The personal satellite assistant: an internal spacecraft autonomous mobile monitor

This paper presents an overview of the research and development effort at the NASA Ames Research Center to create an internal spacecraft autonomous mobile monitor capable of performing intra-vehicular sensing activities by autonomously navigating onboard the International Space Station. We describe the capabilities, mission roles, rationale, high-level functional requirements, and design challenges for an autonomous mobile monitor. The rapid prototyping design methodology used, in which five prototypes of increasing fidelity are designed, is described as well as the status of these prototypes, of which two are operational and being tested, and one is actively being designed. The physical test facilities used to perform ground testing are briefly described, including a micro-gravity test facility that permits a prototype to propel itself in 3 dimensions with 6 degrees-of-freedom as if it were in a micro-gravity environment. We also describe an overview of the autonomy framework and its components including the software simulators used in the development process. Sample mission test scenarios are also described. The paper concludes with a discussion of future and related work followed by the summary.

designed. The physical test facilities used to perform ground testing are briefly described, including a micro-gravity test facility that permits a prototype to propel itself in 3 dimensions with 6 degrees-of-freedom as if it were in an micro-gravity environment. We also describe an overview of the autonomy framework and its components including the software simulators used in the development process.
Sample mission test scenarios are also described. The paper concludes with a discussion of future and related work followed by the summary. and also can be commanded by simple speech commands and human motions.

Mission Roles
The two primary mission roles of the PSA is being designed to address are to improve spacecraft crew productivity and to decrease mission risk by serving as part of an integrated spacecraft systems health management system.

PSA RATIONALE
The impetus for the PSA project is to develop an intelligent system that increases crew productivity and reduces risk for manned missions. It was originally an element of the NASA

Cross-Enterprise
Technology Development program and has subsequently been adopted by the Intelligent Systems project in the NASA Computing, Information, and Technology program, which primarily focuses on enhancing its autonomy, and the NASA Engineering Complex Systems (ECS) program, which is responsible for the overall development effort to raise its technological readiness level.
The ECS program was formulated by NASA to address --grooving concems._about-the. agency-'-s abilit-y -to develop, operate, and maintain the complex systems required to meet our current and future mission objectives.
During the program formulation phase the ECS program identified four common problem classes associated in most NASA systems: After the docking, the PSA returns to Mary and supports her through the rest of the day. After Mary and the rest of the human crew goes to sleep, the PSA recharges in preparation for its assigned patrol duty. It's given several priority areas to monitor where recent mishaps have occurred; a fire and a pressure leak are high on the ground crew's list to monitor.
As the PSA moves out on its evening patrol, it passes by another PSA system at the sensor location that it had identified earlier as having a calibration error. The sensor in question has since failed. The other PSA is now providing substitute sensing until a new sensor is installed.

REQUIREMENTS
In order to develop a system that can meet the goals previously described, we defined a number of requirements.
. biological creatures. Our current approach is to develop a system that can do self-localization using a combination of stereo-cameras to build depth maps and sense motion by means of optic flow algorithms and fuse this with data from a 6-DOF inertial measurement unit (accelerometers), and proximity sensors. As necessary, we can mitigate risk by engineering the environment as needed. ......

Requirement 3--"station-keep"
on command by maintaining a fixed position and orientation relative to its environment.
Note that the environment, i.e., the ISS, is continually in motion as it orbits the Earth and performs minor attitude adjustments.
Although the device is useful if it is held by a crewmember or fixed to a surface, the more that can be done with the device that doesn't require crew time, the better.

