Date: Fri, 14 Oct 1988 17:40-EDT From: space-tech-request@CS.CMU.EDU To: "~/st/lists/stdigest" Subject: Space-tech Digest #11 [ Not much happening this week. Feel free to start some new topics! -- Marc ] Contents: Ollie Eisman Re: Sails too flimsy???? Steve Abrams Why not lasers? Marc Ringuette CMU Mars Rover project ------------------------------------------------------------ Date: Fri, 7 Oct 88 23:47:57 MDT From: SPACE EXPLORATION To: sedspace@doc.cc.utexas.edu, space-tech@cs.cmu.edu Subject: Re: Sails too flimsy???? Steve mentions that he would like to employ lasers for the small satellite's communication system. This sounds interesting, but I would like to point out that the same technology to deploy the solar-sails can be used to deploy a parabolic antenna. (or even use the solar-sails as a reflector, as Steve has said before.) I guess I would just like to hear some advantages of lasers over more conventional radio up/down links. (bandwidth is obvious) Ollie ------------------------------ Date: Sun, 9 Oct 88 20:03:37 CDT From: sedspace@doc.cc.utexas.edu (405986289 abrams) To: space-tech@cs.cmu.edu Subject: Why not lasers? >I guess I would just like to hear some advantages >of lasers over more conventional radio up/down links. Simple...cost. By putting an expensive receiver on the craft, inexpensive transmitters could be made & maintained down here. Relatively cheap lasers, a telescope, and a modulation front-end for dirtside and a charge-coupled device for space. Great aim isn't necessary, you don't need an FCC license, every university can afford a laser. The specific wavelength can be chosen to minimiz e atmospheric absorption and filters place at the other end to filter out bandit s who don't know the wavelength. Some sort of encryption/password scheme can prevent unauthorized, yet knowledgeable, access... No big deal...I didn't think it would be so controversial...radios are fine with me, except that aluminim sails are transparent at radio wavelengths... Steve ------------------------------ Date: Thu, 13 Oct 1988 01:18-EDT From: Marc.Ringuette@DAISY.LEARNING.CS.CMU.EDU To: space-tech@cs.cmu.edu Subject: CMU Mars Rover I promised earlier to describe what's happening here at CMU (that's Carnegie-Mellon University in Pittsburgh) with our Mars Rover project. Mars Exploration for the late 90's ================================== The Mariner probes are the best look at Mars we've had so far. Our pictures of a rock-strewn desert, and our knowledge of Martian soil composition, come from the two Mariners which returned pictures and data. The landing sites were chosen to be the flattest they could find, in order that the landers wouldn't tip over. This had the side-effect that the areas were geologically fairly boring. The next step in Mars exploration is to send a flexible, mobile robot to Mars to collect and study samples from different areas. The importance of the mission is split between observation, on-site testing of samples, and return of samples to Earth. The requirements for this vehicle are pretty stiff if we are to try one of the more ambitious and more useful of the possible missions. The trickiest part of the problem is to do autonomous motion and sampling. Light takes between 10 and 40 minutes to travel the round trip between Earth and Mars, so a vehicle operated from Earth would be extremely slow. Even worse, NASA's Deep Space Network has other jobs to do, and the rover will spend half its time on the far side of Mars. This virtually requires a vehicle which can move and take samples using on-board computers and Artificial Intelligence. The mechanical design of the vehicle is also difficult. The design is most highly constrained by a very low power budget - a few hundred watts to run a 1-ton vehicle. The CMU Mars Rover Project ========================== The CMU rover project, at Carnegie-Mellon University in Pittsuburgh, is a 3-year project to build a prototype rover which can 1. travel several hundred kilometers, reliably, over the period of about a year, traversing 1 meter obstacles and ravines 2. take core samples, aim instruments, and perform sampling and experiments as flexibly as possible 3. collect about 5 kilograms of samples and transport them to a return vehicle for return to Earth 4. weigh no more than about a ton 5. operate on about 300 watts of continuous power, supplied by a Radioisotope Thermal Generator (RTG) 6. operate efficiently even when not in communication with Earth The project has three main research areas: Mechanical Design, Sensing, and Control. The first group is building the vehicle, and is headed by Red Whittaker, a mechanical engineer who recently constructed a robotic vehicle to clean up Three Mile Island. The second group, sensing, is headed by Takeo Kanade, a Computer Science professor who has been involved with the NAVLAB autonomous truck. They are using a laser rangefinder and computer vision software to maintain a terrain map on board the rover. The third group is headed by Tom Mitchell, who does Artificial Intelligence work. His group is designing the software to do motion and sampling without human intervention. The Ambler ========== The original proposal had been for a rover with large, soft wheels which could ignore small obstacles. However, the mechanical design group determined that a walking rover could better satisfy the reliability, stability, and power requirements of the mission. A few hundred watts is almost no power at all, so a wheeled vehicle loses because it puts so much power into its ground interactions. A legged vehicle is mechanically more