AI in engineering & robots

How do a million robots avoid bumping into each other?

Warehouse robots, self-driving cars and software that designs lighter, stronger parts all use AI to plan, predict or react. Engineering is where AI leaves the screen and moves real things in the real world.

Warehouse grid with two robots following planned paths around shelves TWO ROBOTS · PATHS TIMED SO THEY NEVER MEET
In a robotic warehouse, software plans every robot’s path to reduce collisions and congestion so hundreds of them can share the floor, sometimes by making one robot wait a moment while another passes.

What AI is doing right now

Four ways AI is building and moving things

Engineers use AI to plan routes, drive cars, design parts and teach robots to move, often with sensors as the AI’s eyes.

01

Running robot warehouses

Robots carry shelves and packages across huge warehouse floors. The hard part is coordinating all of them at once.

Real exampleIn 2025 Amazon deployed its one millionth robot and an AI model called DeepFleet, which Amazon says cuts its robots’ travel time by 10%. Amazon

02

Driving cars with no driver

Self-driving cars combine cameras, radar and lidar, then use AI to recognize everything around them and plan each move.

Real exampleBy March 2026, Waymo was giving about 500,000 paid robotaxi rides per week across 10 U.S. metro areas. TechCrunch

03

Designing better parts

With generative design, an engineer sets the goals (strength, weight, cost) and the software tries thousands of shapes.

Real exampleGeneral Motors and Autodesk used generative design to create a seat bracket that was 40% lighter and 20% stronger than the original. Autodesk

04

Teaching robots to walk

Instead of programming every step, engineers let an AI practice in a simulator first, where a virtual robot can trip over obstacles safely before the real one takes a step.

Real exampleETH Zurich’s four-legged robot ANYmal learned to walk this way. It then hiked up a Swiss mountain, climbing 120 meters in 31 minutes without a fall, 4 minutes faster than the time estimated for human hikers. ETH Zurich (2022)

Try it yourself

Plan a robot’s path

Click or tap squares to add or remove shelves (or use the arrow keys and Enter). The robot instantly finds the shortest route to the packing station, the same basic problem warehouse software solves over and over, every second.

22squares to travel
12shelves on the floor
123squares the planner checked

Shortest route found in a blink: 22 squares. Add shelves and watch it re-plan. Can you block the robot completely?

This uses breadth-first search, a classic planning method. Real warehouses also have to plan for hundreds of robots moving at once, which is where AI comes in: predicting traffic jams before they happen. How the planner searches, and how that differs from learning: Search and planning →

The pattern behind every AI tool

Data → pattern → prediction

Every “AI in…” page follows the same five steps. Learn them once and you can understand AI in any field.

1 · Data

Sensors everywhere

Cameras, lidar, GPS and position markers report where every robot, car or part is, many times a second.

2 · Pattern

Where trouble happens

Traffic jams, pedestrians stepping out, parts that bend under stress: patterns that repeat.

3 · AI

Plan the next move

The system predicts what will happen and chooses an action: turn, slow down, reroute.

4 · Check

Simulation & testing

Millions of miles or moves in simulation, then careful real-world testing with safety rules.

5 · Use

Moving the real world

Faster deliveries, safer roads, lighter parts and fewer breakdowns.

Think like a scientist

What AI can’t do here

Knowing the limits is part of understanding the tool.

Surprises are dangerous

Rare situations, like an odd object in the road or a worker in an unexpected spot, are hardest to train for.

Sensors can be fooled

Heavy rain, snow, glare or dirt on a camera can confuse what the AI “sees.”

Who’s responsible?

When a robot or self-driving car makes a mistake, deciding who is at fault is still being worked out.

Jobs change

Robots take over some tasks, while creating jobs in maintenance, programming and supervision.

Where people stay in charge: engineers set safety limits, test systems before they go live, monitor them in operation, and step in when the robots get stuck. Many robot fleets have people supervising remotely.

Careers

Jobs where people work with this AI

More and more technical jobs involve working alongside robots and the software that runs them.

Robotics engineer

Designs robots and the software that controls them.

Working with AIUses AI so robots can recognize objects and plan movements.

Robotics technician

Installs, maintains and repairs robots in warehouses and factories.

Working with AIUses AI-driven alerts that predict which robot needs service before it breaks.

Autonomous vehicle safety engineer

Tests self-driving systems and finds situations where they fail.

Working with AIBuilds simulations to test AI drivers on rare and risky events.

Operations manager for automated sites

Runs a facility where people and robots work together.

Working with AIUses AI dashboards to balance robot traffic, staffing and safety.

Classroom activity

Program a human robot

40 minutesGrades 4–10Masking tape, paperNo computers

An unplugged activity: students program a classmate “robot” through a taped grid, then discover why coordinating many robots needs smarter planning.

Teacher tip: tape the 6×6 grid before class. Make the squares big enough to stand in, about 40 cm (16 in) wide. If your floor has square tiles, just mark the corners.

  1. Set up (5 min): On the taped 6×6 grid, place a few paper “shelves” as obstacles and mark a start square and a “packing station.”
  2. Program (10 min): In teams, write instructions using only FORWARD (move one square), LEFT and RIGHT (turn 90° on the spot). A classmate “robot” follows them exactly.
  3. Optimize (10 min): Rewrite to use the fewest instructions. Then count only your FORWARDs: that is the number of squares traveled, the number the simulator above keeps as small as possible. Is the route with the fewest instructions also the shortest?
  4. Add traffic (15 min): Run two or three robots at once from different corners. What goes wrong? Invent rules (right of way, waiting) that prevent crashes.

Discussion questions

  • Why is planning for 1,000 robots harder than for one?
  • What should a robot do when it doesn’t know what an object is?
  • Which jobs in a robot warehouse still need people? Why?

Fresh from the lab

Latest research in engineering & robots

New studies are added every week and explained in plain language. Not every study uses AI, so ask yourself where AI could help. See all research →

New stories in this area are on the way. Browse all research →

Put curiosity to work

Read it. Try it. Question it.

Explore new AI research, or take a paper-first investigation into your classroom.