PSA PROTOTYPES AND TEST FACILITIES
,,e PSA project is using an iterative, rapid prototyping development approach. We started by envisioning the end product in our minds. In 1998, the project's first year, a concept modeI, shown in Figure 1, was developed.
The project's primary loci are mitigating risk on issues 1 -3. Issues 4 -6 are also being worked but at a lower level. In addition, si_maificant effort was invested in enNneering prototypes, simulators, and test facilities in order to validate that we have captured and addressed the salient issues.
For example, spacecraft technoIogy had basically solved the propulsion and attitude control problem by means of coldgas thrusters and reaction wheels. Although reaction wheels that met our requirements were not commercially available, we were convinced they could readily be developed. Safety concerns of having and refueling a high-pressure tank that would enable extended operation soon caused us to rule out cold-gas as a primary means of propulsion. However, it does remain as a option for a special-purpose PSA, such as one for exploring depressurized regions of a spacecraft. Instead, we began examining fans and blowers for propulsion. With this model, we souaht to answer the question, "if we could build it, would we want to?" The effort spent in developing a handheld mockup was well worthwhile. Not only did team members find having a physical concept model useful to convey ideas, we found people outside the project gTasped what we were doing much more readily when we used the model in our presentations.

Technology Challenges
Having fleshed out the concept, we began a critical analysis of the problems and risks as well as the needed technologies currently or imminently available, and those that needed to be developed. The tall-pole issues we identified were: which is shown juxtaposed to the concept model in Figure 2.
In the Model 1 design, no attempt was made to conform to the packaging eventually required. By equipping the Model 1 with a stereo-vision systems, we are able to perform vision-based localization and visual servoing. My means of a wireless Ethernet, a high-level model-based autonomous control system located on a remote server commands the ModeI 1 to execute various missions.
algorithms thatuseoneto fourstereo-pair cameras• The Model 2 wasdeveloped to support up to four stereo pair cameras, in addition, a 12" sphere was created directly from C_d) drawings using a stereo-lithography process.
Propulsion and attitude control 6-DOF (X, Y, Z, yaw, pitch, roll') were achieved using 6 fan pairs located in 6 ducts• It would have been preferable to use reaction wheels for several reasons including they would provide tighter control, quieter operation, and _eater energy efficiency. However, reaction wheels that met our specifications would have to be custom built so they were scheduled to be implemented as part of the Model 3 prototype. truth. We use this external localization capability to measure the accuracy of the onboard position estimation system. In addition, for unit test purposes we can configure the onboard control system to use the externally-calculated position data stream as feedback to achieve its commanded position and velocity, trajectories.

PSA Model 2
While the Model 1 was being tested, development of a 6-DOF Model 2 and an accompanying 6-DOF micro-gravi b' test facility was underway.
We were well aware that the position and velocity estimation problem in a 3-DOF system was much simpler than in a 6-DOF system. The existing algorithms we examined that work fine in a 3-DOF system did not work at all in a 6  The Model 2 became operational in 2001 and currently is being tested on a pitch & yaw _mbal shown in Figure 3. It is 12" in diameter to enable the use of commercially available components as much a possible and to give space to make changes after it was constructed.
For example, the primary core is comprised of a PC104 stack of boards. The gimbal is mounted on a pressurized plate, which permits it to float on the Moreover. there are operational advantages to quic_y being able to replace a batter)' rather than waiting for it to recharge. ,another design consideration is that batteries that can be charged and discharged quickly tend to not hold as much energy as ones that don't. Using the moment,cn'p_ wheels, we will have the ability to use excess momentum to recharge batteries, acting somewhat like a self-winding watch, but this will be limited by the maximum recharge rate of the batteries. Moreover, the critical propulsion measure is how quickly can the PSA stop, which is a function of its mass, velocity, and the maximum battery discharge rate. Initially. a hybrid batter), approach seems to hold the most promise.

PSA Model 5
The

AUTONOMY FRAMEWORK AND S/3,!ULATORS
An autonomy framework designed to address the previously discussed operational requirements has been developed and is depicted in Figure 5.

Communication
Manager--responsible for managing message traffic and executing certain message handIers. Serves same role in both off-board systems.

