Autonomy 1.5: Smarter Systems Without the Billion-Dollar Bet

There’s a lot of noise today around the future of autonomy.

Some are doubling down on Autonomy 1.0, classical systems engineering: modular pipelines (perception -> fusion -> planning -> control), reliable and certifiable… but increasingly hard to scale in complex, dynamic environments.

Others are pushing toward Autonomy 2.0, end-to-end deep models: pixels -> actions, minimal structure… but with major challenges around explainability, data requirements, and certification.

The exciting direction is Autonomy 1.5

A cognitive hybrid approach that combines the best of both worlds. At its core:

• System 1: a primitive sensorimotor + actuation layer (perception, fusion, control, execution, safety loops)

• System 2: a learning-based decision layer operating on a semantic understanding of the world

The key enabler: Latent World Models

Instead of modelling the full physical world, Autonomy 1.5 focuses on learning compact, decision-relevant representations.

👉 Not “what the world looks like” 👉 But “how the world evolves when we act”

These latent world models allow the system to:

• Predict outcomes before acting

• Evaluate multiple alternatives quickly

• Handle uncertainty and partial information

• Scale to multi-agent, real-world scenarios

And importantly, they exist across hierarchy levels (from platform to system-wide decisions), not as a single monolithic model.

Why this matters

Autonomy 1.5 enables something fundamentally new:

👉 Systems that don’t just react, they understand, predict, and decide

The added value of Autonomy 1.5

• Scalable decision-making in complex, multi-agent environments

• Explainability through structured, semantic reasoning

• Efficient learning without requiring massive end-to-end datasets

• Separation of concerns (perception vs. decision vs. actuation)

• Faster iteration via simulation and data flywheel

• Compatibility with existing systems (no need to start from scratch)

We don’t need to choose between rigid engineering and black-box AI.

👉 Autonomy 1.5 is the practical path forward, bringing intelligence into systems without losing control.