · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · (5)

3.2 Acceleration Controller The acceleration con-trol is composed of two part: wheel-side velocity controllerand motor-side velocity controller. We propose two ways ofcontroller design corresponding to target performance. Thetarget performance is high accuracy and high command re-sponse of acceleration.

For high accuracy, both of controllers are I-P controllerto converge wheel-side velocity and motor-side velocity toreference without deviation. We call this controller “po-sition Integral-Proportional controller and velocity Integral-Proportional controller: IP-IP controller”.

For high command response, wheel-side velocity con-troller is PI controller and motor-side velocity is P controllerin order to converge to reference as fast as possible. We callthis controller “position PI and velocity P controller: PI-Pcontroller”.3.2.1 IP-IP Controller IP-IP controller is shown in

Fig. 9. A transfer function from acceleration reference ALref

to acceleration AL is illustrated in Eqs. (6) ∼ (10). Param-eters of the controller in Table 4 is designed by pole place-ment method and coefficient comparison. Eq. (11) expressesa reference model from ALref to AL. Poles: ωp are fourfoldroot and the value is determined to high values in the experi-ments, in which acceleration does not vibrate as much as pos-sible. And then making comparison between coefficients ofEq. (6) and Eq. (11) constructs four simultaneous equationsin Eq. (12).

AL

ALre f=

a0

s4 + a3s3 + a2s2 + a1s + a0· · · · · · · · · · · · (6)

a0 =KpMKpL

TiMTiL· · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · (7)



Fig. 9. Block diagram of IP-IP controller

Table 4. Parameters of IP-IP controllerSymbol Name Valueωp Pole of transfer function 55 × 2π

KpM Proportional gain of motor-side velocity controller 1.38 × 103

TiM Integral time of motor-side velocity controller 3.70 × 10−3

KpL Proportional gain of wheel-side velocity controller 6.52 × 102

TiL Integral time of wheel-side velocity controller 1.45 × 10−2

a1 =KpMKpL

TiM+

KpMDL

TiM Jall

+KpMKwJall

+

DLKwJmJallN2

+DmKwJmJall

· · · · · · · · · · · · (8)

a2 =KpM

Tim+ KpM

DL

Jall

+Kw

(1

Jall+

1JmN2

)+

DMDL

JM Jall· · · · · · · · · · · · · · (9)

a3 = KpM +DM

JM+

DL

JL· · · · · · · · · · · · · · · · · · · · · · · · (10)

AL

ALre f=

ω4p

(s + ωp)4· · · · · · · · · · · · · · · · · · · · · · · · · · · · · (11)

⎧⎪⎪⎪⎪⎪⎪⎪⎨⎪⎪⎪⎪⎪⎪⎪⎩

a0 = ω4p

a1 = 4ω3p

a2 = 6ω2p

a3 = 4ωp

· · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · (12)

3.2.2 PI-P Controller PI-P controller is shown inFig. 10. A transfer function from acceleration reference ALref

to acceleration AL is illustrated in Eqs. (13) ∼ (16). Parame-ters of the controller in Table 5 is designed by pole placementmethod and coefficient comparison. Eq. (17) expresses a ref-erence model from ALref to AL. Poles: ωp are triple rootand the value is determined by experiments. And then mak-ing comparison between coefficients of Eq. (13) and Eq. (17)constructs three simultaneous equations shown in Eq. (18).

AL

ALre f=

b0

s3 + b2s2 + b1s + b0· · · · · · · · · · · · · · · · · · (13)

b0 =DMKw

JM JallN2+

KpMKpL

TiL

+KwKpM

Jall+

DMKwJM Jall

· · · · · · · · · · · · · · · · · · · · · · (14)

b1 =KwKpM

Jall+ KpMKpL

+ Kw

(1

Jall+

1JMN2

)+

DMDL

JM Jall· · · · · · · · · · · (15)

b2 = Kpm − DM

JM− DL

Jall· · · · · · · · · · · · · · · · · · · · · · · · · (16)

