TUTOR

M23 · Level IV — Autonomous & Embodied Intelligence

Autonomous Systems & Decision Making

Engineer autonomous systems that make and execute decisions under uncertainty, partial observability, dynamic environments, multiple constraints, timing limits and interacting agents.

Topics

  • Observe
  • Estimate
  • Predict
  • Plan
  • Decide
  • Act
  • Monitor
  • Recover
  • Replan

Technical concepts

  • 01What Makes a System Autonomous?
  • 02Autonomy Levels as Architecture, Not Labels
  • 03State, Belief and World Model
  • 04Partial Observability and POMDPs
  • 05Belief-State Update
  • 06Prediction Under Uncertainty
  • 07Reachability and Future Sets
  • 08Mission Objectives, Utility and Risk
  • 09Risk-Sensitive Decision Making
  • 10Planning Under Dynamic Conditions
  • 11Receding-Horizon Autonomy
  • 12Decision-Making with Temporal Constraints
  • 13Anytime Planning
  • 14Search, Planning and Heuristics
  • 15Heuristics and Admissibility
  • 16Hierarchical Planning
  • 17Task and Motion Planning
  • 18Behaviour Under Uncertainty
  • 19Other-Agent Prediction
  • 20Game-Theoretic and Interactive Decisions
  • 21Multi-Agent Coordination
  • 22Communication as a Decision
  • 23Distributed State and Consistency
  • 24Safety Monitors
  • 25Runtime Assurance Architecture
  • 26Fault Detection and Isolation
  • 27Recovery and Containment
  • 28Human Oversight and Intervention Boundaries
  • 29Autonomy Under Communication Loss
  • 30Adversarial Environment Reasoning
  • 31Scenario Generation
  • 32Scenario Coverage
  • 33Counterfactual Evaluation
  • 34Causal Reasoning for Autonomous Decisions
  • 35Uncertainty Propagation
  • 36Robustness vs Performance Trade-offs
  • 37Quantum Technology Bridge: Autonomous Decision Systems

Practical implementation

  • 01Laboratory I: POMDP Belief Simulator
  • 02Laboratory II: Belief-Aware Planning
  • 03Laboratory III: Dynamic Replanning
  • 04Laboratory IV: Anytime Planning
  • 05Laboratory V: Risk-Sensitive Planner
  • 06Laboratory VI: Multi-Agent Coordination
  • 07Laboratory VII: Other-Agent Prediction
  • 08Laboratory VIII: Safety Monitor and Fallback
  • 09Laboratory IX: Fault-Injection Autonomy
  • 10Laboratory X: End-to-End Autonomous Mission
  • 11Laboratory XI: Adversarial Scenario Suite
  • 12Laboratory XII: Human Escalation Architecture

Tools and technologies

  • simulation
  • Linux VM/container
  • scenario generation
  • probabilistic state estimation planning/search
  • sensor fusion
  • multi-agent simulations
  • adversarial environments
  • safety monitors

Project

Closed-loop autonomous mission with safety monitor, fallback and scenario evidence

Expected learning outcomes

  • 01Can formalise autonomous decision problems under partial observability.
  • 02Can maintain and evaluate belief/state estimates.
  • 03Can define mission goals, risk and hard constraints.
  • 04Can plan and replan under dynamic conditions.
  • 05Can use time-bounded/anytime planning.
  • 06Can reason about other-agent behaviour and communication.
  • 07Can measure uncertainty propagation across the stack.
  • 08Can implement runtime safety monitors and fallback modes.
  • 09Can test faults, rare scenarios and adversarial behaviour.
  • 10Can design human escalation/authority boundaries.
  • 11Can produce a reproducible autonomy evidence package.

Assessment

  • 01Daily practice
  • 02Laboratories
  • 03Research reproduction
  • 04Integrated project
  • 05Technical viva
  • 06Engineering record

Source document: TUTOR Frontier Module 23 Autonomous Systems Decision Making