Kurt Schwehr schwehr _at_ cs stanford edu 8/93 Revised: Aug. 17, 1993 HTML Upated: Feb, 1995
A.1.0 Autonomous Rovers A.1.1 Introduction A.1.2 Mission Ready Rovers A.1.2.1 Flown Rovers A.1.2.1.1 Lunokhod 1 & 2 A.1.2.1.2 Mars 2 & 3 A.1.2.2 Currently Fliable Rovers A.1.2.2.1 Marsokhod A.1.2.2.2 Rocky IV A.1.3 Development Platforms A.1.3.1 Ambler A.1.3.1.1 SITE-Recon Mission A.1.3.2 Attila A.1.3.3 Dante-Virgil A.1.3.4 Robby A.1.3.5 Others A.1.4 Virtual Reality A.1.4.1 Introduction A.1.4.2 Telepresence A.1.4.3 Virtual Environments and Geographic Information Systems A.1.4.4 Entertainment A.1.4.5 Conclusion
There are currently two camps of robotics that tend to argue back and forth along the major division in this field: simple, small "micro-rovers" verses larger more complicated rovers. There are definite trade offs between the two design approaches. I will cover both types of rovers and will present a concept for a combined mission that tries to use different scales of rovers to make a more robust system. Using micro rovers could send a large number of small micro-rovers in the space and weight required for one or two regular rovers to Mars. There would be less concern over loosing a percentage of micro rovers compared to loosing on of 1 or 2 larger rovers. On the other hand, a single larger rover can have more onboard computational power and larger scientific instruments.
There is also the split between legged vehicles, wheeled vehicles, and vehicles that integrate the two paradigms. Legs provide a more stable platform for instruments, are more energy efficient for motion, and cause less environmental damage to the area explored, but they require higher computational loads in order to figure out where to place each foot. Wheeled vehicles are mechanically simpler, computationally less expensive and more reliable (?), but require more power for moving the same distance and leave behind wheel tracks.
Price per Unit: $60M (delivered to martian surface) Total Mass: 100 kg Life-Time: 1 Terrestrial Year Covered Distance Goals: 100 km/Year, 50 to 100m/Day Velocity: <180m/h Electrical Power: 10 to 20W (RTG) Chassis: 31 kg Articulated Frame with 6 cone wheels, by Vniitransmach Payload: 14.5 kg (8 kg science + drilling system) Width: 60 cm Length: 90 cm (Ref. Runavot, 1993; JPL Universe, 1993) The Marsokhod 96 Table A.1.2.2.1-1
The two main problems with this small of a vehicle are 1) the communication systems and many of the instruments do not scale down well with the rest of the rover and 2) with this design philosophy, this vehicle is not usable for telepresence operations. The scaling problems can be solved by sending multiple rovers, each with a different set of science instruments, and by relaying the rovers data through a single uplink station on the surface of Mars.
For a recent press demonstration, the JPL prototype weighed in at 7 Kg and is 61 cm long by 38.5 cm wide by 36 cm high. It has six 13- centimeter (5-inch) diameter wheels made of strips of stainless steel foil with cleats to provide traction. Rocky runs on 5 watts of solar energy, used during the day. At night, the electronics are turned off, and the keep-alive batteries run the unit. Rocky, was controlled by a Macintosh PowerBook for the demo. Rocky is able to ascend slopes of 26¡. It is equipped with a visible infrared spectrometer, a color camera, a rock chipper, and a soft-sand scoop takes soil samples. Additionally, the rover can place a seismometer on the surface (Ref. JPL Universe, 1993). The control system is based on the subsumption architecture of Brooks, known as behavior-based control. The system currently runs on a tiny 1-MIP Motorola 6811 micro controller with 40k of memory (Ref. Gat et al., 1993.) A space-qualifiable version of Rocky IV scaled to approximately 4 kg is currently being designed for a planned 1996 MESUR Pathfinder launch (See Section 2.2.2.3.4).
