Design of Rover Wheel Testing Apparatus

Size: px

Start display at page:

Download "Design of Rover Wheel Testing Apparatus"

Dorothy Quinn
5 years ago
Views:

1 Design of Rover Wheel Testing Apparatus Daniel Flippo University of Oklahoma Abstract A Rover s wheel performance is one of the limiting factors in its ability to successfully traverse other planets. More research focusing on wheel materials, design, and optimized tread patterns needs to be conducted so that a rover can reach more of its potential. A wheel test bed is being fabricated to test new wheel designs and to collect data which will be used to map the relationship between a wheel s design and its performance on different planetary terrains in different mission roles. The Suspension and Wheel Evaluation and Experimentation Testbed (S.W.E.E.T.) can measure traction, sinkage, lateral force, compliancy, and rolling efficiency for each wheel tested. The data collected by the testbed will be used to understand and fabricate wheels that are specifically suited for particular planetary mission tasks. These tasks will go beyond the current missions of exploration and include such roles as regolith excavation, equipment transportation, and equipment manipulation. 1. Introduction America has had a successful rover presence on the Moon and Mars. The latest two Mars rovers have started their forth year of exploration on the red planet. Their mission is to take pictures and analyze Martian terrain. These two rovers, dubbed Spirit and Opportunity, have vastly exceeded NASA s expectation of 90 day missions. They have sent back some amazing pictures of the Martian landscape and terrain, but their exploration is only scratching the surface of Mars s geographical features. Mars features Olympus Mons, which is the largest mountain in the solar system, towering at a height of 27 km, and is also home to Valles Marineris, a network of canyons that wind 4000 km along the planet s terrain and range from seven to ten kilometers deep [14]. Why is NASA not exploring these features instead of the flat plains of sand and small to medium sized stones it currently examines? The reality is that NASA is understandably squeamish about taking their 400 million dollar rovers to such dangerous locations for fear the equipment may get damaged or immobilized. Even in the current mission, NASA s fears proved to be valid when Opportunity became mired in a sand dune for five weeks [3]. Engineers eventually inched it out, but if NASA had wheels that gripped better in sand they might have avoided this mishap entirely. It is a simple design change to make a wheel perform better in sand, but most likely that change will be at the expense of performance in other areas such as rolling efficiency and turning ability. This leads one to ask what is the best wheel design for every mission? An optimized tread, wheel size, and overall design should vary according to the mission s specific goals. The wheel tread for a rover doing long distance exploration through sand dunes should be different than the tread for a rover that works in a local arena digging samples or clearing landing sites. A "one tread fits all" approach limits the rover s potential. Very little research has been done in the area of interplanetary wheel tread and wheel design as opposed to the tires used on Earth that don t have the material and weight constrictions. As NASA and other agencies expand their endeavors to other worlds, and establish a presence on Mars and again on the Moon, they must pay close attention to the basics of wheel and tread design and wheel soil interaction. At some point in the future wheels will not be the only form of locomotion on other planets. Legs and other robotic forms will be used to overcome many of the transverse problems inherent in natural terrain, but until that time, rover designers will need wheeled rovers that will traverse to more places of interest on Mars, as well as do the work of preparing the planet for human occupation. As expectations for rover performance rise, rovers will become bigger and more complex and it will be critical that wheels have optimized performance with respect to their missions. 1

To fill the gap in the understanding of rover wheel design and wheel to soil interaction, testing machines have been designed by various institutions.

NASA now uses devices such as the variable terrain tilt platform (VTTP), at JPL, to gain a better understanding of entire rover systems in a sloped environment.

