One of the most significant challenges in the development of an autonomous human-scale robot is the issue of power supply. Perhaps the most likely power supply/actuator candidate system for a position or force actuated human-scale robot is an electrochemical battery and dc motor combination. This type of system, however, would have to carry an inordinate amount of battery weight in order to perform a significant amount of work for a significant period of time.
A state-of-the-art example of a human scale robot that utilizes electrochemical batteries combined with dc motor/harmonic drive actuators is the Honda Motor Corporation humanoid robot model P3. The P3 robot has a total mass of 130 kg (285lb), 30 kg (66 lbs) of which are nickel-zinc batteries. These30 kg of batteries provides sufficient power for approximately 15–25 min of operation, depending on its workload. Operation times of this magnitude are common in self-powered position or force controlled human-scale robots, and represent a major technological roadblock for designing actuated mobile robots that can operate power-autonomously for extended periods of time.
1.1. Figure of Merit
Assuming that a given power supply and actuation system can deliver the requisite average and peak output power at a bandwidth required by a power-autonomous robot, three parameters are of primary interest in providing optimal energetic performance. These are the mass-specific energy density of the power source (Es), the efficiency of converting energy from the power source to controlled mechanical work(n) , and the maximum mass-specific power density of the energy conversion and/or actuation system(Ps) . A simple performance index is proposed by forming the product of these parameters
A.P = Es*n*Ps ------------- (1)
where A.P is called the actuation potential. Such a figure of merit is justified by the fact that a system with high power-source energy density, high conversion efficiency, and high actuator power density will be the lightest possible system capable of delivering a given amount of power and energy.
In the case of a battery-powered dc-motor-actuated robot, the energy density of the power source would be the electrical energy density of the battery, the conversion efficiency would be the combined efficiency of the (closed-loop controlled) dc motor and gear head, and the power density of the energy conversion and actuation system would be the rated output power of the motor/gear head divided by its mass. In the case of a gasoline-engine-powered hydraulically actuated system, the energy density of the power source would be the thermodynamic energy density of gasoline; the conversion efficiency would be the combined efficiency of the internal combustion engine(converting thermodynamic energy to shaft energy), hydraulic pump (converting shaft energy to hydraulic energy), and the hydraulic actuation system (converting hydraulic energy to controlled mechanical work); and finally, the power density of the energy conversion and actuation system would be the maximum output power of the hydraulic actuation system, divided by the combined mass of the engine, pump, accumulator, valves, cylinders, reservoir, and hydraulic fluid of the hydraulic system.
With regard to this figure of merit, batteries and dc motors capable of providing the requisite power for a human scale robot offer reasonable conversion efficiency, but provide relatively low power-source energy density and a similarly low actuator/gear head power density. A gasoline-engine-powered hydraulically-actuated human-scale robot would provide a high power-source energy density, but a relatively low conversion efficiency and actuation system power density.
1.2. A Monopropellant Powered Approach
Liquid chemical fuels can provide energy densities significantly greater than power-comparable electrochemical batteries. The energy from these fuels, however, is released as heat, and the systems required to convert heat into controlled, actuated work are typically complex, heavy, and inefficient. One means of converting chemical energy into controlled, actuated work with a simple conversion process is to utilize a liquid monopropellant to generate a gas, which in turn can be utilized to power a pneumatic actuation system. Specifically, monopropellants are a class of fuels (technically propellants since oxidation does not occur) that rapidly decompose (or chemically react) in the presence of a catalytic material. Unlike combustion reactions, no ignition is required, and therefore the release of power can be controlled continuously and proportionally simply by controlling the flow rate of the liquid propellant. This results in a simple, low-weight energy converter system, which provides a good solution to the design tradeoffs between fuel energy density and system weight for the scale of interest.