Fig. 10. Block diagram of PI-P controller

Table 5. Parameters of PI-P controllerSymbol Name Valueωp Pole of transfer function 70 × 2π

KpM Proportional gain of motor-side velocity controller 1.31 × 103

TiM Integral time of motor-side velocity controller -KpL Proportional gain of wheel-side velocity controller 3.70 × 102

TiL Integral time of wheel-side velocity controller 3.04 × 10−5

Fig. 11. Block diagram of wheel-side velocity estimator

AL

ALre f=

ω3p

(s + ωp)3· · · · · · · · · · · · · · · · · · · · · · · · · · · · · (17)

⎧⎪⎪⎪⎪⎪⎨⎪⎪⎪⎪⎪⎩b0 = ω

3p

b1 = 3ω2p

b2 = 3ωp

· · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · (18)

3.3 Wheel-side Velocity Estimator Wheel-side ve-locity estimator is shown in Fig.11. Wheel-side velocity ωL

is calculated from motor-side velocity ωM and joint torque τs

by Eq. (19). Joint torque is measured by the torque sensor and4th order butterworth filter is used for elimination of noise ina high frequency of the torque sensor. Cut-off frequency ofthe butterworth filter is configured as 300 Hz.

ωL = ωM +τssKw· · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · (19)

4. Evaluation Values

For estimating proposal control method, we introduce threeevaluation values, “Sensation mass”: Ms, “Sensation error”:Me and “Response time”: Tr. These three values can be in-terpreted on time series data as shown in Fig. 12. “Sensa-tion mass”: Ms is weight which user feels during movingthe mobile cart and is calculated by Eq. (20). “Sensation er-ror”: Me is deviation between Ms and the virtual mass Mv

while sensation mass reaches virtual mass. This value is cal-culated by Eq. (21). “Response time”: Tr is a period aftermotor starts to move until sensation mass reaches the virtualmass. Time when sensation mass reaches the virtual mass isdefined as time when sensation mass enters between virtualmass Mv ± 5%.

Ms(t) =Fo(t)Ac(t)

· · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · (20)

Me =1

Tste

∫ Tste

0

|Ms(t) − Mv|Mv

× 100 dt · · · · · · · · · (21)



Fig. 12. Graphical interpretation of evaluation values

Fig. 13. Block diagram of motor-side velocity IP control

Fig. 14. Block diagram of motor-side velocity P control

5. Simulation

In this section, we show effect of proposal controllersbased on the evaluation values by simulation. In simulator,an user applies constant force to the cart and sensation errorand response time are evaluated during operation on power-assisted mode and rehabilitation mode.5.1 Condition Strict dynamic model is used as model

of plant. Reference of operating force is step-like signal fil-tered by low-pass filter of which cut-off frequency is 10 Hzand amplitude of the reference is 8 m/s2(converted acceler-ation value). The low pass filter is assumed to be frequencyresponse characteristic of human. For comparison with ourproposal method, we perform simulation in the case of ac-celeration control with only motor-side velocity controllershown in Figs. 13 and 14. Parameters of the controller is usedin Tables 4 and 5.5.2 Result As illustrated in Table 6, the deviation

of sensation mass of IP-IP controller is the least and the re-sponse time of PI-P controller is shortest in all of the method.Time response of sensation mass is shown in Figs. 15 ∼ 16.In the figures, red line is time response of sensation mass incase of proposal control, blue line is time response of sen-sation mass in case of only motor-side velocity control andgreen line is virtual mass. This result indicates that at leastone integrator is required for the precise acceleration control.5.2.1 IP-IP Controller Sensation mass follows accu-

rately to virtual mass shown in Fig. 15. Sensation error andresponse time are not quite different between IP-IP controland IP control. This is because motor-side IP controller sup-pressed effects of the motor-side viscous friction effectively

