-
Book Overview & Buying
-
Table Of Contents
30 Agents Every AI Engineer Must Build
By :
This chapter examines how autonomous agents must be redesigned when operating in the physical world, where actions are constrained by irreversibility, latency, and finite energy. It introduces a fundamental architectural divide between depth and breadth: depth requires precise, real-time control of a single physical system, while breadth requires coordination across multiple interconnected domains with complex dependencies.
To address these competing demands, the chapter presents two complementary architectures. The Embodied Intelligence agent solves the depth problem through a multi-rate control hierarchy, belief-state planning, and strict enforcement of an admissible action set that ensures safety at every layer. The Domain-Transforming Integration agent solves the breadth problem by modeling cross-domain systems as a typed knowledge graph and propagating influence to estimate cascading effects.
These architectures are brought together in a drone mission case study...