Monopropellants, originally developed in Germany during World War II, have since been utilized in several applications involving power and propulsion, most notably to power gas turbine and rocket engines for underwater and aerospace vehicles. Modern day applications include torpedo propulsion, reaction control thrusters on a multitude of space vehicles, and auxiliary power turbo pumps for aerospace vehicles. This seminar describes the design of a monopropellant-powered actuation system appropriate for human-scale self-powered robots, and presents theoretical and experimental results that indicate the strong potential of this system for high energy density human-scale robot applications. Specifically, with regard to the figure of merit described before .The proposed approach is projected to provide a significantly greater power-source energy density and actuation power density relative to batteries and dc motors, and is projected to provide a higher conversion efficiency and significantly greater actuation system power density relative to a gasoline-powered hydraulic system.
2. DESCRIPTION OF MONOPROPELLANT ACTUATION SYSTEM
The monopropellant-powered actuation system is similar in several respects to a typical pneumatically actuated system, but rather than utilize a compressor to maintain a high-pressure reservoir, the proposed system utilizes the decomposition of hydrogen peroxide (H2O2) to pressurize a reservoir. Peroxide decomposes upon contact with a catalyst. This decomposition is a strongly exothermic reaction that produces water and oxygen in addition to heat. The heat, in turn, vaporizes the water and expands the resulting gaseous mixture of steam and oxygen. Since the liquid peroxide is stored at a high pressure, the resulting gaseous products are similarly at high pressure, and mechanical work can be extracted from the high-pressure gas in a standard pneumatic actuation fashion.
The conversion of stored chemical energy to controlled mechanical work takes place as follows. The liquid H2O2 is stored in a tank pressurized with inert gas (called a blow down tank) and metered through a catalyst pack by a solenoid-actuated control valve. Upon contact with the catalyst, the peroxide expands into oxygen gas and steam. The flow of peroxide is controlled to maintain a constant pressure in the reservoir, from which the gaseous products are then metered through a voice-coil-actuated four-way proportional spool valve to the actuator. Once the gas has exerted work on its environment, the lower energy hot gas mixture is exhausted to atmosphere.
3. MONOPROPELLANT ACTUATOR PROTOTYPE
A prototype of the monopropellant-powered actuation system depicted in Fig. 1 was fabricated and integrated into a single degree-of-freedom manipulator, as shown in Fig. 2.
The primary objective of building the prototype was to demonstrate tracking control and to conduct experiments characterizing the actuation potential described by (1). The propellant is stored in a stainless steel blow down propellant tank, and is metered through a two-way solenoid-actuated fuel valve through a catalyst pack and into a stainless-steel reservoir. The catalyst pack consists of a 5-cm-long (2 in), 1.25-cm-diameter (0.5 in) stainless-steel tube packed with catalyst material. A pressure sensor measures the reservoir pressure for purposes of pressure regulation. The high-pressure hot gas is metered into and out of a 2.7 cm (1-1/16 in) inner diameter, 10 cm (3.9 in) stroke double-acting single-rod cylinder by a four-way spool valve, modified for proportional operation by replacing the solenoid actuator with a thermally isolated voice coil. The valve spool displacement is measured with a differential variable reluctance transducer (DVRT) in order to enable closed-loop control of the valve spool position. The pneumatic cylinder is kinematically arranged to produce a bicep-curling motion upon extension of the piston, as illustrated in Fig. 3.
Control of the system is achieved using three separate control loops. The first and simplest is the pressure regulation of the reservoir. Pressure feedback from the pressure sensor switches the solenoid fuel valve with a thermostat-type on-off controller that regulates the reservoir pressure to 1515 kPa (220 psig). The second control loop provides a high-bandwidth (i.e., approximately10 Hz) position control of the valve spool. Finally, the valve spool position is commanded by an outer control loop, which controls the angular motion of the single-degree-of-freedom manipulator. The outer control loop utilizes a rotary potentiometer to provide arm angle measurement for a position, velocity, acceleration (PVA) feedback controller, which commands the valve spool position.