Table 6. Simulation resultType of controller IP-IP IP PI-P P

Power-assist mode Sensation error Me [%] 1.69 1.72 1.84 2.47Mv:40 [kg], M:80 [kg] Response time Tr [s] 0.110 0.124 0.019 0.438Rehabilitation mode Sensation error Me [%] 1.15 1.46 1.36 3.67

Mv:80 [kg], M:40 [kg] Response time Tr [s] 0.049 0.101 0.023 0.225

Table 7. Experimental result

Type of controller IP-IP IP PI-P PPower-assist mode Sensation error Me [%] 1.23 1.50 2.36 3.30

Mv:40 [kg], M:80 [kg] Response time Tr [s] 0.505 0.665 0.128 5.150Rehabilitation mode Sensation error Me [%] 0.66 2.74 3.13 6.51

Mv:80 [kg], M:40 [kg] Response time Tr [s] 0.231 0.303 0.180 1.573

and wheel-side viscous friction is not so large. As a result,it is considered that effect of wheel-side controller did notappear.5.2.2 PI-P Controller Sensation mass follows fast to

virtual mass in Fig. 16. However, steady-state deviation is leftfrom 0.6 s to 1.0 s. This is because that motor-side P controldid not suppress the effect of motor-side viscous friction ef-fectively. Moreover, in the case of only motor-side P control,it takes time for convergence of sensation mass and so boththe sensation error and the response time are large.5.3 Robustness against Wheel-side Inertia Fluctua-

tion We confirmed robustness of the proposed methodagainst the wheel-side inertia fluctuation because of the slipof the wheel. The wheel-side acceleration and acceleration ofthe car body were compared in Fig. 17. The operating con-dition shown in section 5.1 was used. In this simulation, slipratio λ in Fig. 18 is about 0.01, while the nominal slip ratioλn is 0.10. Despite the large difference between slip ratio λand nominal slip ratio λn, the difference between the wheel-side acceleration and the acceleration of the car body is small.Therefore, the proposed method is robust against the wheel-side inertia fluctuation because of the slip of the wheel.

6. Experiment

We demonstrate effectiveness of proposal controller byconverging sensation mass to virtual mass precisely and fastby experiment on the power-assisted mode and the rehabili-tation mode.6.1 Condition While controlling acceleration by our

proposal controller, the mobile cart is moved back and for-ward straight for 2 m by pushing or pulling control leverwith hand (shown in Fig. 19). For comparison with our pro-posal method, we made experiments in the case of acceler-ation control with only motor-side velocity controller shownin Figs. 13 and 14. The parameter of controller is used inTables 4 and 5.6.2 Result As illustrated in Table 7, the deviation

of sensation mass of IP-IP controller is the least and the re-sponse time of PI-P controller is shortest in all of the method.Time response of sensation mass is shown in Figs. 20 ∼ 21.In the figures, red line is time response of sensation mass incase of proposal control, blue line is time response of sen-sation mass in case of only motor-side velocity control andgreen line is virtual mass.6.2.1 IP-IP Controller Sensation mass follows ac-

curately to virtual mass shown in Fig. 20. However, unlikesimulation, response time of IP-IP control is shorter than



(a) Power-assisted mode (b) Rehabilitation mode

Fig. 15. Time response of Sensation Mass in the simulation of IP-IP control (red line: the proposed method, blueline: only motor-side velocity control method)


Fig. 16. Time response of Sensation Mass in the simulation of PI-P control (red line: the proposed method, blueline: only motor-side velocity control method)

(a) Power-assisted in IP-IP control (b) Rehabilitation in IP-IP control (c) Power-assisted in PI-P control (d) Rehabilitation in PI-P control

Fig. 17. Comparison for car acceleration and wheel-side acceleration

(a) Power-assisted in IP-IP control (b) Rehabilitation in IP-IP control (c) Power-assisted in PI-P control (d) Rehabilitation in PI-P control

Fig. 18. Slip ratio of the car body in the simulation

motor-side IP control. The cause of this difference is con-sidered that the weight of the mobile cart pushed the tire ofthe wheel to the ground and the friction with the ground in-creased.6.2.2 PI-P Controller Sensation mass follows fast

to virtual mass in Fig. 21. Like the simulation, in the caseof only motor-side P control, sensation error and response

time are large. However, its value is much larger than that ofsimulation. The cause of difference is also considered to befriction with the ground as in the case of IP-IP control.