Price per unit: $2.5M (includes landing system) Total Mass: 7.5 kg Life-Time: 7 Days (minimum) Coverage Distance Goals: 100 m Velocity: ~1 m/min. Electrical Power: 100W Hr/Day (Solar) + 150 W Hr non-rechargeable battery Chassis: Rocker/Bogie 6 wheel system Power Consumption: 14.7 W Hr/Day; 8.0 W Hr/Night(Ref. Reynolds, 1993) 1996 JPL Rocky IV
Mass: 2500 kg Height: 4.1 to 6.0 m Width: 4.5 to 7.1 m Locomotion: 6 legs - circulating gate Speed: 35 cm/min. Power Consumption: Steady-State: 1400 W Laser Scanner: + 210 W Leg Motion: + 150 W Horizontal Body Motion: + 600 W Vertical Body Motion: + 1800 W
Code named: "Bulwinkle"
The SITE-Recon mission proposed in Section 2.2.2.5 is based on a modified cargo vehicle. The mission as whole is meant to demonstrate working versions of a number of untested systems by sending them on a useful Mars precursor mission. In particular, the mission contains an unusual array of robotic rovers. The mission proposes two groups of rovers: 1) 10 Marsokhod Rovers and 2) one second generation Ambler rover (named "Bulwinkle") that is parent to 10 Rocky microrovers. The driving force behind this assortment of rovers is that we do not know what the "best rover configuration is so we must try a number of different combinations in actual field use to gain more hands-on experience.
The primary reason for the choice was the large amount of local computational power available on the Ambler (currently two Sun workstations) that allow the Ambler and subordinate rovers to be almost entirely autonomous short of deciding the high level mission goals. This is important, in that the mission design already has ten Marsokhod rovers which use extensive human interaction to perform their missions.
A large part of this mission is a technology demonstration. The combination of ten Rocky IV rovers and an Ambler allow for complicated grouping behaviors to be tested. This will increase the useability of the group without having complex computational facilities located on each and every rover. See Gat's paper (Ref. Miller, 1990) for more detailed description of grouping behaviors in simple/small rovers. The group behaviors can be implemented with simple high level commands from the Ambler which is a base that provides a communication base/beacon/repeater with an antenna that is high enough up that it should be visible by most of the group at any time. The Ambler will be able to carry one large communications system that will allow the group to communicate back to the mother ship (the cargo vehicle) over a much longer ground distance with higher data rates.
The taller Ambler will allow us to mount a meteorology station above the surface layer and take this station a good distance from the landing site. Along with this meteorology station will go a nice imaging system. As Professor Don Lowe of Stanford University has said, a lot of excellent geology can be done with good, high resolution imaging with a zoom lens. If we were able to mount such an imaging system so high above the martian surface and be amble to move it around, we would be able to explore areas a long was away from the roving group without actually have gone to these sites.
It is assumed that the Ambler will have a much larger range than the Marsokhod. As the Ambler makes its traverse, it serves at a refueling depot for its daughter rovers (i.e.. recharging of batteries). This extends the effective use of the Rocky rovers.
Finally, the mission already contains ten Marsokhods, which only use up a small portion of the available 4000 kg payload available on the cargo vehicle. It would be silly to propose that we send another ten to the same area when we have an opportunity to test other systems. Admittedly, a sample return vehicle would probably fit nicely in this space/weight, but this would be a repetition of many previous proposals - numerous authors have discussed a variety of sample return missions. We need to try out some new ideas and get creativity flowing. This is an excellent opportunity to combine two different rover methodologies and do lots of science at the same time.
At the 1992 California desert rover tests, a ramp was added to the Russian Marsokhod rover so Attila and Pebbles could ride piggy back on the Marsokhod (Ref. Bullock, 1993.) The idea was to provide a smaller vehicle that could carry a sampling tool into tight areas and also as an extra set of eyes should the parent rover get stuck. This is very similar to the SIMM '93 SITE-Recon mission concept (See Section 2.2.2).
Amazingly, the vehicle was designed, built, and deployed in only a year. The vehicle has eight legs, weights 450 kg and is 1.8m x 2.5m x 3m in size. The complete system consisted of four nodes: Dante, Virgil, Base Station (Erebus Hut), and Mission Control (NASA Goddard Space Flight Center) Dante's journey was computationally expensive: it used a total of 5 sun4 and 3 Motorola 68030 VME computers throughout the system. See Table A.1.3.3-1 for more of the vehicle parameters.
The project was lead by William "Red" Whittiker of the Field Robotics Center, Carnegie Mellon University who planned and executed the project. The project was also co-lead by Philip R. Kyle of the New Mexico Institute of Mining and Technology. Dante is a synthesis of the previous work done at CMU on the NavLab and Ambler projects. "The robot demonstration project had three objectives: to test telerobotic capabilities; to test the use of such sophisticated hardware in a very harsh and demanding environment; and to test the use of advanced computer programs which would enable machines such as the Dante robot to act under a form of machine intelligence." The development and telepresence aspects of the process were considered a success despite the projects failure to reach the caldera floor and David Lavery, NASA Telerobotics program manager for the project feels "[t]he prototypes are worthy contenders for inclusion in any further planetary exploration."
