TARKKA

The tool for the situational awareness.
PRODUCTS

Software suite for environment perception

The main functionality is obstacle detection and tracking, but also has software modules for segmentation, classification, and traversability mapping that can be used to tailor a customer-specific perception solution. The goal of the perception system is to provide accurate obstacle information for risk assessment, and therefore reduce the amount and severity of vehicle collisions. Risk assessment is out of scope for TARKKA.

Advanced obstacle detection approach that utilises high-precision localisation systems such as PAIKKA, to measure obstacle presence on planned routes, further ahead than KILPI is capable of – even up to 100 metres. It is typically used as a standalone system in Driver Assistance systems, or in AGVs/AMRs as an additional soft stop before actual safety mechanisms are triggered. TARKKA SW also includes the basic obstacle detection functionalities of KILPI.

TARKKA has been designed and tested to function only in specific conditions:

  • Scenery: Accounts for the non-movable elements associated with the operating environment (e.g., buildings and other structures).
  • Environmental and weather conditions: Applies to weather and atmospheric conditions that may be apparent within the operating environment.
  • Dynamic Elements: Accounts for ego-motion of the machine and all movable elements within the operating environment, an example of this being other machines or vehicles.

 

SUPPORTED SENSORS

Paikka supports the following sensor modalities:

  • 3D LIDAR(s)
  • IMU
  • Wheel odometry
  • RTK-GNSS
VIEW DIAGRAM

FEATURES

TARKKA has two layers for obstacle detection, without considering risk assessment or control of vehicles required after obstacle detection has been performed.

The first level, the Reactive Layer, is covered by the same functionality as the KILPI. It is a basic obstacle detection approach that monitors the immediate surrounding of the AGV for lidar visible obstacles. It does not require localisation information to function. Instead, it takes velocity and steering angle as an input, and creates a region of interest e.g. in front, in the back, or in the sides of the robot depending on lidar visibility. It is typically used as a soft stop before actual safety mechanisms (physical safety bumper or safety lidar) are triggered.

The second level, the Predictive Layer, is an advanced obstacle detection approach that utilises high-precision localisation systems such as the PAIKKA, to track obstacles on planned routes, further ahead than the Reactive Layer is capable of. The Predictive Layer is the level where a more advanced approach for Obstacle Detection is implemented. The Predictive Layer can observe contents of a polygonal free space model, so-called “Virtual Tunnel”. The tunnel is essentially defined as “all the space that a vehicle requires to move safely ahead”, considering width and height of the vehicle. Predictive Layer requires PAIKKA to function, see further details in PAIKKA specification and it has four main software modules:

  • Obstacle Detector: Detects obstacles along the trajectory from a single lidar scan.
  • Detection filter: Filters irrelevant detections to improve tracking accuracy, and to avoid false alarms.
  • Multiple Object Tracker: Associates detections to tracks. Tracks each obstacle’s position, size and velocity over time. Assigns a unique identifier to each track.
  • Collision Monitor: predicts obstacle movement inside the virtual tunnel. It calculates the obstacle’s time and distance when it’s predicted to enter and exit the virtual tunnel.

BENEFITS

Safety
Real-time driving assistance and collision avoidance even in low visibility conditions due to rain, snow and dust.
Flexibility
Various object detection and tracking algorithms can be adapted and trained to support intelligent and autonomous functions
Efficiency
Utilizes sensor fusion and intelligent sensor systems to enable fast integrations.

VIDEOS

Skoda
WIP (Old material)
Trombia Free short

PICTURES