7. Conclusion

In this paper, we propose the human-friendly accelerationcontrol for mobile cart with torque sensor built in driving



wheel. The acceleration controller realizes the high com-mand response and a steady state with less deviation enoughthat sensation mass converges to virtual mass in the simula-tion and experiment. In addition, it is shown that it can beapplied to two applications of the mobile cart: the power-assisted mode and the rehabilitation mode.

Fig. 19. Situation of experiment condition


Fig. 20. Comparison of time response of sensation mass in case of IP-IP control (red line: the proposed method,blue line: only motor-side velocity control method)


Fig. 21. Comparison of time response of sensation mass in case of PI-P control (red line: the proposed method,blue line: only motor-side velocity control method)

However, there are two future problems. One is deviationbetween virtual mass and sensation mass because of wheelslip. Wheel slip disturbs precise acceleration control andcause the deviation. We think that wheel slip can be sup-pressed by adding velocity limiter to the velocity controller.The other is an incorrect action because of offset error offorce sensor. The offset error generates unnecessary accel-eration reference and causes the incorrect action. We triedadding a high pass filter to the impedance controller in orderto eliminate the offset error. But the high pass filter dete-riorates the operability because it eliminates not only offseterror but also low frequency band component of operatingforce. As a result, we think that a physical switch is neededat the control lever in order to detect whether users contactwith the cart or not.

We will develop more practical and useful control methodwhile considering the problems.

References

( 1 ) R. Murai and F. Matsuno: “Field Experiment of Feasibility for Offering Ser-vice by an Mobile Robot in Hotel and Airport”, Journal of the Robotics So-ciety of Japan, Vol.36, No.4, pp.279–285 (2018)

( 2 ) S. Maki, K. Shirane, and R. Masaki: “A Map for Logistics Support MobileRobot and Its Application”, Journal of the Robotics Society of Japan, Vol.33,No.10, pp.732–737 (2015)

( 3 ) M. Nordin and P. Gutman: “Controlling mechanical systems with backlash—



a survey”, Automatica, Vol.38, pp.1633–1649 (2002)( 4 ) N. Amann, J. Bocker, and F. Prenner: “Active Damping of Drive Train Os-

cillations for an Electrically Driven Vehicle”, IEEE/ASME Transactions onMechatronics, Vol.9, No 4 (2004)

( 5 ) S. Yamada and H. Fujimoto “Precise Joint Torque Control Method for Two-inertia System with Backlash Using Load-side Encoder”, IEEJ Journal ofIndustry Applications, Vol.8, No.1, pp.74–82 (2019)

( 6 ) Y. Kawai, Y. Yokokura, and K. Ohishi: “High Back-drivable Pseudo I-PDTorque Control Using Load-side Torque Observer With Torsion Torque Sen-sor”, IEEE 14th International Workshop on Advanced Motion Control, Auck-land New Zealand, pp.1751–80 (2016)

( 7 ) Y. Kuroki, Y. Kosaka, T. Takahashi, E. Niwa, H. Kaminaga, and Y.Nakamura: “Cr–N Alloy Thin–film Based Torque Sensors and Joint TorqueServo Systems for Compliant Robot Control”, Proc. IEEE International Con-ference on Robotics and Automation (ICRA), pp.4954–4959 (2013)

( 8 ) J. Lee, C. Lee, N. Tsagarakis, and S. Oh: “Residual Based External TorqueEstimation in Series Elastic Actuators Over a Wide Stiffness Range: Fre-quency Domain Approach”, IEEE Robotics and Automation Letters, Vol.3,No.3, pp.1442–1449 (2018)