4.1. Load Profile
Since the actuator relies on gas as an energetic medium, and since the actuation system is not designed to utilize energy resulting from condensation of the steam (steam quality less than 100%), the energy required to vaporize the water will not be recovered and as a result the conversion efficiency is lower than if actuation system included partial condensation. The best possible efficiency would occur when partial condensation is allowed to occur within the actuator and also when the load profile of the piston is designed to allow isentropic expansion from high pressure down to the lowest pressure possible (atmospheric pressure). In particular the most efficient load profile is such that the expansion of the peroxide reaction products is isobaric until all propellant mass is in the actuator, at which point the expansion becomes isentropic and continues as such until the cylinder pressure reaches atmospheric. Partial condensation occurs as a result of this load profile, leaving 70% quality steam in the actuator. This load profile would yield a theoretical efficiency of 39 %( calculated theoretically) for the 70% peroxide solution at a supply pressure of 220 psig.
4.2. Uninsulated Experiments
Experiments were conducted to measure the previously calculated conversion efficiency. A 70% peroxide solution was used as the propellant to maintain acceptable temperatures for commercially available components. For these experiments, the single-degree-of-freedom manipulator was commanded to move the 11 kg mass through a 30-degree amplitude, 1-Hz sinusoidal motion. The work output was computed indirectly by measuring the angle and, in post-processing, computing the actuation torque using a model of the load. The instantaneous power and average power could then be calculated. The propellant mass consumption was measured indirectly by recording the pressure of the nitrogen gas in the blow down tank, assuming an isothermal process inside the constant-volume tank, and calculating the volume occupied by the nitrogen from the ideal gas equation, which in turn yields the volume of propellant in the tank. Since the propellant is a liquid, the mass of propellant used is easily computed from the known volume and density. The conversion efficiency is then computed over an integer number of cycles with the heat of decomposition of 70% hydrogen peroxide solution.
Based on these measurements, the experimentally determined conversion efficiency was found to be 6.6%. Note that the electrical power required to operate the valves was neglected in this analysis. The measured average combined electrical power required by the fuel and gas valves was approximately 2 W. Since this is only about 3% of the average work delivered by the actuator, this electrical power can be legitimately omitted from the analysis. The significant discrepancy between the measured conversion efficiency of 6.6% and the calculated upper bound of 16% is due to two major factors. The first is inefficiency in control and the second is heat loss. Specifically, the thermodynamic model assumed that no gas was exhausted during a given monotonic segment of the trajectory, and that no energy was lost as heat. Regarding the former, any overshoot of the desired trajectory will violate the assumed monotonicity of the trajectory, and therefore will result in an intermittent exhaust of hot gas and a resulting decrease in the efficiency. The existence of such intermittent exhaust is evident in the oscillations exhibited in the power delivered to the load which is shown in Fig. 4 plotted against the theoretically required power .Regarding inefficiency due to heat loss, the external surfaces of the catalyst pack, reservoir, and actuator were hot during the experiments, thus indicating the presence of heat flow. In order to more quantitatively assess the degree of heat loss, the prototype was instrumented with thermocouples so that the rate of heat loss could be estimated from surface temperature measurements referenced to tables associated with heat loss from uninsulated steam piping . This measurement yielded an estimated heat loss rate of 140W. Note that the average measured mechanical power output was approximately 60 W. The prototype lost twice as much energy in the form of heat as it delivered in the form of work. Taking into account this heat loss, the conversion efficiency of the prototype was recalculated to be 10 %
4.3. Insulated Experiments
In order to improve the measured conversion efficiency, the catalyst pack, reservoir, and actuator were wrapped in insulating tape, as shown in Fig. 5, and measurement of the conversion efficiency was repeated. For the insulated case, the experimentally determined conversion efficiency was found to be 9 %.Thermocouple measurement of the surface temperatures, as previously described, yielded an estimated heat loss rate of 73 W, approximately half of the uninsulated case. Using this heat loss rate, the theoretically calculated efficiency was 12 %, the difference presumably due to control inefficiency (i.e., intermittent exhausts).
