Design and Analysis of a Smart Power Management System for Ultracapacitor-Powered Robotic Platform

Daniel P. Muffoletto, Collin Mandris, Shola Olabisi,

Kevin M. Burke and Jennifer L. Zirnheld Energy Systems Institute

University at Buffalo, The State University of New York Buffalo, NY, United States

Harry L. Moore, Jr. and Hardev Singh US Army ARDEC

AMSRD-AAR-MEM Picatinny Arsenal, NJ, United States

Abstract- Capacitors are well known for their long lifetime

and ability to rapidly charge and discharge. However, these attributes come at the expense of a lower energy density than current rechargeable battery chemistries (NiMH, NiCad, Li-thium, etc). Recent advances in double-layer electrolytic capa-citors have increased their energy storage capabilities, allowing for the possibility to power unmanned systems that have tradi-tionally been powered by rechargeable batteries. Capacitive energy storage is particularly interesting for collaborative ro-botic networks, where small scouting robots are powered for short missions by ultracapacitors, then return to a mobile re-charging station to rapidly recharge and return to their mis-sion. A system implementing capacitive energy storage, unlike rechargeable battery storage, requires additional power condi-tioning circuitry to overcome the capacitor’s disadvantages, including a DC-DC converter to draw power as the capacitor’s voltage lowers and current limiting to overcome the nearly short-circuit initial charging conditions. A robotic platform implementing smart charge and discharge circuitry for one to several kJ of capacitive storage was built and the performance is analyzed and discussed.

Keywords – Capacitive Energy Storage; Collaborative

Robotics; DC-DC Power Conversion; Ultracapacitors

I. INTRODUCTION Collaborative robotic systems are inspired by the natu-

ral coordination of ants and bees, which as a collection can carry out a task that no single organism can accomplish. They are also fault tolerant in that they can still operate after the loss of a single member [1]. Collaborative robotic sys-tems seek to carry out tasks with the same coordination and high redundancy. Energy storage and recharging solutions are sought to minimally impede the system’s progress and to recharge a robot quickly so it may return to work with minimal delay.

Energy management solutions for mobile recharging stations that seek out worker robots [2] and for worker ro-bots that can transfer energy amongst themselves [3] have been presented, but with little focus on the energy storage mechanism used. In all cases, the robot should be charged quickly. Not only is a robot that is charging unproductive, but the robot and its mobile charging station become an ob-stacle to all other robots that are working in the area.

This work was supported by a fellowship from the New York State Space Grant Consortium.

The use of ultracapacitors as the only energy storage medium in a given system is beginning to be commercia-lized. An example is an electric screwdriver boasting a 90 second charge time introduced by Coleman [4]. Ultracapa-citors are also an attractive solution for collaborative robot-ics, as it would allow for rapid charging of the robot. The major drawback of capacitive storage is its low energy den-sity compared to batteries or fuel cells, which can be miti-gated in this scenario since a recharging robot can always be nearby. In this scenario their inherently low energy density can be compensated with frequent charging, and since the cycle lifetime of ultracapacitors is orders of magnitude greater than that of batteries (hundreds of cycles compared to nearly 500000 cycles), such a power system could outlast a similar one powered by rechargeable batteries alone [5].

A design for a small (24cm x 16cm x 10cm) robotic platform implementing capacitive energy storage is pre-sented in this paper, along with the circuitry required to effi-ciently and rapidly charge and to efficiently and maximally recover the energy stored in the capacitor. Simulated charg-ing and discharging characteristics are compared with cor-responding experimental results.

II. SYSTEM DESIGN Electric double layer capacitors, or ultracapacitors, are

among the highest energy dense capacitors currently availa-ble and were chosen for this design. The characteristics of the capacitors utilized are summarized in Table 1.

Since most single ultracapacitors are currently rated at or below 2.7V, and the voltage across the capacitor decreas-es in proportion to the square root of the energy remaining ( 1 2⁄ ), a DC-DC converter is needed to boost the voltage to a usable level as the energy is drained from the capacitor. The 5V operating voltage was selected so that a single voltage rail could supply the minimum requirements of a small robotic platform – a microcontroller, motors and telemetry.

