1. INTRODUCTION

Human and robotic exploration of space are often presented as mutually exclusive alternatives – why should we spend a huge amount of money to send humans to Mars when the six-years-plus missions of the Mars Exploration Rovers (MERs) Spirit and Opportunity have made so many important discoveries at a tiny fraction of the cost? In fact, the time when we finally send humans to Mars, presumably a few decades from now, will not mark the end of the involvement of unmanned systems in the exploration of Mars. Instead, robots and other unmanned systems will continue to play many critical roles on Mars, and the presence of humans will strongly affect the characteristics of the robotic systems we build. In advance of the first human landings, the descendants of Spirit and Opportunity will survey candidate landing sites, locate ice and mineral resources, establish power, communications, and navigation infrastructure, and construct underground habitats. Once humans have landed, mobile robots will continue to explore and to preview sites for human exploration, identifying targets of interest and possible hazards. They will also perform ongoing construction tasks, and transport equipment, supplies, and people. This paper discusses some of the major roles that unmanned systems will play in the context of the evolution from unmanned exploration sorties through base site selection and construction to base operations, describes some of the capabilities these robots will require, considers the technology context for the development of these systems, and outlines some of the major challenges to their realization.

2. ROLES FOR ROBOTS: EXPLORATION, TRANSPORTATION, AND WORK

We can distinguish several archetypal roles for robots on Mars: exploration, transportation, and physical work, on the basis of the capabilities they will require in terms of size, strength, power, mobility, navigation, manipulation, and command and control modalities.

The role of an exploration robot is to travel across the surface of Mars while carrying a payload of instruments to sense the features of its environment. Its size and strength is dictated by the size and mass of its payload. It must have enough power to support the operation of its payload instruments, but beyond this the amount of power available may dictate the pace of operations, based on power requirements for mobility and communications. Its mobility capabilities, and therefore some aspects of its size and shape, are dictated by the nature of the terrain it is intended to traverse. A rover intended to travel overland need only be able to avoid damage or entrapment, but eventually we will need exploration robots with "exotic" mobility capabilities to explore much more severe (but highly interesting) terrain such as cliff faces (Huntsberger et al. 2007). The route of an exploration robot typically carries it to places that have not been visited before, although mission planners can and do make use of overhead imagery. The mission plan is repeatedly and incrementally modified in response to what is encountered, and manipulation requirements are generally minimal, reflecting payload needs. Both before and after the arrival of humans on Mars, exploration robots will preview areas of interest before humans visit to reduce risk by identifying hazards, to identify areas of highest interest, and to maximize the efficient use of human attention. All the mobile robots that have been deployed on Mars to date have been light-weight solar-powered exploration "rovers."

Once we have determined the site of the first Martian base, robots will provide logistical support through the transportation of supplies (fuel, oxygen, water, food, etc) and equipment from one place to another within this general area of operations. Robot size and strength will be dictated by the specifics of what it is to be transported, and general-purpose tractors may be used to tow specialized trailers for cargo, liquids, and so forth. Making repeated transits between specific points within the base area of operations will greatly ease the challenges of mobility and navigation, since we can create and follow improved "roads" and leverage navigation infrastructure such as beacons. Manipulation capabilities will be required only for loading and unloading cargo or transferring fuel. Eventually we will need "optionally manned" vehicles to transport humans as drivers or passengers. This suggests that we will need some unmanned ground vehicles with size at least comparable to an All Terrain Vehicle (ATV) or golf cart. The dune buggy sized Lunar Rover Vehicles (LRVs) successfully used on Apollo 15, 16, and 17 provide a good reference point.

Later, robots will be employed to perform physical work in support of the construction of the Mars base: site preparation, road clearing, drilling, excavation (NASA, 2009), manufacture of bricks and/or other materials, construction of structures, and assembly and installation of equipment. These robots will have to be strong, they will require much more power than basic exploration or transportation robots, and they will need the mobility to move about in the "construction zone." For excavation and similar heavy construction tasks a back-of-the-envelope calculation suggests that the obvious terrestrial models – a small Bobcat or forklift – might be overkill, since the excavation of 10x10x10 meters in 800 days (26 months) using 5 robots would require each unit to move only about 1/4 m3 of regolith per day. Well-defined heavy tasks that do not require precision (such as excavation) will be performed autonomously by teams of robotic vehicles working pretty much continuously, day and night. Sophisticated dexterous manipulation capabilities will be required to autonomously perform more precise tasks ranging from structure construction to equipment installation and hookup.

Work robots will require much more power than rovers, and chemical fuels provide an obvious power source. Methane (or other hydrocarbons) and oxygen produced from atmospheric CO₂ and water mined from ground ice could be used in internal combustion engines or possibly in fuel cells. (Will the water vapor in the engine exhaust create a local snowfall?) Vehicle refueling will of course be necessary, and will likely be most efficient when tankers transfer fuel to the work robots at the jobsite. In some cases, a single power source might provide power via electrical tethers to multiple work robots.

