AGV localization solution JACK-SLAM

for automated forklifts

Overview
JACK-SLAM is an indoor, camera-based visual localization system that provides real-time 6 DOF position (X, Y, Z, Rx, Ry, Rz) for robots and mobile assets without the need for beacons or changes to the venue. It uses natural, stable landmarks already present in the environment to deliver centimeter-level localization accuracy suited to industrial applications.

Details
JACK-SLAM focuses on localization (not path planning). It supplies real-time pose information that navigation stacks, trackers, and control systems use to make movement decisions indoors. The system extracts and tracks visual features such as corners, ceiling elements, shelf edges and color variations instead of relying on installed infrastructure.

Key capabilities

• Compatible with standard monocular, stereo, and depth cameras.
• Available as JACK-SLAM software or as ready-to-use THE-JACK-BOX hardware.
• Can complement existing LIDAR-based navigation systems.
• Applicable to robots, AMRs, AGVs, forklifts, cleaning machines and other mobile industrial assets.

Performance highlights

• Position update rate: up to 20 results per second (20 Hz).
• 6 DOF output: X, Y, Z positions plus Rx, Ry, Rz rotations.
• Centimeter-level indoor localization precision under typical conditions.
• THE-JACK-BOX hardware: operates on 9–60 VDC and is mountable on various mobile platforms.

Advantages

• Cost-efficient: uses existing camera hardware and avoids beacon infrastructure and installation.
• Fleet scalable: maps and updates can be shared across a fleet with automatic map updates when revisiting known areas.
• Robust in mixed-traffic environments: maintains localization performance around humans, robots and moving machinery.
• Tracker mode: THE-JACK-BOX can monitor multiple forklifts or mobile machines in real time to support routing and collision reduction.
• Flexible deployment: choose hardware (THE-JACK-BOX) or integrate JACK-SLAM software into embedded Linux systems.
• Customizable interfaces: engineers can adapt sensors, camera setups and software behavior for specific use cases.

How it works
The algorithm detects and tracks approximately 100–200 distinctive image points per frame and computes the instantaneous pose of the camera-bearing device from their motion across frames.

• Relies on stable features such as ceiling elements, corners, shelves and color contrasts.
• Supports camera orientations toward ceilings to exploit fixed reference elements.
• Operates in environments where LIDAR may be intermittently occluded.
• Improves map consistency with loop-closure mechanisms.

Technical specifications

• Position update rate: up to 20 results per second (20 Hz).
• Degrees of freedom: 6 DOF (X, Y, Z, Rx, Ry, Rz).
• Localization precision: centimeter-level indoor accuracy.
• Feature tracking: ~100–200 points identified and tracked per image frame.
• Camera inputs supported: monocular, stereo and depth cameras.
• Deployment options: software license or THE-JACK-BOX hardware (ready-to-use tracker or embedded integration).
• Power (THE-JACK-BOX): 9–60 VDC.
• Typical use cases: forklifts, AGVs, AMRs, cleaning machines, drones and service robots.
• Operates without beacons or venue modifications and is designed to complement existing navigation stacks (including LIDAR).