Computational Challenges for Model-based Autonomous Systems Prof. - - PowerPoint PPT Presentation

Space Systems & Artificial Intelligence Laboratories Massachusetts Institute of Technology

Computational Challenges for Model-based Autonomous Systems

Artificial Intelligence & Space Systems Laboratories Massachusetts Institute of Technology

• Prof. Brian C. Williams

Space Systems & Artificial Intelligence Laboratories Massachusetts Institute of Technology

Idea: Support programmers with embedded languages that avoid commonsense mistakes, by reasoning from hardware models.

Polar Lander Leading Diagnosis:

• Legs deployed during descent.
• Noise spike on leg sensors

latched by software monitors.

• Laser altimeter registers 50ft.
• Begins polling leg monitors to

determine touch down.

• Latched noise spike read as

touchdown.

• Engine shutdown at ~50ft.

Reactive Model-based Programming

Model-based Executives automate all reasoning about system interactions.

• Scheduling
• Command

confirmation

• Diagnosis
• Commanding
• Configuration
• Repair . . .

Space Systems & Artificial Intelligence Laboratories Massachusetts Institute of Technology

Model-based Programs Interact Directly with State

Embedded programs interact with plant sensors and actuators:

• Read sensors
• Set actuators

Embedded Program S Plant Obs Cntrl

Programmers must map between states and sensors/actuators. Model-based programs interact with plant state:

• Read state
• Write state

Model-based Embedded Program S Plant

Model-based executives map automatically between states and sensors/actuators.

S’ Model-based Executive Obs Cntrl

Space Systems & Artificial Intelligence Laboratories Massachusetts Institute of Technology

Example: The model-based program sets the state to thrusting, and the model-based executive . . . .

Determines that valves

• n the backup engine

will achieve thrust, and plans needed actions. Deduces that a valve failed - stuck closed Plans actions to open six valves

Fuel tank Fuel tank Oxidizer tank Oxidizer tank

Deduces that thrust is off, and the engine is healthy

Closed Closed

Valve Valve

Open Open Stuck Stuck

• pen
• pen

Stuck Stuck closed closed

Open Open Close Close

• 0. 01
• 0. 01
• 0. 01
• 0. 01

0.01 0.01 0.01 0.01

inflow = outflow = 0

Modeling Complex Behaviors through Probabilistic Concurrent Constraint Automata

• Complex, discrete behaviors
• Anomalies and uncertainty
• Physical interactions
• Timing
• modeled through concurrency, hierarchy and non-determinism.
• modeled by probabilistic transitions
• modeled by discrete and continuous constraints
• modeled by simple temporal networks

System Model

Commands Observations

Control Program Plant Deductive Controller

Model-based Executive

(Livingstone, Titan, Kirk..)

Model-based Program

Model-based Autonomy Architecture

State goals State estimates

Tracks likely state trajectories

Mode Estimation Mode Reconfiguration Finds best target

Plans reactively

Control Program Sequencer Performs lazy scheduling Performs lazy scheduling Searches for optimal feasible threads of execution Searches for optimal feasible threads of execution

Plans Plan Failures

Computational Challenges:

• Propositional Satisfiability
• Optimal CSPs
• Graph-based Planning
• Scheduling

Executes concurrently Preempts Asserts and queries states Chooses based on reward Expresses temporal and

resource constraints

Closed Closed

Valve Valve

Open Open Stuck Stuck

• pen
• pen

Stuck Stuck closed closed

Open Open Close Close

• 0. 01
• 0. 01
• 0. 01
• 0. 01

0.01 0.01 0.01 0.01

inflow = outflow = 0

Space Systems & Artificial Intelligence Laboratories Massachusetts Institute of Technology

Future autonomy requires reasoning about hybrid discrete/continuous systems

Detecting subtle failures Compute hybrid of:

• temporal constraint problem
• mixed integer linear program

Compute hybrid of:

• HMM belief update
• Kalman filtering

Coordinating fleets of agile vehicles

X Y

d

x x d x x d y y d y y d

p q q p p q q p

− ≥ − ≥ − ≥ − ≥

• r
• r
• r

x x d Mb x x d Mb y y d Mb y y d Mb b

p q pq q p pq p q pq q p pq pqk k

− ≥ − − ≥ − − ≥ − − ≥ − ≤

=

∑

1 2 3 4 1 4

3 and and and and

Probabilistic mode transition

• ld estimate:

new estimate: continuous state evolution within mode