SHAKEY: First Mobile Intelligent Robot
Stanford Research Institute develops SHAKEY, the first mobile robot capable of perceiving its environment and acting autonomously.
Overview
Between 1966 and 1972, Stanford Research Institute (SRI) developed the SHAKEY robot, the world's first mobile robot capable of perceiving its environment, planning paths, and acting autonomously. SHAKEY was equipped with a television camera, optical rangefinder, and bump sensors, able to navigate its environment, push objects, and perform simple tasks. It integrated visual perception, path planning, logical reasoning, and control execution in one system, pioneering the classic perception-planning-action paradigm in robotics. The SHAKEY project produced numerous important contributions, including the A* search algorithm and the STRIPS planning system, profoundly influencing later robotics and AI planning fields.
Study Notes
Background, mechanics, impact, and key takeaways in one read
Background & Context
From 1966 to 1972, the Stanford Research Institute (SRI) built SHAKEY, widely recognized as the first mobile robot that could "reason about its own actions." Proposed by Charles Rosen and funded by DARPA from 1965, its goal was to let an autonomous device function in real environments. Roughly human-sized, SHAKEY carried a TV camera, laser rangefinder, and bump sensors, linked by radio to a large computer (first an SDS-940, later a PDP-10). In rooms strewn with blocks and ramps, it understood commands like "push the block off the platform," planned paths, avoided obstacles, and acted—integrating perception, world modeling, planning, and execution for the first time.
Key Points
- 1Built at SRI from 1966 to 1972, it was the first mobile robot to reason logically about its own actions.
- 2Proposed by Charles Rosen, with key contributors including Nilsson, Raphael, Fikes, and Hart at SRI.
- 3It first applied the STRIPS planner, which decomposed high-level commands into sequences of basic actions.
- 4For navigation it used the co-invented A* heuristic search algorithm to plan optimal, obstacle-avoiding paths.
- 5Its software split into perception, world modeling, planning, and execution—a template for later autonomous systems.
- 6It earned an IEEE Milestone award in 2003 and joined the Robot Hall of Fame in 2004.
How It Works
SHAKEY's breakthrough was "think before you move." It first built a symbolic world model from camera and rangefinder data (e.g., "block A is in room B"). A high-level command is translated by the STRIPS planner: STRIPS uses operators expressed as preconditions–action–effects, chaining backward from the goal state to the needed action sequence, and replans when unexpected obstacles appear. Pathfinding uses A*, which estimates each node's "cost so far + heuristic to goal" and expands the least total estimate first, guaranteeing optimality while beating blind search. Think of STRIPS as a manager breaking "deliver the file" into tasks, and A* as a courier picking the shortest route on a map. The control architecture decouples perception, modeling, planning, and execution—still the standard robot paradigm today.
Impact & Legacy
SHAKEY validated the perceive–plan–act framework; its STRIPS and A* became cornerstones of planning and search, with A* still used in game pathfinding, map navigation, and Mars rovers. Its limits were stark: constrained by 1960s compute and sensors, it worked only in simple preset labs, often taking hours per command, and its retractable arm was never built. Its real value was architectural—marrying symbolic reasoning with sensor feedback and pointing the way to modern autonomous systems. The layered architecture it pioneered remains the reference design for autonomous robots today.
Key Takeaways
- Robot intelligence = perception/modeling + symbolic planning + execution; all three must connect.
- A* balances optimality and efficiency via "cost so far plus heuristic"—a model search algorithm.
- Layered, decoupled architecture (perceive/plan/act) matters more for scalability than raw hardware.
Glossary
A classic planning system using preconditions–action–effects operators; first applied on SHAKEY.
A graph search combining cost-so-far with a heuristic estimate; optimal yet far more efficient than blind search.
The classic autonomous-system loop of perceiving, planning, and acting; first integrated on a robot by SHAKEY.
Related Links
Related Events
Stanford Cart: Autonomous Driving Pioneer
Stanford Cart achieves autonomous navigation under Hans Moravec, traversing obstacle courses using stereo vision.
Stanford Cart: Autonomous Driving Pioneer
Stanford Cart achieves autonomous navigation under Hans Moravec, traversing obstacle courses using stereo vision.