DANTE
Propulsion: 4 legs on each of 2 frames Height: 2.5 m Width: 1.875 m (foot to foot) Length: 3 m (foot to foot) Ground Clearance: 0 to 1.45 m Speed: 2 m/min. Slope Handling: 30* without tether Telemetry: 400 m fiber optic (in tether) Power: Virgil via tetherVIRGIL
Height: 1 m Width: 1.2 m Length: 4 m Weight: 4540 kg Speed: 30 kph Track: 2.6 m Ground Clearance: 0.1 to 0.4 m Power: Internal combustion engine & 5 kW Honda generator Dante-Virgil TABLE A.1.3.3-1
Robby is rather large: 3 segments, for a total of about 4 meters long and 2 meters with; six wheels, 1 meter in diameter each. It weights in at 1200 Kg. Robby's design is able to surmount obstacles 50% larger than the wheel diameter.
The general population has a strange view of virtual reality . This comes from seeing such things as Hollywood movie "Lawnmower Man." When asked about VR, they often bring up thinks like cyberpunks, "virtual sex" and Star Trek the Next Generation's "Holodeck." This does not have much of a relation to the current technology. It is important to clear your mind of the tremendous hype while thinking about what it can do for planetary exploration.
However, despite being way out of our technological reach, the Holodeck does bring in the right feeling of what virtual reality is about. In a number of episodes the crew of the Enterprise explore copies of the real world. It allows them a deeper view a situation or a system. The Holodeck is very similar too and must of come from "the Cave" - an MIT project that surrounds a person in a room that has screens on all surfaces, thereby immersing the person in the environment projected by the video screens. A good example of the power of this immersion can be experienced at such places as the Disney Epcot Center's theater in the round. This is essentially a large circular auditorium that has movie screens completely surrounding the audience. This total immersion has a much greater power to instill the immensity of such things as the Grand Canyon as a simple flat screen could. The audience feels like they are there, because everywhere they turn is still the Grand Canyon. As the film drops into the canyon, the standing audience has to grab hand railings for balance since optic righting reflexes are stimulated. The goal is to make the person feel like they are really there.
Geologists have said that, "The complex yet subtle nature of geological materials requires powers of observation, pattern recognition, and synthesis not possessed by automated devices. Field study...absolutely requires human geologists to be involved intimately." Telepresence technology provides us a way to get the scientists deeply involved in their field areas even when they are out of human reach, but we must be careful in applying telepresence: "if the remote operation becomes too cumbersome...the operator will concentrate more on mechanical aspects of the work and less on the intellectual ones" (Ref. Taylor and Spudis, 1989).
Rudimentary telepresence is achievable using only limited computational power. For example, the Telepresence Demonstration Project at NASA Ames achieved usable telepresence with their remotely piloted submersible using a simple Motorola 68000 based Commodore Amiga system (Ref. Schwehr, 1992). Unfortunately, most planetary exploration applications will need more computational power and more complex programs to function up to expectations. Using this platform, the group has shown that telepresence can be an effective tool for scientific exploration.
The Telepresence Demonstration Project, run by C.R. Stoker, uses a submersible remotely operated vehicle (ROV), a Deep Ocean Engineering SuperPhantom II, as a development platform. The group added a head tracked camera and science instruments to the ROV for an initial testbed. The vehicle has opperated successfully in several very different environments: a coral reef in the Florida Keys, the south shore of Lake Tahoe, and the ice covered Lake Hoare of the Antarctic Dry valleys. The trip to Antarctica is the most important of these expeditions because according to C.R. Stoker, "Antarctica is the most Mars-like environment on Earth." (Ref. NASA PR 92-147) The ROV is a big win for the group in allowing scientists to get around physiological constraints of diving, thus allowing the scientists to spend more time exploring their field areas. This is very similar to the constraints on astronauts due to life support systems. The telepresence system of the ROV succeeded in helping to improve the pilot's situation awareness while exploring underwater. In all three cases the vehicle was restricted to being less than 1100 from the pilot at the control console.
While the ROV was in Antarctica, the software was extended on the ice by a brilliant field engineer to allow for an operator to control the ROV from Ames while the vehicle dived under the ice near McMurdo base. The test was the first success of direct telepresence via satellite link. In December, 1992, the CMU Dante robot used this same satellite link to control the robot on Mt. Erebus from state side.
Recently, Butler Hine's group at NASA Ames conducted further remote tests with a Marsokhod Rover located in Moscow at the IKI laboratory (Ref. NASA PR 93-84). The tests were designed to verify this technology for use on the Russian Mars 96 mission. Hine describes the control system at NASA Ames as "a 'tele-operator interface' because it is a combination of virtual reality and telepresence. We can drive the vehicle by looking through the rover's cameras, which is telepresence. We also can drive it using a computer-generated graphic simulation, which is virtual reality."