TABLE 1 SPECIFICATIONS FOR MAXWELL TECHNOLOGIES BOOSTCAP BCAP0350 E250 ULTRA-

CAPACITOR [6] Rated Voltage 2.5V (2.7V Surge) Nominal Capacitance 350F +/- 20% Maximum Continuous Current 20A (1.5kA Peak) Leakage Current 1mA Mass 60g Volume 50mL

978 -1-4244-7129-4/10/$26.00 ©2010 IEEE 651 643

There are two competing options to confcapacitors and the supporting DC-DC convertconfiguration, several capacitors are placed inthe whole capacitor bank is operating above ttage, and a buck converter is used to regulatebank’s voltage down to 5V. The problemsproach are that the series arrangement of caprequire an equalizing circuit for each cell, andwould not run when the capacitor is under 5Vlizing circuit could be as simple as a passive rwhich would dissipate power continuously, more complicated active network that bypassenearing their surge voltage rating [6]. This apalso require a large capacitor bank, since mantors would be required to increase the voltasystem could always operate above 5V, eventors are drained. The 5V limit is due to the fconverter can only convert to a lower voltageovercome by using either a buck-boost converconverter, which comes at the cost of either loor greater complexity, respectively.

The topology selected for this applicatioor more capacitors in parallel with a boost coconvert the voltage on the capacitor, as shoThis topology eliminates the need for a balaand only requires a single DC-DC conversionever, since the boost converter can only opsome minimum voltage, some of the energy of the capacitor will be wasted, and so a bwith a low operating voltage was desired. Thiconductor MAX1709 Step-up was chosen fotion, which can operate down to 0.7V input aup to 2.4A at 5V, which is sufficient for thcircuitry and two small DC motors. For a 3operating from 2.5V down to 0.7V, roughly 8citor’s maximum energy remains on the capaboost converter shuts off; this corresponds to 80J out of 1kJ stored. Note that this energy isrobot recharges quickly (except for the enerself-discharge), as the energy will be left onwhen it starts to charge.

Also outlined in Fig. 1 is a buck convused when charging the robot to step down athe 2.5V and up to 20A needed to charge the cdocking at 40V instead of 2.5V, the losses

Fig. 1. Design of power management system on robo

figure multiple ter. In the first n series so that the system vol-e the capacitor s with this ap-pacitors would d that the robot

V. A cell equa-resistor ladder, or could be a

es cells that are pproach would

ny 2.5V capaci-age so that the n as the capaci-fact that a buck e. This can be rter or a SEPIC ower efficiency

on utilizes one onverter to up-own in Fig. 1. ancing network n stage. How-

perate down to storage ability

boost converter e Maxim Sem-or this applica-and can output

he digital logic 350F capacitor

8% of the capa-acitor when the

approximately s not lost if the rgy lost due to n the capacitor

verter which is a 40V input to capacitors. By due to contact

resistance and wiring resistance supply could be minimized. Thucarrying high current to the capacneeded. Consequently, contact dwhen the robot makes electricalstation is reduced.

III. SIMULA

Using the topology outlined efficiency of the boost convertersimulated the charging and discharobotic platform.

A. Capacitor Charging SimulatiFor the discharge simulation

robot would draw a constant 0.4Ater to power the motors and anysuming that the input power wasε 0.81 to the required output p5V and 0.3A) and using a simplif

a differential equation for the cacharge was found to be

In MATLAB, we simulated the the parameters previously mentioResults of the simulations are sho

B. Capacitor Recharging SimulaSeveral approaches for recha

simplest method utilizes a curre2.5V source, but the drawback ocurrent drops off quickly, as can charging method is to continuoutage or series resistance so that thconstant current at 20A.

Fig. 2. Simulated discharge of 2.5V, 3converter outputi

otic vehicle

from the charging power us, only a short connection citor on the robot itself was deterioration due to arcing l contact with its charging

ATIONS in section II, and the 81%

r given in its datasheet, we arging characteristics of the

on n, it was assumed that the A from the DC-DC conver-y supporting circuitry. As-s converted with efficiency power of Pout=1.5W (given fied model for a capacitor,

⁄ (1)

apacitor voltage during dis-

. (2)

capacitor’s discharge with oned using Euler’s method. own in Fig. 2.

ation arging were explored. The ent limiting resistor and a of this approach is that the be seen in Fig. 3. An ideal

usly vary the charging vol-he capacitor charges with a

50F capacitor through DC-DC ing 1.5W.

652 644

Fig. 3. Plot of voltage and current for a capacitor charginresistance

A simpler solution was realized by usinpower resistors in series with MOSFETs to swseries resistances as the capacitor’s voltage simulated result is shown in Fig. 4. These methods are summarized in Table 2. The solution was used in the final design. Althoucharge as fast as the constant current approaadded benefit of using two 50W resistors thabe replaced in the event of thermal failure incomplicated and expensive circuitry.

