The Perseverance Mars rover has demonstrated a new onboard global localization capability that has the potential to significantly improve long-distance autonomous navigation [1]. It allows the rover to precisely determine its position by comparing panoramic images from its navigation cameras with high-resolution orbital maps, and use that knowledge to reset its onboard position uncertainty model back to nearly zero meters. Leveraging a fast, commercial off-the-shelf (COTS) Snapdragon 801 processor from the Ingenuity Helicopter Base Station contained within the rover body, the system performs complex image-processing tasks to accurately determine the rover's position relative to orbital maps. This approach of diverting compute-intensive operations to a faster but nonradiation hardened CPU overcomes the limitations of the rover's 20-times slower, radiation-hardened primary processor, and is a paradigm shift in planetary robotics. We focus on the characterization, implementation and mitigation strategies required to use a non-radiation-hardened COTS processor in the harsh space environment. We describe the software-based memory checking and fault isolation techniques developed to address memory corruption potentially caused by aging or radiation effects. Furthermore, we present the performance and operational results from the deployed system on Mars, including its accuracy and robustness. The paper also discusses lessons learned during the development and flight deployment process and outline how this technology will fundamentally change future planetary navigation to support more ambitious and efficient autonomous missions by enabling continued use of commercial co-processors even in the presence of degraded memory capabilities.

The Perseverance Mars rover needs to drive long distances between regions of scientific interest to collect a diverse set of samples. Position knowledge is needed for navigating to the region of interest. Planetary mobile robots accumulate position uncertainty as they move. Globally localizing the robot to an orbital map of Mars removes this uncertainty. To date, this has been performed manually on the ground by humans for mobile surface and aerial robots. This can be accurate but requires communication between planets. This takes significant time and the need for it limits how far Perseverance can autonomously navigate without ground-in-the-loop.This paper describes a new onboard approach for performing global localization, much of which already has been successfully demonstrated on Perseverance. Our Censible technology uses a modified census transform to achieve sub-meter global localization accuracy that is robust and practical, and whose performance matches human-directed localizations from the first two and a half years of the mission to within 0.5 meters on average with no outliers. We use the fast processor on the Ingenuity Helicopter Base Station mounted in the Perseverance rover to perform the localization. It was originally installed to coordinate communication with Ingenuity. This effort developed the interfaces and radiation mitigation methods needed to enable its use as a rover co-processor. The system is designed to limit operations impact and requires no daily input from rover operators other than whether or not to perform global localization, but also allows strategic configuration options if desired. We discuss the lessons learned from developing and deploying this new technology on a flight mission, and describe how global localization is expected to increase science return and change how planetary mobile robots navigate.

The Mars 2020 (M2020) Perseverance Rover is NASA's most advanced planetary rover mission to date. It includes a novel Sample Caching Subsystem (SCS) which will collect rock cores for possible future return to Earth, as well as an improved mobility system with enhanced autonomous navigation which will enable it to traverse faster and farther than prior rovers. The development of both systems required extensive flight software and flight hardware testing. To support this testing, we developed the Surface System Development Environment (SSDEV) and used it for a wide variety of testing. SSDEV is a bundled subset of M2020 Flight Software which runs on commercially available Linux computers and can be combined with multiple backend options for simulation and hardware control. The SSDEV architecture enabled our teams to perform much more testing of flight software and flight hardware than would have otherwise been possible. As a secondary benefit, the SSDEV-based test campaigns also helped our teams enter the operations phase of the mission with greater readiness of operations products and tools. In this paper, we summarize the motivation for SSDEV, provide an overview of the SSDEV architecture, list several examples of how SSDEV was used, and summarize lessons learned. SSDEV is not a substitute for integrated testing with flight-like avionics, but it enabled substantially more testing than would have otherwise been possible and also provided some unique benefits. We recommend architectures like SSDEV to future projects that need to perform extensive hardware and software testing using a limited set of flight-like avionics.