2 To fill the gap in the understanding of rover wheel design and wheel to soil interaction, testing machines have been designed by various institutions. In 1971, NASA tested the Lunar Rover Vehicle s wheels on a testing device called a dynamometer system which measured load, sinkage, pull, torque, horizontal, and angular velocity [10]. NASA now uses devices such as the variable terrain tilt platform (VTTP), at JPL, to gain a better understanding of entire rover systems in a sloped environment. The VTTP is a 16 x 16 ft table that can tilt up to 25 degrees and can be left bare or covered with terrain [8]. At the Massachusetts Institute of Technology a testing device dubbed the "Field and Space Robotics Laboratory terrain characterization testbed" tests a single driven wheel through different mediums to better understand wheel to soil interaction [5]. A similar device is used at Tohoku University to refine rover steering and other parameters [15]. Other comparable devices test wheels for Earth s terrain [9, 12, 2] This paper describes a testbed that aids in our understanding of wheel to soil interaction, as well as tests new wheel designs. The following sections will describe the testbed s design, testing mediums, and proposed experiments, while the final section will summarize the information and give a small outline of how the experimental data will be used. enough to be used to test other assemblies such as a suspension system or an entire rover. Figure 1: Testbed 2.1. Motion To simulate motion on the testbed, the table moves under the test leg. This motion is facilitated by three DC geared omni-directional wheels that are offset 120 degrees from each other (Figure 2) 2. S.W.E.E.T. design The purpose of this testbed is to test new wheel designs and collect data that will be used to map the relationship between a wheel s design and its performance on different terrains. The testbed (Figure 1) has a 10 x 10 ft footprint and is fabricated from modular aluminum. A weighted drop down test leg, incorporating a driven wheel and a six axis, force torque sensor, stays stationary in the X and Y directions but allows movement along the Z-axis via a counterbalance system. S.W.E.E.T. differs from those discussed earlier in that the table can move in the X and Y directions underneath the test stand, as well as rotate about the Z-axis. This added advantage gives the apparatus the unique ability to measure forces and torques in a true turn. This testbed can measure traction, sinkage, lateral forces, turning efficiency, compliancy, and rolling efficiency for each wheel tested. The table can move along the cm. This X and Y -axis at velocities faster than 20 sec is more than needed considering Spirit or Opportunity but it will allow for testing the emerging, faster rover concepts [11]. S.W.E.E.T. is also large Figure 2: Motor configuration To transform the desired cartesian table motion of X, Y, and Θ to the angular speed of each of the three motors, a transformation equation (1) was used [13]: 1 ξ = R(Θ) 1 J1f J2 φ (1) In this equation, ξ is a vector containing the desired table motion parameters of X,Y, and Θ, rep2

3 resented as: ξ = Ẋ Ẏ Θ (2) R Θ is the rotational transformation matrix that is dependent on the angular position of the table Θ. cos(θ) sin(θ) 0 R(Θ) = sin(θ) cos(θ) 0 (3) J f is a matrix made up of three constraint equations, one for each wheel. J f transforms the desired motion values of ξ into the control parameter, φ, which is a vector containing the three motor angular speeds. Figure (3) shows how J f is dependent on the three angles and the distance, l, the wheel is away from the center of the rover. maps a relation between the table motion (Ẋ, Ẏ, Ż) and the motor angular velocities (ω 1, ω 2, ω 3 ). ω 1 ω 2 ω Test leg = 1 J f R(Θ) r wheel Ẋ Ẏ Θ (7) The test leg (Figure 4) is a fully adjustable assembly that hangs from the center of the pyramid shaped apparatus, and is free to move along the Z-axis using linear rod bearings. The test wheel is powered by a FaulHaber DC motor geared down to (43:1) and then again geared down (2:1) via a bevel gear set. An encoder is used on the motor for PID control. The test leg also holds the force torque sensor, discussed in section 2.4. Figure 3: A Swedish wheel and its parameters [13] J 1f = sin(α 1+β 1+γ 1) cos(α 1+β 1+γ 1) lcos(β 1+γ 1) (4) For the configuration in this apparatus, β and γ are zero, making J f dependent only on α. If all three wheels are taken into account and set up in matrix form, the result is equation (5): J f = sin(α 1 ) cos(α 1 ) l sin(α 2 ) cos(α 2 ) l sin(α 3 ) cos(α 3 ) l (5) J 2 is a matrix holding the radius values of each wheel. Since each radius is the same J 2 can be simplified to a scalar value r. J 2 = rad rad rad 3 (6) Taking into account equations (2) and (6), we can simplify equation (1) into equation (7) which 2.3. Electrical Figure 4: Test leg Most of the electrical system is housed entirely in the control box shown in Figure (5). Three 240 Watt power supplies along with three 40 Amp PWM controlled H-bridges provide power and control for the main motors. The control box is also equipped with a 200 Watt power supply that powers the encoders, testing wheel, and force 3

Figure 5: Control box Figure 7: Force torque sensor installation torque sensor. Another H-bridge controls the test wheel and has a current sense that is used to record the test wheel wattage.