C. Scalability The setup can currently be scaled so t

storage capacity of the robot can be increasimilarly rated capacitors in parallel. Since are programmed to switch based on the measu

TABLE 2

CHARGING CURRENTS AND TIMES FOR SEVERAL APPCharging Method Average

Current Peak Current

RC Exponential 5.6 A 29.1 A Constant Current 20 A 20 A 2 FET Switches 12.2 A 29.1 A

Fig. 4. Charging capacitor using two MOSFETs to lowetance as capacitor voltage increases

ng through a fixed

ng several high witch on lower rises, and this three charging two-MOSFET

ugh it does not ach, it had the at could easily nstead of more

that the energy sed by adding the MOSFETs

ured capacitor

PROACHES Charging Time

1.8 min 0.5 min 0.8 min

er the series resis-

voltage, the charging curves wilthe time scale will be stretched otimes longer when using 2 capawere to be done, the charging current limiting power resistors,replaced in favor of components each capacitor could be charged a

IV. EXPERIMENT

Fig. 5 shows the robotic plaplete with a buck converter for chter for discharging the capacitor,for remote control and an omnidThe docking station and contactdiagramed in Fig. 6 were designeproach from any direction to mawith the contacts to facilitate aufuture.

The robot’s ATmega32 mgrammed to receive driving commote control and to drive thSN754410 Quadruple Half-H Briprogrammed to monitor the capainformation to switch the approdetermined by the simulations. Wdrive itself off the charger.

Fig. 5. Robotic test platform

Fig. 6. Design of robot's

ll look similar, except that over a longer period (e.g., 2 acitors in parallel). If this

circuitry (buck converter, and MOSFETs) could be rated for higher power, as

at a maximum of 20A.

AL RESULTS atform that was built, com-harging and a boost conver-, as well as an RF receiver directional docking station. ts on the robotic platform ed so that the robot can ap-ke an electrical connection

utonomous behavior in the

microcontroller was pro-mmands from a wireless re-he motors through a TI idge Driver IC. It was also citor’s voltage and use that

opriate series resistance as When fully charged, it will

at its docking station

charging station

653 645

The robot was charged at the docking station, and the capacitor’s voltage was logged using a data logging circuit (Logomatic V2 Serial SD Datalogger from Sparkfun Elec-tronics), and was then read into MATLAB for analysis af-terward. An example of this set of data is shown in Fig. 7, with the charging portion of the curve highlighted in Fig. 8. Note that the discharge curve depends on the current draw required by the robot, but when an appropriate output cur-rent is assumed in simulation the plots agree well. The ro-bot took longer to charge than in simulation, as evidenced by the voltage plots. This can be attributed to the buck con-verter entering a short-circuit protection mode, which shows up as the spikes on the plot. In this mode, the buck conver-ter is lowering its output voltage, waiting for the perceived short to be removed. Future work to switch resistances at better times would prevent these over-current conditions.

V. CONCLUSION The design for a robotic platform is presented in this

paper. By choosing to use ultracapacitors as the energy storage medium, the robot can achieve high drive time to charge time ratios. We observed drive times 5 to 6 times

longer than it takes to charge. When used in a collection of collaborative robots, this can be used to maximum benefit as a couple of robots can be designated to be a mobile recharg-ing platform for a collection of worker robots. Because the ultracapacitor powered robot presented here charges at a voltage and current most energy dense batteries packs could supply (40-60V at approximately 1A), this recharging robot could carry large banks of these batteries to charge other robots.

REFERENCES [1] Y. Cao, A. Fukunaga, and A. Kahng, "Cooperative

mobile robotics: Antecedents and directions," Autonomous Robots, vol. 4, pp. 7-27, 1997.

[2] A. Couture-Beil and R. T. Vaughan, "Adaptive mobile charging stations for multi-robot systems," presented at the Proceedings of the 2009 IEEE/RSJ international conference on Intelligent robots and systems, St. Louis, MO, USA, 2009.

[3] M. Kubo and C. Melhuish, "Robot trophallaxis: Managing energy autonomy in multiple robots," Proceedings of “Towards Autonomous Robotic Systems”(TAROS-2004), p. 77—84, 2004.

[4] J. Miller and A. Burke, "Electrochemical capacitors: challenges and opportunities for real-world applications," Electrochemical Society Interface, vol. 17, pp. 53-57, 2008.

[5] W. Ching-Kuo, H. Han-Pang, and S. Cheng-Han, "Dynamic analysis of the hybrid recharging system with super-capacitors on the armed cleaner robot," in Advanced Intelligent Mechatronics, 2009. AIM 2009. IEEE/ASME International Conference on, 2009, pp. 1533-1538.

[6] D. Y. Jung, Y. H. Kim, S. W. Kim, and S.-H. Lee, "Development of ultracapacitor modules for 42-V automotive electrical systems," Journal of Power Sources, vol. 114, pp. 366-373, 2003.

Fig. 7. Capacitor voltage during operation of robotic platform, recharged 3 times.

Fig. 8. Capacitor voltage measured while robot is charging

654 646