NASA's Mars Science Laboratory (MSL) mission landed the Curiosity rover on Mars on August 6, 2012. As of August 6, 2019 (sol 2488), Curiosity has driven 21,318.5 meters over a variety of terrain types and slopes, employing multiple drive modes with varying amounts of onboard autonomy. Curiosity's drive distances each sol have ranged from its shortest drive of 2.6 centimeters to its longest drive of 142.5 meters, with an average drive distance of 28.9 meters. Real-time human intervention during Curiosity drives on Mars is not possible due to the latency in uplinking commands and downlinking telemetry, so the operations team relies on the rover's flight software to prevent an unsafe state during driving. Over the first seven years of the mission, Curiosity has attempted 738 drives. While 622 drives have completed successfully, 116 drives were prevented or stopped early by the rover's fault protection software. The primary risks to mobility success have been wheel wear, wheel entrapment, progressive wheel sinkage (which can lead to rover embedding), and terrain interactions or hardware or cabling failures that result in an inability to command one or more steer or drive actuators. In this paper, we describe mobility trends over the first 21.3km of the mission, operational aspects of the mobility fault protection, and risk mitigation strategies that will support continued mobility success for the remainder of the mission.

The Mars Science Laboratory (MSL) rover, Curiosity, completed its second Martian year, 1337 sols (1374 Earth days), of operation on May 11, 2016, and its third Martian year, 2006 sols (2061 Earth days), of operation on March 28, 2018. During this time the rover successfully drilled twelve full depth drill holes into the Martian surface and analyzed the recovered material using onboard instruments, giving us new insights into the potential habitability and geologic diversity of ancient Mars. During the second Martian year, four holes were drilled into the mudstones of the Murray formation: 'Confidence Hills' (Sol 759), 'Mojave 2' (Sol 882), 'Telegraph Peak' (908) & 'Buckskin' (Sol 1060); while four more holes were drilled into the sandstones of the Stimson formation: 'Big Sky' (Sol 1119), 'Greenhorn' (Sol 1137), 'Lubango' (Sol 1320) & 'Okoruso' (Sol 1332). During the third Martian year, four additional holes were drilled into the Murray formation: 'Oudam' (Sol 1361), 'Marimba' (Sol 1422), 'Quela' (Sol 1464) & 'Sebina' (Sol 1495). In this paper, we will give a brief overview of the rover sampling hardware and nominal drilling protocols, followed by a discussion of how these protocols were refined and altered early during the course of Curiosity's second year on Mars. In addition, we will describe the 'Bonanza King' (Sol 724) drill campaign, the mission's first 'successful failure', and how it influenced these changes. We will also briefly discuss the events leading up to the drill feed fault on Sol 1536, which resulted in suspension of all drill activities for the remainder of the third Martian year. Finally, we will present scientific highlights obtained from each drill site utilizing MSL's onboard instrumentation (SAM & CheMin), results enabled by the drill's ability to excavate sample at depth and transfer it to these instruments.

NASA's Mars Science Laboratory Curiosity rover landed in August 2012 and began experiencing higher rates of wheel damage beginning in October 2013. While the wheels were designed to accumulate considerable damage, the unexpected damage rate raised concerns regarding wheel lifetime. In response, the Jet Propulsion Laboratory developed and deployed mobility flight software on Curiosity that reduces the forces on the wheels. The new algorithm adapts each wheel's speed to fit the terrain topography in real time, by leveraging the rover's measured attitude rates and rocker/bogie suspension angles and rates. Together with a rigid-body kinematics model, it estimates the real-time wheel-terrain contact angles and commands idealized, no-slip wheel angular rates. In addition, free-floating "wheelies" are detected and autonomously corrected. Ground test data indicate that the forces on the wheels are reduced by 19% for leading wheels and 11% for middle leading wheels. On the ground, the required data volume increased by up to 129%, and drive duration increased by up to 25%. In flight, data collected over 3.6 km and 149 drives confirmed a reduction in wheel current, correlated with wheel torque, of 18.7%. The new algorithm proved to use fewer resources in flight than ground estimates suggested, as only a 10% increase in drive duration and double the drive data volume were experienced. These data indicate the promise of the new algorithm to extend the life of the wheels for the Curiosity rover. This paper describes the algorithm, its ground testing campaign and associated challenges, and its validation, implementation, and performance in flight.

NASA Mars Rover operations are planned anew every day, in contrast to traditional deep space missions where operations can be planned weeks or months in advance. Verification of the rovers' spacecraft commands must be performed quickly, implemented robustly, and must consider multiple possible outcomes. In this paper we discuss the architecture behind one component of our system, a software tool named RP-check that evaluates command sequences created by the Rover Planner (RP) team. RP-check is one component of the Rover Sequencing and Visualization Program (RSVP), our primary tool for commanding and validating the mobility, arm, and turret subsystems on the rovers. RP-check is the spacecraft sequencing equivalent of programming language analyzers like lint [6]. It automates the analysis of sequences of commands, evaluating their conformance to best practices described by mission-spanning Flight Rules and Rover Planner team-specific recommendations. It checks hundreds of rules and typically completes its assessment in less than 30 seconds, enabling Rover Planners to continuously validate and refine their command sequences many times a day while planning activities. RP-check owes its success to a simple framework that enables it run quickly and makes it easy to add new rules. Built and maintained by members of the Rover Planner team, new rules and test cases can be validated and deployed quickly to ensure that any issues of concern that arise during operations will be addressed in all future plans.