A PCI-6602 DAQ board handles all the digital and counter, PWM signals to the H-bridges, while a USB-6020 DAQ board handles all the force torque analog signals, as well as encoder signals of the four

When executed, the program initially reads from a test procedure file into an array that controls the program during a test, allowing for different experiments.

While the test is running, data from the test procedure file is appended to sensor readings and written to a test result file. Shown in Figure (8) is one of the control screens in the program.

4 Figure 5: Control box Figure 7: Force torque sensor installation torque sensor. Another H-bridge controls the test wheel and has a current sense that is used to record the test wheel wattage. The apparatus utilizes two National Instruments data acquisition boards linked to a computer. A PCI-6602 DAQ board handles all the digital and counter, PWM signals to the H-bridges, while a USB-6020 DAQ board handles all the force torque analog signals, as well as encoder signals of the four motors and the Z-axis displacement 2.5. Programming Since the Testbed uses two National Instruments data acquisition boards, all the control and measurement programming was done in Labview 8.2. The programming incorporates PID control algorithms for each of the motors. When executed, the program initially reads from a test procedure file into an array that controls the program during a test, allowing for different experiments. The test procedure file also can direct the program to control motion by force feedback, simulating dragging and other parameters that the user may wish to incorporate. While the test is running, data from the test procedure file is appended to sensor readings and written to a test result file. Shown in Figure (8) is one of the control screens in the program. This particular screen is used to visually inspect and adjust the PID coefficients for better control Force torque sensor Like the devices at MIT and Tohoku University [5, 15], the main sensor used in the Testbed is located on the test leg, along with the sensor s adjustable gain amplifier, just above the wheel (Figure 7). It is a six degree of freedom force torque sensor machined from 2024 aluminum, incorporating 32 strain gages networked together [4]. This sensor gives us a full view of the forces and torques being applied to the test wheel. Both the sensor and the amplifier were made at the University of Oklahoma. Figure 8: Labview interface control screen Figure 6: Force torque sensor device 3. Testing medium 4

5 Different terrain mediums will be used for testing. Those of particular interest are mediums that mimic the terrain of Mars and our Moon. Sand will be the most common and will come in several grain sizes, densities, and slopes. Rocks will also be used as a medium in varying sizes and orientations depending on the test. Each of the differing terrains will be housed in bins that are easily rolled in and out of the machine for the different tests being done. 4. Experiments Several tests are possible for the apparatus in such categories as traction, sinkage, turning efficiency, compliancy, and rolling efficiency in varying terrain. The following are a few of the tests that are possible with this apparatus. The first, is intended to test static traction in each wheel and is the simplest to perform. With the wheel weighted down, the test motor is ramped up over a certain period of time, from zero to full speed while the table is motionless. As the power is increased to the wheel, the measured force increases along its line of motion. At some point, the force overcomes the frictional force between the wheel and the media, causing the wheel to slip. This point is apparent in the results when the force in the direction of the line of motion levels off. This can be used to determine the static coefficient of friction between the terrain and the wheel being tested. The second test is similar to [1] and is concerned with wheel sinkage on loose terrain. For this test, the table will be moving along with the wheel but will simulate a drag by using a force feedback loop. This loop will move the table along the line of motion as long as there is a set amount of force felt by the wheel. The force, as well as the speed, are all prescribed in the test procedure file. As the wheel spins, it will sink in the loose terrain and the Z-axis encoder will measure the displacement incurred. A variation to this test, which is similar to [7, 6] is to orient the terrain on a slope simulating a hill. This apparatus is large enough that a sloping bin can be installed on the table and the test carried out as before. Steering is a very important part of a rover s design, so another test would be to discover how efficiently a wheel can turn in varying terrains [15]. Several variations to this experiment are possible depending upon what type of turning mobility is being tested. The test apparatus can simulate Ackerman steering by moving the table beneath the stationary wheel in a curving motion and measuring the forces and moments incurred. If a skid steer wheel is being tested, the table is programmed to go at an angle, dependent upon the rover s geometry, while the test wheel rotates at a set speed. A third experiment is to test for compliancy with each wheel. The test wheel can be driven off a set height and dropped onto varying terrains and with varying weights. The Z displacement and all accelerations can be measured through this test to see which wheels do better under varying conditions. The final initial experiment proposed, with this testbed, measures the rolling efficiency of a rover wheel. This is done by recording the wattage used by the test wheel as it traverses different terrains, at different weights, and with opposing forces. This will give us a better understanding of how tread and overall design affects efficiency. 5. Conclusions In this paper, a testbed for rover wheels was described. The physical parameters of the machine as well as the electrical system and control program were explained. Uses and proposed testing experiments designed for this apparatus were also detailed. The data obtained from these experiments, in the form of performance values in several categories, will allow attempts at linking a wheel s design using a neural network. This link will be used by a genetic algorithm to evolve an optimized wheel for a given mission. The testbed has been completed and preliminary results will be reported at the conference. 6. Acknowledgments This work was supported in part by a grant from Malin Space Science Systems, Inc. The author would also like to acknowledge the support and help of Dr. David Miller and those at the University of Oklahoma robotics lab. References [1] C Brooks, K Iagnemma, and S Dubowsky. Visual wheel sinakge measurement for planetary rover mobility characterization. Autonomous Robots, 21(1):55 64, August [2] Koichiro Fukami, Masami Ueno, Koichi Hashiguchi, and Takashi Okayasu. Mathematical models for soil displacement under a rigid wheel. Journal of Terramechanics, 43(3 SPEC ISS): , Soil displacement;soil bin test;rigid wheel;displacement increment;gaussian func- 5

