System Development

Risk assessment and and emergency collision avoidance


Today, contactless sensor technology mainly based on 2D laser scanners, that monitors the danger area around the AGV is the state of the art implementation for human protection.

The risk assessment and emergency collision system is significantly improved by PAN-Robots. With the additional knowledge of the AGV motion and the semantic 3D map, safety areas no longer need to be defined manually, but rather adapt automatically to the AGV dynamics and environment. In addition to the 2D safety by Laserscanners, 3D perception is provided by the omnidirectional stereo camera.


Table 1:  Comparison of the current risk assessment and emergency collision avoidance system and the PAN-Robots approach

Current solution PAN-Robots approach
Definition of safety areas static, manually defined Automatic adaptation to AGV dynamics (velocity and yaw rate)
Environment perception 2D on-board safety laser scanners 2D on-board safety laser scanner system,3D omnidirectional perception system, On-board fusion of the above systems, Control center global live view data
Obstacle type static objects, on ground, dynamic objects (assumed static) static and dynamic objects, on ground, hanging or protruding objects


The principal behavior of the AGV remains unaltered: obstacles detected in the warning field lead to a deceleration of the AGV, obstacles in the protective field trigger a complete stop immediately. In addition, dynamic objects predicted to enter the AGV driving path are considered for risk assessment and collision avoidance as well.

The proposed system thus not only decreases initial deployment efforts, but at the same time greatly enhances worker safety and accident prevention, supporting the vision of zero accidents caused by AGVs.

Pallet loading and unloading functions


The task of the stereo camera is to provide precise pallet positions for loading and unloading operations.

For loading (picking) operations the AGV needs the precise position and orientation of the pallet. For unloading (placing) operations the target pallet position is required. This enables the AGV to correctly plan the route for approaching the target and performing the required operation.



Control Center (CC) and operator HMI


Besides handling the global environmental representation and communication, the AGV fleet coordination and mission management, the control center is in charge of providing a synchronized overall system and a human-machine interface (HMI) for operators.

To handle the synchronization, the control center operates as a time server, distributing a global time reference to all PAN-Robots AGVs and infrastructure perception systems.

The HMI allows the visualization of live information about the system and offers the possibility to the user to manually define custom missions.


Global object representation environmental model and communication

Information (as persons, dynamic objects, etc) acquired by the PAN-Robots sensing systems are gathered and collected into a centralized system, called global live view. In particular, the global live view module receives information from the infrastructure sensors, and from the PAN-Robots AGVs, and constantly updates the global information regarding the position of each AGV, and the information acquired by the infrastructure sensors.

The global live view is then in charge of performing multi-sensor data fusion. In fact, different kinds of sensors provide complementary information: which is fused to obtain a consistent, robust and accurate representation of the environment.

The global live view running on a centralized system sharing the output of the data fusion process with the AGV and the Control Center.


Global navigation and local path planning

Global navigation is in charge of defining optimal paths from picking to placing positions based on the information contained in the global live view provided by the control center.

On-board local path planning addresses the need for autonomous obstacle avoidance, allowing the PAN-Robots AGV to locally deviate from the route map, in order to overcome obstacles without manual intervention.


-     Global navigation


For each AGV with an assigned mission, the global navigation needs to define the path to be traveled along the route map, in an optimized manner. Global navigation needs to define a traffic model, and exploit this model to globally optimize the performance of the AGV fleet, in terms of average time for mission completion.

-     Local path planning


Local path planning is in charge of computing deviations from the assigned route in order to circumnavigate obstacles. In presence of an obstacle in the current segment of the route map, the local path planning system defines a local deviation path that allows the PAN-Robots AGV to avoid the detected obstacle based on the data from the global live view (sent from the control center) and the onboard perception.