Control Algorithms

Modular controllers for precise movement, tracking, and avoidance.

Kompass provides a suite of battle-tested control algorithms. These range from classic geometric path-followers to modern, GPU-accelerated planners that handle dynamic obstacle avoidance or vision target following in real-time.

Algorithm Suite

DWA

Velocity Space Planning A real-time, GPU-accelerated planner that samples reachable velocities to balance goal progress and obstacle clearance.

DWA
Pure Pursuit

Geometric Path Tracking Calculates curvature to reach a lookahead point. Includes a reactive search layer to deviate for obstacles.

Pure Pursuit
Stanley Steering

Axle-Based Feedback Optimized for Ackermann platforms, providing smooth exponential convergence to a reference path.

Stanley Steering
DVZ

Reactive “Bubble” Control A highly efficient method that models safety zones as deformable perimeters for rapid avoidance.

DVZ
Vision Follower (RGB)

2D Visual Servoing Maintains target centering and relative scale using standard monocular cameras.

Vision Follower (RGB)
Vision Follower (RGB-D)

3D Target Tracking Uses depth data and DWA-style planning to follow targets while actively avoiding obstacles.

Vision Follower (RGB-Depth)

Kinematic Compatibility

Every algorithm in the Kompass stack is natively compatible with the three primary motion models. The internal logic automatically adapts to the specific constraints of your platform:

  • ACKERMANN: Car-like platforms with steering constraints.

  • DIFFERENTIAL_DRIVE: Two-wheeled or skid-steer robots.

  • OMNI: Holonomic systems capable of lateral movement.

Flexible Parameterization

Each algorithm is fully parameterized. Developers can tune behaviors such as lookahead gains, safety margins, and obstacle sensitivity directly through the Python API or YAML configuration. This modularity ensures the same code can run on a small indoor lab robot or a heavy-duty outdoor platform.