The Mars Science Laboratory (MSL) Curiosity rover experienced increasing wheel damage beginning in October 2013. While the wheels were designed to operate with considerable damage, the rate at which damage was occurring was unexpected and raised concerns regarding wheel lifetime. The Jet Propulsion Laboratory (JPL) has now developed and deployed new software on Curiosity that reduces the forces acting on the wheels. Our new Traction Control algorithm adapts each wheel's speed to fit the terrain it drives over. It does not rely on any a priori knowledge of the terrain, and instead leverages the rover's measured attitude rates and suspension angles, together with a rigid-body kinematics model, to estimate the real-time wheel-terrain contact angles and ideal, no-slip wheel angular rates. In addition, free-floating “wheelies” are detected and autonomously corrected. In this paper, we describe the algorithm, its ground testing campaign and associated challenges, and finally its validation and performance in flight. Ground test data demonstrates reductions in the forces acting on the wheels and validates the wheelie-damping capability. Secondary benefits in some terrains include a reduction in heading deviations while climbing rocks, with a reduction in slip in certain sandy terrains. Preliminary validation from flight data confirms these findings.

The development of the Sampling and Caching Subsystem (SCS) of the Mars 2020 Rover Mission is highly dependent on testing of prototype hardware and software operating in explicit conditions as part of integrated testbeds. To achieve relevant integration of hardware and software while maintaining rapid algorithm development capabilities and high testing throughput, the Controls and Autonomy for Sample Acquisition and Handling (CASAH) software system was developed. CASAH is an implementation of the Intelligent Robotics System Architecture (IRSA), which mimics JPL Flight Software (FSW) in that it is divided into modules that run separate processes that communicate via message passing, each module is assigned an owner that is a single developer, and the operator initiates requests via a text-based interface that interprets sequences of commands. IRSA enables a modular breakdown of CASAH that follows that of 2020 Flight Software, so developers can take an algorithm from a module in CASAH and re-code it into the same module in FSW. As deployment of CASAH has grown to ten testbeds - each with different hardware and objectives - bottom-up design decisions have been intentionally made to keep the system lightweight and maintainable by a very small team. To date, CASAH has been used to run 1393 different tests. This work describes CASAH, the testbeds and functionality it supports, the tools used to manage the development and sharing of code, and the features of the software. Lessons learned over the past three years of development and deployment are provided.

This paper describes the flyaway guidance, navigation, and control system design that was successfully used during the landing of NASA's Mars Science Laboratory on Mars on 5 August 2012. One of the most challenging phases of the Mars Science Laboratory mission is the entry, descent, and landing that starts with the Mars atmospheric entry and ends with the rover Curiosity landing on Mars and the descent stage with the sky crane crashing to the surface. Because the two-way signal's communication time between Earth and Mars is about 26 min and the fact that the end-to-end entry, descent, and landing takes about 7 min, the lander must be controlled completely autonomously. At the last step of the landing, the rover is lowered to the surface by using a new sky crane, which resides on the descent stage of the lander. The sky crane slowly descends toward the Martian surface as the rover is lowered. Once the touchdown is detected, the onboard control system flies the descent stage as far away as possible from the rover. This phase of the landing is called the "flyaway." The flyaway is the last phase, and its main goals are to move the sky crane, which may contain unused propellant, to a safe distance and to avoid plume impingement on the sensitive science instruments on the rover during the flyaway maneuver. The coupled attitude and translational motion is planned, that is, rotational and translational trajectories are time profiled by the onboard guidance algorithm. The trajectories obey the control authority constraints, due to maximum torque and force, while maximizing the divert distance. These guidance trajectories are tracked by using feedback control laws that provide robustness to any model uncertainties and disturbances forces and torques. Simulation results predicted that a desirable performance would be achieved in terms of the flyaway distance and the plume avoidance. Finally, the actual landing on Mars on 5 August 2012 was extremely successful, and the results were consistent with predictions.

