Episodic control

As explained in Delivering Machine Learning Projects: From design to deployment

Episodic control is an autonomy-management technique in which a system independently performs bounded episodes of activity subject to human approval or intervention between episodes.

Episodic control manages autonomy by having a system prepare and execute a bounded episode after human approval. It balances communication constraints, mission value, and operational risk. section 9.2.2

Apply the technique by:

• Preparing a command or action plan for review. section 9.2.2
• Allowing independent execution only after approval. section 9.2.2
• Constraining each episode so its risk remains acceptable between intervention points. section 9.2.2

Mars rover operations

Perseverance uses delegated episodes because communication with Mars makes manual control slow, while full autonomy would expose a high-value mission to unacceptable risk. section 9.2.2

Figure 9.3 The control flow for a delegative remote control rover [1]

Trace the approval-to-execution control flow for a delegative remote-control rover. section 9.2.2

How Delivering Machine Learning Projects: From design to deployment covers it

How deep it goes

Explains approval and autonomous execution, but stops short of specifying episode design or implementation details.

Examples

Uses Perseverance rover operations and mobile phone network cells to illustrate bounded autonomy and localized failure.

What it assumes

Assumes familiarity with autonomous and manual control, while grounding the trade-off in communication constraints and operational risk.

What you can do after

Evaluate autonomy strategies against communication delay, system value, failure tolerance, and remediation requirements.

How it connects

Connects episodic control with delegative control, manual control, subsystem partitioning, failover, replication, and service-level agreements.

liveBookmanning.com

Read Delivering Machine Learning Projects: From design to deployment on liveBook Get the book at manning.com

Related concepts

• delegative systemuses · A delegative system is a system that makes decisions on a user's behalf while providing the human with timely mechanisms to review, correct, or override those decisions.

This explainer was generated by Manning's knowledge graph from the text of Delivering Machine Learning Projects: From design to deployment. The section links point to the parts of the book it draws on.