Autonomy: Executive and Instruments

1
Autonomy Executive and Instruments

• Life in the Atacama 2004Science Technology
  Workshop
• Nicola Muscettola
• NASA Ames
• Reid Simmons
• Carnegie Mellon

2
Outline

• Rover Executive (RE) and Instrument Manager (IM)
  functionalities
• Brief technologic background
• Status
• Expectations for 2004 Field campaign

3
Functionality

• Goal provide high-level rover control that is
• on-board
• goal-directed
• context-aware
• robust
• easily reconfigurable
• in other words, the rovers conscience

4
Software functional diagram
Rover Interface
Science Interface
Telemetry
Goals
MissionPlanner
SciencePlanner
HealthMonitor
TelemetryManager
Plans
Faults
State
State
Plans
Rover Executive
StateObserver
Waypoints
Instrument Manager
State(All)
Mapper
Navigator
Measurements
Odometry
Curve Speed
Stop
Positions
Actions
Geom.
Eval.
Vehicle Controller
Far-field Evaluator
Instrument Controllers
Near-field Detector
Position Estimator
Images
Data
Images
Proprioception
Commands
Commands
5
Rover Executive

• Activates planning (both Mission and Science) and
  receives plans to execute
• Keeps track of status of controlled software
  modules
• Executes recovery actions when faults occur
• Anomalous execution condition
• Task timeout or early end
• Software error causing software module crash
• Requests replanning in case of fault/opportunity

6
Instrument Manager

• Implements science observation interface to
  instruments
• SPI Mosaic rather than individual SPI
  commands in science planner
• Implements tightly coordinated multi-instrument
  observations
• Commands instrument calibration from information
  on (multi-)instrument performance
• Implements context-sensitive data cataloging
• Recovers from instrument faults

7
Rover Executive Model-Based Planning
8
Instrument Manager Procedural Execution
Interface Model
Procedures Store
Agent Relay
Active Procedure
9
Advantages of approach

• Unlike traditional flight software, task
  coordination information is explicit
• In model -gt constraints do A while B
• In procedure -gt tasks and links in task
  structure
• Addresses key complexity of complex flight
  software
• Correctness of interaction between parallel
  tasks
• If task coordination is explicit, it can be
  inspected, analyzed and tested with automated
  tools
• Expected result
• more robust software
• decrease in time and cost to produce validated
  software

10
Reason for differences

• Rover Executive
• Coordination of larger number of parallel faults
• Ability to modify a run-time constraints without
  having to re-compile code
• Coordination between tasks does not need to be
  programmed
• Instrument Manager
• Programming with procedures is perceived as more
  natural especially for nominal operations
• Less overhead due to planning, faster reactions

11
Rover Executive Fault Recovery Status
Tested multiple fault recoveries in
simulation Ready to go for field test on rover
12
Rover Executive Performance
Nominal Execution
Multiple Fault Recoveries
13
Instrument Manager Interfaces to instrument
controllers

• Instrument Manager can currently command the SPI
  and PTU controllers (e.g., take a panorama)
• Current API implementation not frozen
• SPI controller connected to hardware
• PTU operates with stubs
• FI
• VNS early phase of API design
• Plow

14
Expectation for field campaign

• Rover Executive
• 2004 operate rover for one day without
  human intervention in the presence of faults
• 2005 demonstrate quick modification of
  onboard control model while mission is
  ongoing
• Instrument Manager
• 2004 execution of single and multiple
  instrument complex observations
• 2005 dynamic instrument calibration and
  instrument fault handling

15
Requirements for field campaign

• Logging on message incoming and outgoing message
  traffic
• Internal logs
• Ability to inject all faults and explicit field
  scenario that excercize all of them.