6 tion;displacement locus;first derivative;second derivative;. [3] Natalie Godwin and Dolores Beasley. Nasa s opportunity rover rolls free on mars. News Release, [4] Junichi Hayashi. Device for measureing compoents of force and moment in plural directions. Patent , United States Patent Office, May [14] Martin J.L. Turner. EXPEDITION MARS. Springer-Praxis, London, [15] G. Yoshida, K.; Ishigami. Steering characteristics of a rigid wheel for exploration on loose soil. Intelligent Robots and Systems, (IROS 2004). Proceedings IEEE/RSJ International Conference on, 4: vol.4, 28 Sept.-2 Oct [5] Karl Iagnemma, Hassan Shibly, and Steven Dubowsky. A laboratory single wheel testbed for studying planetary rover wheel-terrain interaction. Technical report, MIT Field and Space Robotics Laboratory, 77 Massachusetts Avenue, Room 3-435a, January [6] G. Ishigami, K. Nagatani, and K. Yoshida. Path following control with slip on loose soil for exploration rover. pages , Beijing, China, 2006//. path following control;slip motion;loose soil;lunar/planetary exploration rovers;slip compensation;wheel-and-vehicle model;terramechanics;. [7] Genya Ishigami, Akiko Miwa, Keiji Nagatani, and Kazuya Yoshida. Terramechanics-based model for steering maneuver of planetary exploration rovers on loose soil. Journal of Field Robotics, 24(3): , [8] Randel Lindemann. Platform for testing robotic vehicles on simulated terrain. Technical report, NASA, June [9] UENO MASAMI, OMINE MASAAKI, SHIKANAI TAKESHI, HASHIGUCHI KOICHI, and OKAYASU TAKASHI. A precise prediction model of traveling performance for a rigid wheel on sandy ground. Journal of the Japanese Society of Agricultural Machinery, 61(2): , [10] K.J. Melzer and A.J. Green. Performance of the boeing lrv wheels in a lunar soil simulant. Technical Report 1, Mobility and Environmental Division U.S. Army Engineer Waterways Experiment Station, December [11] Matt Roman, Dr. David P Miller, and Zack White. Roving faster farther cheaper. In Proceedings for FSR, Chamonix, France, July Field Service Robotics. [12] I. Shmulevich, D. Ronai, and D. Wolf. A new field single wheel tester. Journal of Terramechanics, 33(3): , [13] Roland Siegwart and I. Nourbakhsh. Introduction to Autonomous Mobile Robots. MIT Press,

Study on Effect of Grousers Mounted Flexible Wheel for Mobile Rovers

Study on Effect of Grousers Mounted Flexible Wheel for Mobile Rovers Kojiro Iizuka and Takashi Kubota 2 International Young Researchers Empowerment Center, Shinshu University, iizuka@shinshu-u.ac.jp 2