Homotopy Class

• Also called: homotopy class of paths
• Category: Planning & manipulation › Motion planning
• Tags: motion-planning, topology, trajectory-optimization, mobile-navigation

A set of paths with the same start and goal that can be continuously deformed into one another without passing through an obstacle, such as all paths that go left of a given pillar.

Definition

Two paths with the same start and end points are homotopic if one can be bent continuously into the other while the endpoints stay fixed and every intermediate path stays in free space. This relation splits all possible paths into equivalence classes, called homotopy classes. In a plane with one obstacle between start and goal, paths that pass to its left form one class and paths that pass to its right form another; paths that loop around the obstacle one or more times before reaching the goal form further classes, so even a single obstacle produces infinitely many classes, distinguished by their winding number. Each extra obstacle multiplies the possibilities.

The concept matters because most path optimizers work locally. Gradient-based trajectory optimizers and elastic bands deform a path to make it shorter or smoother, but they cannot push it through an obstacle, so they only find the best path within the class they started in. Global planners decide which class to use, often implicitly. A sampling-based planner such as RRT may first find a path through one gap and later discover a shorter path through another, which shows up as a sudden drop in path cost rather than a gradual improvement. Homotopy-aware planners make this explicit by tracking several classes in parallel and optimizing the best path in each.

In practice, planners often use a computable invariant instead of homotopy itself. For planar problems, an H-signature or a winding-angle count around each obstacle labels which side of each obstacle a path passes, which strictly identifies the related but coarser homology class. In the higher-dimensional configuration space of an arm, classes still exist but are hard to enumerate, so they are mostly discussed for mobile robots, vehicles and planar problems.

Why it matters

Knowing which homotopy class a path belongs to explains why a local optimizer gets stuck with a long detour, why a planner's cost jumps suddenly, and why a robot may oscillate between passing an obstacle on the left and the right. Planners that reason about classes avoid these failures and can offer genuinely different route options.

Key applications

• Local planners for mobile robots that keep alternative routes around obstacles
• Explaining step changes in solution cost in anytime and asymptotically optimal planners
• Choosing initial guesses for trajectory optimization
• Multi-robot and crowd navigation where passing side must be agreed
• Tethered robots, where the cable's class around obstacles limits reachable paths

In practice

A warehouse robot uses a local planner that keeps several candidate trajectories, each in a different homotopy class around nearby pallets and people. Every cycle it optimizes each candidate for time and clearance, then picks the cheapest. Without care, two nearly equal options cause the robot to switch sides from one cycle to the next and wobble in place, so the team adds hysteresis: a new class must beat the current one by a margin before the robot changes its mind. They also cap the number of classes explored so the planner stays within its cycle time.

Common misconceptions

• Homotopic paths have similar lengths — homotopy is a topological property; two paths in the same class can differ greatly in length
• There are only as many classes as there are ways around the obstacles — looping around an obstacle repeatedly creates infinitely many classes
• A local optimizer will find the globally shortest path given enough iterations — it stays in the class of its initial path
• Homotopy and homology classes are the same — homology is coarser, and practical signatures such as the H-signature identify homology classes

Who should know this

• Motion planning researchers working on optimal and multi-modal planning
• Mobile robot navigation engineers tuning local planners
• Trajectory optimization practitioners choosing initializations
• Students studying configuration-space topology

Related technologies

• H-signature
• Timed Elastic Band (TEB) planner
• Winding number
• Voronoi diagram
• Visibility graph
• CHOMP and TrajOpt

Related terms

• Configuration Space
• Elastic band
• RRT*
• Anytime Motion Planning
• Narrow Passage Problem
• Motion Planning

Related projects

• OMPL: Sampling-based motion planning library powering MoveIt and countless robotics applications
• MoveIt 2: Motion planning framework for ROS 2 robot arms, with kinematics and collision checking
• PyRoki: Modular Python toolkit for kinematic optimization across CPU, GPU, and TPU via JAX
• mink: Python inverse kinematics library built on MuJoCo supporting 20+ robot platforms
• Ruckig: Real-time jerk-constrained time-optimal trajectory generation for robots and machines
• NeuPAN: Real-time end-to-end planner that maps raw lidar points directly to robot actions