1. NASA Ames Research Center
Executive(s)
Life in the Atacama Design Review December 19, 2003
Nicola Muscettola
Vijayakumar Baskaran
NASA Ames Research Center
NASA Ames Research Center

2. Description and Motivation
Model-Based Executive(s) for LITA
Rover Executive
Instrument Manager
Modules were presented in September 2003
NASA Ames Research Center

3. Rover Architecture using IDEA
Operator
Interface
Offboard
Telemetry
Router
Mission-level agent
Mission
Planner
Science
Planner
Waypoints
IDEA
Mission Executive
Mission
Executive
Status
Science
Executive
System-level agents
Waypoint
Curvature,
Speed & Time
Curvature,
Speed & Time
Navigator
Health
Monitor
Onboard
Stop
Terrain Evaluation
Odometry
Stereo
Mapper
State
Estimator
Vehicle
Controller
Status
Status
Sensors
Images
NASA Ames Research Center

4. Description and Motivation
Model-Based Executive(s) for LITA
Rover Executive
Instrument Manager
Modules were presented in September 2003
Motivating System Requirements: 3.1, 4.4, (but what does that mean really?), 5.5, 6.0
NASA Ames Research Center

5. NASA Ames Research Center
Key Requirements
Rover Executive: provide goal-driven commanding for high-level traverse and science operations
“Go to waypoint” and “Perform Detailed Investigations”
Instrument Manager: allows single goal commanding
“Do mosaic” instead of sequence of individual snapshots
Executive invokes planner(s) if necessary to optimize course of action
Executive translates between abstract goals and low-level commands
Executive issues commands and responses as a consequence of internal (synchronous) or external (asynchronous) events
Executive must be robust to wide range of faults (to be defined)
Executive must guarantee response within required time limits
NASA Ames Research Center

6. Software Organization
Science
Interface
Rover
Interface
Telemetry
Goals
Science Planner
Mission
Planner
Health
Monitor
Telemetry
Manager
Measurements
Plans
Plans
Faults
State
State
Plans
Measurements
Rover
Executive
State Observer
State (All)
Positions
Waypoints
Stop
Actions
Instrument
Manager
Position
Estimator
Mapper
Navigator
Odometry
Curvature & Speed
Far-field Evaluator
Near-field Detector
Vehicle
Controller
Instrument
Controllers
Images
Measurements
Data
Images
Proprioception & Sun Image
Commands
Commands
NASA Ames Research Center

7. Design Considerations
Informing Design Principle:
The logic of execution and recovery shall be fully inspectable
A model-based approach combines a declarative model, a general-purpose plan synthesis engine and heuristics
A model-based approach is expected to allow quick incremental encoding of new recoveries from failure conditions as they are discovered
Metrics:
Coverage of failure conditions
Reaction time (as a function of model complexity)
The design is influenced from Remote Agent and Gromit rover experience
Reimplementation of the DS1 Executive as a model-based system vs the old procedural Executive in ESL
Gromit (ATRV Junior) rover executive reimplement PRS-based executive from LAAS/CNRS
NASA Ames Research Center

8. NASA Ames Research Center
Technical Approach
Multiple executives (Rover Executive and Instrument Manager) vs single executive
Coupling between rover components
Reaction time
Currently focusing on single executive (consistent with Dave’s architecture concept)
Modeled constraints can be switched on or off to support direct teleoperation modes (monitored mode of requirements 6.0)
NASA Ames Research Center

9. Implementation Issues
Prototype tests
Simulator
Mission Simulator Facility? Not clear if it is usable for real-time simulation
Rover
Multi-day scaled-down autonomous traverse with Gromit
Expected timeframe: April 2004
Risks
Complexity of incremental modeling
Mitigations: full engagement of trained Ames personnel, training, better modeling tools/languages
Reaction latency
Mitigations: faster planning framework (new EUROPA), use of procedural task expansion (TDL)
Issues
Relation of executive to health monitoring and telemetry server. External modules or internal modules?
What happens to executive during hybernation? Current design assumes it is always “active”, i.e., capable of reacting. Is this reasonable?
Schedule issues: need to firm up ASAP instrument models, message/command dictionary, mission operations concept.
NASA Ames Research Center