The time delay in response of the telepresence system for all of these tests was between 1/3 of a second for direct control up to a few seconds for telepresence via satellite. These results show that for on sight or "on planet" situations, straight telepresence control of a vehicle or other robotic system works quite well, but there are several reasons to want to add layers between the operator and the machine. The most dramatic reason, is for teleoperation of robotic explorers on other planets from Earth. For example, the distance between the Earth and Mars causes round trip delays of 11 min. 20 sec to 40 min. 50 sec depending on the positions of the two planets. With time delays of more than a few seconds, direct telepresence becomes more of a detriment and will cause the operator to quickly become fatigued (and damn bored!) The solution is to use predictive systems that model the environment and allow an operator to control a computer generated rover that acts like the real thing with out the delay. The operator can then run through a series of operations with the model and then wait for the conformation that the rover has achieved this set of goals. Predictive systems are currently being developed at NASA Ames in Hine and McGreevys' groups and in Schenker's Man-Machine Systems Group at JPL.
Another reason for wanting virtual reality between the operator and the vehicle is to allow the person to work through several different scenarios to see the simulated results of each. Once he decides on the proper sequence a events, the operator can move through the sequence with confidence or have the simulator send off the correct actions of the operator that were generated and saved during the tests.
"One can integrate and automate many of the vehicle functions that are normally driven manually. Teleoperation may thus be raised to a supervisory level, relieving the operator of tedious tasks as piloting the vehicle from point-to-point. Automated control may be employed at the operator's discretion to free him of tiring tasks that range from the most mundane to the most complicated." (Ref. Gwynne et al., 1992) An excellent example comes from a joint project between Stanford's Aerospace Robotics Laboratory (ARL) and the Monterey Bay Aquarium Research Institute (MBARI), which is developing technologies for use on several different remotely operated submersible vehicles. The goal of their project is to design task level controls that are off-loaded onto local computers which should free the operator to focus more on planning and decision making than on the difficulties of piloting a vehicle. The group has developed "and demonstrated the capability of combined camera and vehicle tracking of underwater targets" - i.e., the group created a system that can autonomously follow objects (such as a plastic turtle or a fish) as they move through the water. (Ref. Marks et al, 1992). When integrated with the MBARI ROV, this system should make following fish for hours on end a much more pleasant task for the ROV crew - allowing the scientist to keep following a fish without the highly experienced pilot always having to run the show.
McGreevy and Stoker (1991) conclude that "geologic field work consists of highly integrated perceptual, cognitive, manipulative, and locomotive behaviors organized around a hierarchy of scientific goals and methods, and it is unlikely that automation and robotics will be able to replace this human capability any time soon. Thus, for missions beyond the mere rote gathering of rocks, that is, for planetary surface field work, human presence is required. To effectively explore large areas of Mars or other terrestrial bodies, telepresence offers the best combination of man and machine."
As these massive data bases continue to expand we need a good way to explore this model of Mars. Virtual environments (VEs) are the creation, in virtual reality, of physical locations using a variety of data sets. At the moment we have Viking and soon to come is the Mars Observer Data sets with which to create a virtual Mars to explore. An immersive interactive environment will allow scientists to explore Mars and get the best intuitive/visceral feel for the planet. It will also be valuable in getting astronauts intimately acquainted with the geologic setting before they even leave for Mars. They will be able to explore any part of Mars anytime through the trip to re-explore any location.
Geologist like to have an oblique view of area - i.e. like from a plane. "[The geologists] confessed to relying heavily on vision; theirs is a highly spatial business" say McGreevy and Stoker (1991). A VE can provide a close equivalent to a plane for the astronauts. The simulation will let them be able to quickly see from any angle they choose, even those not physically possible (like inside a mountain!)
A Marssuit heads up VR display can present the astronaut with terrain (topos), reference material, orbital photos while he is in the field. Similar systems have frequently been proposed for use in the zero-G environment around the space station Freedom (Ref. Fisher et al, 1988) as an EVA Spacesuit visor display. This system will allow easy hand-free note taking that is linked the astronauts location. The main impediment, speech recognition, should be up to speed by mission time.
A system similar to a ÒDatagloveÓ (Ref. Fisher, 1986) can be integrated into the suit to allow an astronaut to more easily interact with a virtual reality system. This system would sense the position and orientation of the astronautÕs arms, hands, and fingers. A system like this would be light weight and very trim (not bulky) with the use of small fiber optic flex sensors (patented by VPL research, Inc.)