4.4. Experimentally Determined Actuation Potential
Having measured the conversion efficiency, the mass-specific power density of the actuator and the mass-specific energy density of the power source need to be determined in order to calculate the actuation potential (1). The former is found by determining the mass and the maximum output power of the energy conversion and actuation system. Though finding the mass is a trivial task, characterizing the maximum deliverable power is not as straightforward due to the dependence upon several factors, including the supply pressure, the valve flow coefficient of the proportional valve, and the nature of the load, among others. In order to base the actuator power density solely on measured data the maximum deliverable power was estimated by using the peak power consistently measured during the previously described efficiency experiments. As evidenced by the data in Fig. 4, the actuator can consistently generate peak power of 150W, as indicated by the dashed line overlaid on the plot. The mass of the actuation system was obtained by weighing the components of the actuator shown in Fig. 2. The mass of each component is summarized in Table 1
As indicated in the table, the total actuation system mass is 1.5 Kg, thus resulting in an actuation system power density of 100 W/Kg. This would increase for a multi-degree-of-freedom system, since such a system would only include a single fuel valve, catalyst pack, pressure reservoir, and pressure sensor. Having determined the actuator power density, only the power-source energy density need be found in order to calculate the actuator potential. As previously mentioned, the heat of decomposition of 70% hydrogen peroxide propellant is 2.0 MJ/Kg. The propellant must be stored, however, in a pressurized blow down propellant tank, and as such a legitimate characterization of the energy density should include the mass of a tank. Based on available data for a composite over wrapped propellant tank, the mass of a propellant tank for a volume on the order of 10 liters would conservatively decrease the mass-specific energy density of 70% peroxide from approximately 2.0 MJ/Kg to approximately 1.7 MJ/Kg. Based on this and the measured values of conversion efficiency and actuator power density previously described, the actuation potential for this single-degree-of-freedom system, as given by (1), would be 15.3 KJ KW/Kg2. As previously mentioned, the power density will increase for a multi-degree-of-freedom system, and thus so will the actuation potential. For a six-degree-of-freedom system, for example, the total actuation system mass would be 5.2 Kg, or 870 g per actuator. The reservoir used in the single-degree-of-freedom experiment was oversized, and is appropriately sized for a power-comparable six-degree-of-freedom system. The actuation system power density would therefore increase to 172 W/Kg, and the corresponding actuation potential to 26.4 KJ KW/Kg2 for the six-degree-of-freedom system.
For purposes of comparison, the best commercially available rechargeable batteries have energy densities of approximately180 KJ/Kg (e.g., Evercel M40-12 nickel zinc, or SAFT 27 10 LAS silver zinc). A rare-earth permanent-magnet dc motor with a harmonic drive gear head with output characteristics capable of achieving the trajectory specified by Table I, has a power density of approximately 48 W/Kg. Note that this remains invariant, regardless of the number of degrees of freedom. Finally, one can assume that the overall conversion efficiency would be the combined efficiencies due to pulse width-modulation (PWM) control, the motor, and the gear head. The PWM efficiency was estimated to be 95%, the motor efficiency calculated for the desired trajectory to be 90% (i.e., the resistive power loss in the motor windings was calculated given the desired torque), and the harmonic drive gear head efficiency was estimated based on manufacturer data to be 65%. The resulting actuation potential for this type of system would therefore be 4.8 KJ KW/Kg. The poorly insulated single-degree-of-freedom experimental setup with 70% peroxide therefore exhibited an actuation potential more than three times a state-of-the-art battery/dc motor system. A similar six-degree-of-freedom system would exhibit an actuation potential over fives times the battery/dc motor system.
4.5. Projected Performance for High-Test Propellant
Though improvements can clearly be made with improved insulation and control performance, the most obvious means of improving the actuation system performance is to substitute a fully concentrated version of the propellant (i.e., 100%hydrogen peroxide) in place of the 70% solution used in the previously described experiments. Though procedurally quite simple, such experiments cannot be performed on commercially available pneumatic components, due to the high decomposition temperatures. Specifically, the adiabatic decomposition temperature of 100% peroxide is approximately 1000 0 C (1800 F), compared to approximately 230 0 C (450 F) for a 70% solution. Rather than conducting experiments using 100% peroxide, one can obtain a reasonable estimate of performance with projections based upon the experiments conducted with 70% solution. Upon replacing 70% propellant with 100% (technically 99.6%), at least two of the three parameters forming the actuation potential figure of merit would be expected to increase. Specifically, since the propellant contains more peroxide per unit mass, the heat of decomposition increases by a factor of 1.45 from 2.0 MJ/Kg to 2.9 MJ/Kg.