It is anticipated that the Mars Science Laboratory rover, named Curiosity, will traverse 10–20 km on the surface of Mars during its primary mission. In preparation for this traverse, Earth‐based tests were performed using Mars weight vehicles. These vehicles were driven over Mars analog bedrock, cohesive soil, and cohesionless sand at various slopes. Vehicle slip was characterized on each of these terrains versus slope for direct upslope driving. Results show that slopes up to 22 degrees are traversable on smooth bedrock and that slopes up to 28 degrees are traversable on some cohesive soils. In cohesionless sand, results show a sharp transition between moderate slip on 10 degree slopes and vehicle embedding at 17 degrees. For cohesionless sand, data are also presented showing the relationship between vehicle slip and wheel sinkage. Side by side testing of the Mars Exploration Rover test vehicle and the Mars Science Laboratory test vehicle show how increased wheel diameter leads to better slope climbing ability in sand for vehicles with nearly identical ground pressure. Lastly, preliminary data from Curiosity's initial driving on Mars are presented and compared to the Earth‐based testing, showing good agreement for the driving done during the first 250 Martian days.

The design of a surface robotic system typically involves a trade between the traverse speed of a wheeled rover and the terrain-negotiating capabilities of a multi-legged walker The ATHLETE mobility system, with both articulated limbs and wheels, is uniquely capable of both driving and walking and has the flexibility to employ additional hybrid mobility modes. This paper introduces the Sliding Gait, an intermediate mobility algorithm faster than walking with better terrain-handling capabilities than wheeled mobility.

The All-Terrain Hex-Limbed Extra-Terrestrial Explorer (ATHLETE) is a modular, heavy-lift vehicle being developed to support NASA operations on the lunar surface. This agile system consists of a symmetrical arrangement of six limbs, each with six articulated degrees of freedom and a powered wheel. The design enables transport of bulky payloads over a wide range of terrains and is envisioned as a tool to mobilize habitats, power generation equipment, and other supplies in for long-range lunar exploration and lunar outpost construction. The first-generation prototype transports payloads of up to 300 kg in terrestrial testing, with flight models projected to carry at least 15 metric tons in a lunar gravity environment.

This paper presents the development, validation, and deployment of the visual target tracking capability onto the Mars Exploration Rover (MER) mission. Visual target tracking enables targeted driving, in which the rover approaches a designated target in a closed visual feedback loop, increasing the target position accuracy by an order of magnitude and resulting in fewer ground-in-the-loop cycles. As a result of an extensive validation, we developed a reliable normalized cross-correlation visual tracker. To enable tracking with the limited computational resources of a planetary rover, the tracker uses the vehicle motion estimation to scale and roll the template image, compensating for large image changes between rover steps. The validation showed that a designated target can be reliably tracked within several pixels or a few centimeters of accuracy over a 10-m traverse using a rover step size of 10% of the target distance in any direction. It also showed that the target is not required to have conspicuous features and can be selected anywhere on natural rock surfaces excluding rock boundary and shadowed regions. The tracker was successfully executed on the Opportunity rover near Victoria Crater on four distinct runs, including a single-sol instrument placement. We present the flight experiment data of the tracking performance and execution time. © 2009 Wiley Periodicals, Inc.

We describe a new capability to automatically detect dust devils and clouds in imagery onboard rovers, enabling downlink of just the images with the targets or only portions of the images containing the targets. Previously, the MER rovers conducted campaigns to image dust devils and clouds by commanding a set of images be collected at fixed times and downloading the entire image set. By increasing the efficiency of the campaigns, more campaigns can be executed. Software for these new capabilities was developed, tested, integrated, uploaded, and operationally checked out on both rovers as part of the R9.2 software upgrade. In April 2007 on Sol 1147 a dust devil was automatically detected onboard the Spirit rover for the first time. We discuss the operational usage of the capability and present initial dust devil results showing how this preliminary application has demonstrated the feasibility and potential benefits of the approach.

Visual target tracking (VTT) software has been incorporated into Release 9.2 of the Mars Exploration Rover (MER) flight software, now running aboard the rovers Spirit and Opportunity. In the VTT operation (see figure), the rover is driven in short steps between stops and, at each stop, still images are acquired by actively aimed navigation cameras (navcams) on a mast on the rover (see artistic rendition). The VTT software processes the digitized navcam images so as to track a target reliably and to make it possible to approach the target accurately to within a few centimeters over a 10-m traverse.