Exploration, transportation, and work represent archetypal roles, both in the different goals each addresses and the capabilities each requires in terms of size, strength, power, mobility, navigation, manipulation, and command and control modalities. Any specific robotic system, of course, may perform tasks encompassing aspects of more than one of these roles.

2.1 Vision of the long-term scenario The sequence of unmanned exploration missions will continue, involving ever larger ground robots carrying more and better instruments and more able to take intrusive samples, such as by drilling. Samples may be collected and transported to one place for the long-delayed sample return mission. At some point, interest will begin to focus on one or two sites for more intensive study – making multiple sorties to a site, where, for example, possible evidence of life may have been found, and the beginnings of infrastructure will be installed there, such as communications relays and navigation beacons. Once the decision is made to send people to a specific site, many tonnes of cargo, including many work, transportation, and exploration robots, will be sent over the course of several launch windows. Teams of heavy work robots will excavate the tunnels for the base, supported by transportation robots bringing fuel and carrying loads of regolith from the hole to the tunnel roof. Later, dexterous work robots will work to seal up the underground base, and to install equipment that has been shipped up. Even if our unmanned systems are unable to bring the base into operation alone, it will be prepared so that the first humans to arrive can do so quickly and easily. At the same time, exploration robots will be continuing their sorties from the base, gaining an ever more detailed understanding of the area, collecting numerous samples, and identifying the places people will want to see in person (Gage, 2010).

3. TECHNOLOGY: CONTEXT AND CHALLENGES

3.1 The context of future IT and robotic technologies While the vision for robotics on Mars sketched in the scenario above may not be feasible today, it is clear that we are not sending people to Mars anytime soon, and so we will have many years of technology advances to exploit. Moore’s Law and its analogs (Gray & Shenoy, 2000) suggest that computer processing, data storage, and communications will improve in performance per cost by 5 or 6 orders of magnitude by 2040, which is (regrettably) not an unreasonable target date. Micro- and nano-scale sensor technologies are also advancing very rapidly. This means that we can depend on qualitative changes as well as quantitative ones in how Information Technology (IT) and related technologies will be used. One obvious development to expect is the continuing addition of new "leaves" to the Internet "tree:" "smart devices" that minimize the need for hands-on human management, radio frequency identification (RFID) tags and other tools for "stuff management," "smart places" with distributed cameras and other sensors covering the human area of operations, and "smart people," with wearable sensors integrated with personal information management (lifelogging) tools (Bell & Gemmell, 2009). One of the classic tropes of space science fiction is that the protagonists of the story are embarked on a mission to learn the fate of an earlier team that has mysteriously vanished (e.g., the film Mission to Mars, 2000). Our Mars base will prevent this by maintaining a network of ubiquitous sensors, feeding into a comprehensive log of sensor inputs, system behaviors, states, and activities, and this data will be reliably cached on Mars as well as relayed to Earth. These sensors will of course be used to support robotic operations, and data from robots will be included in the log (Gage, 2007).

While we can safely rely on the prediction that the IT technologies that underpin "intelligent" robots will continue to evolve rapidly, it is much more difficult to predict the state of the art in 2040 for robotics itself. While robotics still appears on many lists of the top ten/twenty technologies for the next decade/century, it sometimes seems that its future is behind it. An iRobot Roomba® does not live up to our idea of the household robots we were promised, the Army’s Future Combat Systems Program has been cancelled, and DARPA’s projects – with some notable exceptions such as BigDog (Boston Dynamics, 2009) – often treat robots only as an evaluation or demonstration domain. Research in robotics is generally focused on specific components and capabilities; no one is articulating a coherent vision for developing the robots of 2030 or 2040, and certainly no program is pursuing such a goal in a coherent manner. It is sobering to recall that the development of unmanned ground vehicles began in the early 1980s with the Autonomous Land Vehicle program and has continued through a succession of programs (e.g., Demo II and Demo III (NRC, 2002)), and later the DARPA Grand Challenges and Urban Challenge (DARPA, 2007), and yet, after more than 25 years of development and of Moore’s Law improvements in IT technology, we still do not see autonomous vehicles driving on our roads. The greatest challenge to developing autonomous robots is the realization of reliable autonomous "intelligent" behavior – combining elements of artificial intelligence (AI) and human robot interaction (HRI) – and it would be unwise to depend on having a general solution to this problem being available, even in 2040. This paper, therefore, discusses issues associated with autonomous or semi-autonomous behaviors in language referenced to the current state of the art, with modest extrapolation. In addition, of course, we are relying on continuing progress across the broad spectrum of technologies that must be integrated to create capable, robust and, reliable robots, including mobility, manipulation, sensors, actuators, control, communications, and power.

3.2 Robotic operations with humans – and before humans arrive Just as the continuing use of robots will make the lives and work of human Martians easier and safer, so too the presence of humans on the planet will make robotic operations more effective and efficient.

