[Paper Review] Behavior Trees in Robotics and AI: An Introduction

Michele Colledanchise, Petter Ögren|arXiv (Cornell University)|Aug 31, 2017

Reinforcement Learning in Robotics56 references184 citations

TL;DR

This paper (book chapter) introduces Behavior Trees (BTs) as a modular, reactive control architecture for robotics and AI, contrasts them with FSMs, and surveys foundations, examples, extensions, analysis, planning, and learning aspects.

ABSTRACT

A Behavior Tree (BT) is a way to structure the switching between different tasks in an autonomous agent, such as a robot or a virtual entity in a computer game. BTs are a very efficient way of creating complex systems that are both modular and reactive. These properties are crucial in many applications, which has led to the spread of BT from computer game programming to many branches of AI and Robotics. In this book, we will first give an introduction to BTs, then we describe how BTs relate to, and in many cases generalize, earlier switching structures. These ideas are then used as a foundation for a set of efficient and easy to use design principles. Properties such as safety, robustness, and efficiency are important for an autonomous system, and we describe a set of tools for formally analyzing these using a state space description of BTs. With the new analysis tools, we can formalize the descriptions of how BTs generalize earlier approaches. We also show the use of BTs in automated planning and machine learning. Finally, we describe an extended set of tools to capture the behavior of Stochastic BTs, where the outcomes of actions are described by probabilities. These tools enable the computation of both success probabilities and time to completion.

Motivation & Objective

• Explain what Behavior Trees are and why they offer reactiveness and modularity over FSMs.
• Show classic BT formulations and execution semantics with illustrative examples.
• Demonstrate BT applications in robotics through Pac-Man and mobile manipulators.
• Survey design principles, extensions, and analytical tools for safety, efficiency, and robustness.
• Outline connections between BTs, planning, and machine learning.

Proposed method

• Define BTs as rooted trees with control flow and execution nodes and describe tick-based execution semantics.
• Present four control-flow node types (Sequence, Fallback, Parallel, Decorator) and two execution node types (Action, Condition).
• Explain control-flow nodes with memory as a mechanism to improve reactivity and discuss memory-less equivalence.
• Provide concrete BT examples for Pac-Man and a mobile manipulator to illustrate modular construction.
• Discuss extensions to BTs (Utility, Stochastic BTs, dynamic expansions) and their implications.
• Survey BTs in planning (PA-BT, ABL) and learning (GP-BT, RL-BT, learning from demonstration).

Experimental results

Research questions

• RQ1 How do BTs generalize earlier switching architectures such as FSMs, subsumption, teleo-reactive programs, decision trees, and sequential behavior composition?
• RQ2 What design principles improve readability, reactivity, safety, and deliberation in BTs?
• RQ3 How can BTs be extended and integrated with planning and learning to enhance autonomy?
• RQ4 What are the practical benefits and tradeoffs of using BTs in real robots and AI systems?

Key findings

• BTs offer modularity and reactivity advantages over FSMs by using two-way control transfers through tree nodes.
• Memory-enabled BTs can reduce unnecessary re-execution and improve efficiency, while non-memory BTs are more reactive.
• BTs can be incrementally built to handle tasks like Pac-Man and mobile manipulation, illustrating modular composition.
• BTs relate to and generalize earlier concepts such as Subsumption Architecture, Teleo-Reactive programs, and Decision Trees.
• BT extensions (Utility, Stochastic BTs) and planning/learning integrations enable richer, adaptive behaviors.
• BTs provide formal tools for analyzing efficiency, safety, and robustness in autonomous systems.