( 9 ) Y. Yamada and Y. Kakinuma: “Sensorless cutting force estimation for full-close controlled ball-screw-driven stage”, The International Journal of Ad-vanced Manufacturing Technology, Vol.87, No.9, pp.3337–3348 (2016)

(10) C. Mitsantisuk, M. Nandayapa, K. Ohishi, and S. Katsura: “Design for Sen-sorless Force Control of Flexible Robot by Using Resonance Ratio Con-trol Based on Coefficient Diagram Method”, Journal for Control, Measure-ment, Electronics, Computing and Communications, Automatic 54, pp.62–73(2013)

(11) S. Hara, T. Yoshiura, Y. Morita, N. Sato, and Y. Yamada: “Control Charac-teristics of Robust Power Assist Cart for Flexible Structures”, Transactionsof the Japan Society of Mechanical Engineers, Series C, Vol.79, No.799,pp.704–717 (2012)

(12) W. Akada and H. Fujimoto: “Study on Power Assist Control of Push CartRobot with Wheel-Side Encoder”, IEEJ International Workshop on Sensing,Actuation, Motion Control, and Optimization (2016)

(13) H.B. Pacejka and E. Bakker: “The Magic Formula Tyre Model”, Proc. 1st In-ternational Colloquium on Tyre Models for Vehicle Dynamics Analysis, heldin Delft, The Netherlands (1991)

(14) K. Fujii and H. Fujimoto: “Traction Control based on Slip Ratio EstimationWithout Detecting Vehicle Speed for Electric Vehicle”, Power ConversionConference, Nagoya (2007)

Ryohei Kitayoshi (Student Member) received the B.S. degree inUndergraduate School of Engineering Science atKyoto University, Japan in 2009 and received theM.S degree in department of systems science gradu-ate school of informatics at Kyoto University, Japan,in 2011. Currently, he is engaged in research anddevelopment of servo motors and industrial robotsat YASKAWA ELECTRIC CORPORATION. He isworking towards the Ph.D. degree at The Universityof Tokyo, Japan. His research interests include auto-

matic controller design, control of servo motors and system identification.He is a student member of the Institute of Electrical Engineers of Japan anda member of Society of Instrument and Control Engineerings

Hiroshi Fujimoto (Senior Member) received the Ph.D. degree in theDepartment of Electrical Engineering from the Uni-versity of Tokyo in 2001. In 2001, he joined the De-partment of Electrical Engineering, Nagaoka Univer-sity of Technology, Niigata, Japan, as a research asso-ciate. From 2002 to 2003, he was a visiting scholar inthe School of Mechanical Engineering, Purdue Uni-versity, U.S.A. In 2004, he joined the Department ofElectrical and Computer Engineering, Yokohama Na-tional University, Yokohama, Japan, as a lecturer and

he became an associate professor in 2005. He is currently an associate pro-fessor of the University of Tokyo since 2010. He received the Best PaperAwards from the IEEE Transactions on Industrial Electronics in 2001 and2013, Isao Takahashi Power Electronics Award in 2010, Best Author Prizeof SICE in 2010, the Nagamori Grand Award in 2016, and First Prize Pa-per Award IEEE Transactions on Power Electonics in 2016. His interestsare in control engineering, motion control, nano-scale servo systems, elec-tric vehicle control, motor drive, visual servoing, and wireless motors. Heis a senior member of IEE of Japan and IEEE. He is also a member of theSociety of Instrument and Control Engineers, the Robotics Society of Japan,and the Society of Automotive Engineers of Japan. He is an associate edi-tor of IEEE/ASME Transactions on Mechatronics from 2010 to 2014, IEEEIndustrial Electronics Magazine from 2006, IEE of Japan Transactions onIndustrial Application from 2013, and Transactions on SICE from 2013 to2016. He is a chairperson of JSAE vehicle electrification committee from2014 and a past chairperson of IEEE/IES Technical Committee on MotionControl from 2012 to 2013.


Documents

Human-friendly Acceleration Control for Mobile Cart with