Conversely, it is the operation of unmanned systems before humans arrive on Mars that will be most challenging, before a handy human can be called upon to free a stuck joint by hitting it with a wrench, and while human intervention is limited by the 6 to 44 minutes round trip communications latency between Earth and Mars. It is this last factor that is most critical.

Today’s MER rovers operate on the basis of a single command cycle per day/sol: the ground team sends a command sequence to the rover, the rover executes the sequence and returns images and other data showing the results of the execution, and the team uses this as the basis for planning the next day’s command sequence. Communication is intermittent and limited in bandwidth, and traffic usually passes through an orbital relay. Schemes supporting multiple (potentially many) command cycles per day are not prohibited by the physics, but the cost/benefit tradeoff depends on a number of factors. A solar-powered rover does not have the energy needed either to support the high bandwidth communications required to support a short command cycle or to gain an optimal payoff from rapid command interactions by working 24 hours (and 39 minutes) a day. And exploratory "roving" intentionally exposes the robot to maximum novelty and a correspondingly increased need for operator interaction.

Good situational awareness will be necessary if a remote operator is going to be able to intervene quickly and effectively, so continuous high bandwidth communications from the robot will be required to allow the ground team to "immediately" detect any anomalies that would require intervention. Fortunately, our robots will be able to use the powerful multi-tier communications system that will be implemented to support all aspects of the Mars exploration enterprise, including continuous multi-gigabit per second "big pipe" communications with Earth, perhaps implemented via free-space optical links. This will have to work in the face of Earth and Mars rotation, solar conjunction, and massive Martian dust storms.

Given reliable high-bandwidth communications, even with long latency, rapid remote operator intervention can ensure high system productivity if the frequency of intervention is kept low. If a robot requires 5 interventions per day, and each intervention takes an hour (45 minutes conjunction communications latency plus 15 minutes operator replan time), then the element’s efficiency will be (24.7-5)/24.7, or almost 80%. The same efficiency would result if we had 30 interventions per day with 6 minutes opposition latency and 4 minutes operator time. Clearly, the goal will be to minimize the frequency of required interventions. This means developing task-level behaviors that can operate at a high level of autonomy. We will employ navigation beacons to provide precise robot localization; we will spread fiducial markings liberally across every object’s surface; we will decompose complex system tasks into simpler component subtasks, and implement controllers to replan locally around as many exception conditions (such as getting stuck) as possible. Beyond considerations of efficiency, we will have to implement effective autonomous behaviors to respond immediately to serious contingencies such as spillage of fuel and other material. We will minimize surprises by virtue of performing repetitive tasks (such as excavation) in a well-known area of operations. Even so, efficient robotic operations on Mars before humans arrive will require significant advances in dynamic/adjustable autonomy, including perception-based route learning, replay, and retroplay, and especially in task-sequence-level autonomous manipulation.

Success in this endeavor will result in unmanned systems that will effectively perform the full spectrum of activities required, beginning with detailed site survey and assessment, then autonomous deployment and operation of a nuclear power plant, initiation of in situ resource utilization (ISRU) operations to produce fuel and oxygen from the atmosphere and indigenous ice, and installation of communications/navigation infrastructure to support repetitive operations within a limited "familiar" area with minimum remoteoperator attention. Base construction will require extended excavation, manufacturing (e.g., bricks), assembly, and installation operations. Finally, it may be possible to begin some base-centric operations prior to the arrival of people, including extended exploratory sorties and sample collection and analysis – the commissioning of a Humans-Optional Base.

4. CONCLUSIONS

Robots and other unmanned systems will play many critical roles in all phases of bringing humans to Mars, first continuing in their current role as exploration rovers, then performing base site assessment, selection and preparation, leading in turn to base construction and operations.

These vehicles will differ from current planetary rovers in significant ways. Some of them will be work robots, requiring much more strength and power than exploration rovers, and will be fueled by methane/oxygen engines or fuel cells, requiring autonomous ISRU fuel production, storage, and distribution. Installed communications and navigation infrastructure will enable structured and/or repetitive operations (such as excavation, drilling, or construction) within a "familiar" area with minimum operator intervention. The critical challenges to effectively using robots will occur before humans arrive on Mars, maintaining efficient long-term operations despite the round trip communications latency of 6 to 44 minutes. Later, the presence of humans in the vicinity will permit proactive maintenance and repair and allow teleoperation and extensive operator interaction.

Even after humans arrive, the single most critical resource on Mars will be human attention. Each human we decide to send to Mars will require a huge investment in mass, and therefore in cost. It will be highly cost effective to create systems and procedures to leverage the attentional energy of each human on Mars – to do the most with the fewest people – and that can only be done by using "smart systems", including robots. The question is NOT "robots OR humans on Mars"; instead, the answer is "robots BEFORE humans and robots WITH humans on Mars."