Off-board Autonomy System
The off board autonomy system is responsible for high level  Figure 5 -PSA Autonomy Framework sensor data provided by the onboard control system, e.g., for diagnosis, and for plan repair, e.g., onboard control system is unable to achieve a waypoint. Architecturally, this system could be integrated onboard the PSA. It is off-board to The declarative planner is called to initialize the plan database and also is called during plan execution as specified by the plan being executed. It is typically caIied to plan for a period of significant duration sufficiently in the future such that the deliberative planner will complete prior to the start time of this period, but not so far in the future that the initial state at the future start horizon is not known with high confidence.

Reactive
Planner--responsible for insuring that the Plan Database is in a state such that the tasks to be executed at a its arguments defined by the token, updating the plan database with the token return values when the procedure ten-ninates, constraining the plan database so that planners only have !imited abi!ity to change the past, and cal!ing the Reactive Planner. as described above, as needed to update the ptan database. The plan runner implemented is described in more depth in [4].

Goal Dialogue
Manager--acts as an arbiter between the autonomous control system and other agents, including people. It retains state regarding its interaction with the other agents, e.g., recalls the subject of a previous sentence spoken by the user. As an arbiter, this element serves two roles: a goal manager and a dialogue manager.
The goal manager essentially acts as a meta-planner for the declarative planner. As stated above, the declarative planner requires a start and end horizon time bounds, an initial state of the timelines at the start time, and a set of goals. The goal manager interacts with the user to determine this information. with simulated crewmembers, payloads, and objects. In Figure 6, a PSA is shown with a crewmember in the [SS U.S. Lab "Destiny" module.  Description--PSA will create an enviromment map of the !SS module by traversing the space in a serpentine path recording the environment sensor readings along the way. During this activity, its path will be blocked by static obstacles (some of which are known of ahead of time) and moving obstacles. At one point the PSA will be interrupted to be teleoperated and then perform a station-keeping task at a location specified by an ISS Rack Locker name, after which it will complete its original environment-mapping task.

Purpose-
. Demonstrate navigation to several waypoints in an environment that has static and dynamic obstacles.
• Demonstrate mixed-initiative execution including autonomous task interruption and resumption, guarded teleoperation, and visual servoing by command.
• Demonstrate generation of a near-optimal 6-DOF route plans

Description--
A fixed sensor in the ISS module sig-nals a high temperature to the Environmental Control Life Support System (ECLSS). However, it is not known whether the sensor is defective or the source or the heat. PSA is given a command by ECLSS to go the fixed sensor location and verifv the temperature at that location. If PSA confirms the fixed sensor is correct, PSA is to locate the heat source and signal the source to ECLSS, will then power down the locker at that location. Once PSA verifies that the temperature has returned to normaI, it returns to its docking bay. If the fixed sensor is not correct, PSA is to stay at that location until the fixed sensor is made operational.
Once PSA verifies the sensor, PSA returns to its docking bay. and is requested to generate a near-optima/ plan to achieve the goals. The goals will be such that it will be necessary to schedule multiple battery recharges in order to achieve them. The operator will dynamically' change the plan prior to its execution. During the execution, PSA will monitor the location of inventory items it senses as it passes by. PSA witI encounter static and dynamic obstacles in the environment.
Due to an inaccurate battery" model, PSA wi]I have to replan to prevent running out of power prior to r_eharo, ino _t the, doc.Vdn_ h._,, Once vq a has completed the goal list, it given a list of inventory items to locate, some of which it passed by. PSA responds with the locations of the items it senses and then generates a plan to explore the areas of the ISS module it did not previously explore in order to locate the other items. designed. We also briefly, discussed the micro-gavity test facility, which allows us to fly the PSA prototypes on the _ound as if they were onboard the ISS. The autonomy framework for intelligent flight vehicle control being developed and tested as part of this project was also presented.
Several sample missions being used to test the proto.types and the autonomous control system were also outlined. We concluded with a discussion of both the shortterm and long-term future work in the area of autonomous mobile vehicles for in-flight spacecraft support.