challenging, but is smoother in operation and very energy efficient. The Ambler has six legs, each of which has two joints which move in the horizontal plane and a telescoping z-axis which stays vertical. The horizontal and vertical directions are totally decoupled - the machine always stays level, and the two horizontal joints also stay level at all times. Each of the six legs is attached to a central pole at a different height, so they can move 360 degrees without running into each other. The bulk of the body hangs from the center pole, close to the ground. Here is a picture: | |--------------------------------- | | | | | | |--------------------------------- -----------------------------| |^ ^ U | | | | |Shoulder Elbow U -----------------------------| | U U | | U U -------------------------- U U | | U U | Body (with RTG, sampling,| U U | computing, robot arm, | U U | instruments) | U U | | U | | | | | | | | | | | | | -------------------------- | | | | | /_\ /_\ (Side view of the body and the lower two of six legs) (the elbow and shoulder move in and out of the page) ============================================================================= _________ _ /--------- __--_\\_ ------// __--__-- \_\_/ //\ __- --__-- \_\_ // __----\ -- | \_\ -//__---- \\ ______--_| |--- | \\ _-- ____--- -- _-___ \\ / --- \ --__-__ / / \ / --|| / / ------ || / / || / / || || (Top view. 5 legs planted, 1 recovering.) (Each leg moves in a different horizontal plane except for the telescoping z-axis of each) ============================================================================= The machine has a reach of about 4 meters and a height of about 4 meters. The laser rangefinder goes on top of the central pole. The Ambler will walk a bit like a crab, with five legs on the ground at all times. When a leg is lifted from behind the vehicle, it is moved all the way to the front of the vehicle to minimize the number of footfalls required. The body slides forward using the horizontal joints only, spending energy only on friction losses and ground sinkage. It moves almost like floating on water. The z-axis is used to hoist the body up and down, and to lift each foot for recovery to the next position. The machine moves very slowly (since the limiting factor is the ability to control the motion reliably, not the motion itself). The body averages a few centimeters per second, which is plenty as long as the machine can operate autonomously. The Software ============ Building the Ambler is about half the project. The other half, which is technically more on the cutting edge, is putting together a software system to reliably (VERY reliably) move the robot from place to place and perform sampling tasks. A terrain map (more or less a contour map of the immediate vicinity of the rover) is maintained by integrating data from the rangefinder. Other information, such as "this is a rock" or "this is black stuff that sticks to your feet" may be attached to the basic map. This allows motion planning to be done accurately. The rover will be controlled by commands from the control center on Earth, such as "go north as long as it's safe" "go back and pick up rock 13 and look at its underside" "follow this path to rock 15 and drill a hole in it, at this angle" "pick up one of those gray pebbles, about half an inch wide" "put this dust in your mass spectrometer" "aim your infrared sensor at anything unusual" The commands won't be in English, but rather will be specified in terms of frames and slots in a knowledge representation system designed by the control group. Realistically, there will probably be a team of geologists fighting over what is most important to do next. Rather than forcing them to do the optimization of exactly what to do, it will be necessary to have a planning system which can do the best thing given a set of goals of varying importance and difficulty. For instance, if one goal is very easy, you might as well do it first rather than a more desirable but much more difficult one. There are also background goals best monitored by the machine, such as "Never go too near a dropoff" and "Don't point your satellite dish away from Earth." A flexible geometric reasoning system will be a component of the software. It's important for the rover to have the ability to pick up a rock, and also to be able to notice if the rock was dropped by accident. This involves creating a general purpose planner which can generate expectations about what will be true in the world if all went as planned, and to check if this really happened. This is my area. ============= The project has just finished its first year (of three or four). A test leg is nearing completion, and the first integration of planning and sensing software and the rangefinder hardware is a few months in the future. We're funded by NASA and interact off and on with groups from JPL, TRW, and Martin Marietta. JPL in particular has a parallel project, designing a more conventional wheeled rover for the same mission. Ours is considered to be the more ambitious and high-risk of the two. The results will be evaluated and the plan is to produce the real thing for a launch in approximately 1998. -------------------------------------------------------------------------- | Marc Ringuette | mnr@cs.cmu.edu | "He slimed me!" | | CMU Computer Science | 412-268-3728(w) | [watch this space for other | | Pittsburgh, PA 15213 | 412-681-5408(h) | quotes from great literature] | -------------------------------------------------------------------------- ------------------------------ End of Space-tech Digest #11 *******************