Additionally, the relatively low conversion efficiencies observed earlier were primarily due to the heat required to vaporize the water in the reaction product. Since the 100% propellant contains less water, less energy is invested in vaporizing the reaction product. Recalculating the expected efficiencies accounting for the reduced water content, the conversion efficiency scales by a factor of 1.56. Assuming that the actuation system power density remains invariant (i.e., that it does not increase with the 100% propellant), the single-degree-of-freedom system shown in Fig. 2 with 100%propellant would be expected to have an actuation potential of 35 KJ KW/Kg2 , which is 7.3 times greater than the battery /dc motor system. A similar six-degree-of-freedom system would exhibit an actuation potential of 60.4 KJ KW/Kg2, more than an order of magnitude greater than the battery/dc motor system. The promise of such performance, which would presumably be further improved with better insulation and light weight components, justifies the fabrication of custom high-temperature pneumatic components.
5. CHALLENGES OF DESIGN
The biggest challenge in using monopropellant as a power source is providing adequate insulation to prevent the heat loss from the system. We have seen from the experimental results that the heat loss exceeds the power output obtained from the actuator. Finding a suitable method to contain this heat loss is the first and the biggest challenge in designing a monopropellant based power supply.
Another problem is the non availability of parts that can withstand the heat produced on the decomposition of 100% hydrogen peroxide. Due to this reason, a hydrogen peroxide solution of lower strength has to be used. Materials more resistant to heat are to be used to make the parts of the system so that it can withstand higher temperatures. This will aid the use of higher concentrations of peroxide thereby increasing the actuation potential of the system.
Monopropellants are highly reactive materials and are even toxic to humans. It has a tendency to catch fire if spilt on clothing. So the persons handling the fuel should be extremely cautious in order to avert possible danger of explosion and intoxication. The conventional power systems do not have such problems. Hence extra care must be taken while selecting the materials to be used in the system.
The selection or rejection of a proposed design depends heavily on its economic aspect. Hydrogen peroxide powered system when compared with battery operated power system is very costly in running condition. Also due to the presence of valves, maintenance costs of the system are high as well. The choice of valves can also influence the reliability of the system. The next challenge in designing monopropellant powered systems is attaining proper co ordination between the different control loops which controls the operation of the system. The emission of hot steam might be an inconvenience to other human workers if such robots are used along with humans. So it is better to limit the use of such robots to places inaccessible to humans. Controlling the emissions from the robot can make it usable along with humans. This can also improve the efficiency of the system.
A power supply and actuation system appropriate for a position or force controlled human-scale robot was proposed. The proposed approach utilizes a monopropellant as a gas generant to power pneumatic-type hot gas actuators. Experiments were performed that characterize the energetic behavior of the proposed system and offer the promise of an order-of-magnitude improvement in actuation potential relative to a battery powered dc-motor-actuated approach. Experiments also demonstrated good tracking and adequate bandwidth of the proposed actuation concept.
Steam powered robots are a possibility in the future provided the limitations of the existing prototype is done away with. A better actuation potential can be obtained by providing better insulation to the prototype thereby reducing the heat loss. Another challenge before researchers is to manufacture parts that can withstand the high temperatures generated on decomposition of 100% H2O2 .With the introduction of better controls, fuel and insulation, the se robots could function effectively and economically.
The proposed power supply was found to be a feasible solution to the problem of providing a long lasting power supply to robots that can actually work. Moreover the power output could be easily adjusted by controlling the rate of flow of the monopropellant. Although a full size human scale robot powered by a monopropellant is yet to be made, the experimental results obtained from a single degree of freedom manipulator proves the feasibility of such a system.