Obstacle Avoidance

A robot senses an obstacle and adjusts its path or motion ahead of time so it doesn't run into it.

Obstacle avoidance means a robot detects an obstacle while moving and steers around it while still reaching its goal. It's a real-time sense-decide-act process: sensors such as lidar, depth cameras, or ultrasonics detect the obstacle, and a planning or control module adjusts the path. Approaches fall into roughly three categories: computing a collision-free route up front during global planning, using algorithms like A* or RRT; reacting locally in real time, such as a mobile robot's costmap plus dynamic window approach, or a robot arm's collision checking or control barrier functions; and learning avoidance behavior directly from sensor data with methods like reinforcement learning. It's different from 'collision detection (robot safety)': obstacle avoidance steers around a collision before it happens, while collision detection notices a collision after it has already occurred and stops or backs away. Moving obstacles, such as walking people, are much harder to handle than static ones, since the robot also has to predict how they'll move.

ExampleA vacuuming robot follows its planned cleaning route and encounters a slipper freshly dropped on the floor; the local planner steers around it in real time based on the lidar data and updates the obstacle into its map.

Also called

Collision Avoidance

Related

Motion Planning · Collision Checking · Costmap · Dynamic Window Approach · Control Barrier Function · Collision Detection (Robot Safety)

Sources

Wikipedia: Obstacle avoidance
Wikipedia: Motion planning