✧ PARS is a new research & development project, granted within the CENTRIC project: Center for knowledge transfer to enterprises in the ICT field, which will be implemented in collaboration with a research team from the Ștefan cel Mare University of Suceava. 

The Autonomous Recognition and Support Platform (PARS) research project envisages a solution to assist public authorities in critical situations that call for quick response times, as well as optimal resource coordination and information flow. The proposed solution allows for the fast identification of weak points and the generation of dynamic solutions adapted to the situation on the ground. This will be made possible through the implementation of a software and hardware ecosystem made up of a command center and an autonomous swarm of drones able to carry out a wide range of missions, such as reconnaissance, alert, and tracking.  

Within this project, ASSIST Software will carry out research and development activities and will be responsible for UI/UX prototyping, software architecture elaboration, and software development of multiple components, such as frontend, backend, and ML. Furthermore, our team has taken on the task of creating an API connector that enables agents to be trained within Unreal Engine 5, and a semi-procedural simulated environment in UE5, in which the agents (drones) will be trained. Another exciting task is to develop an Image Recognition module able to identify real-life objects after being trained with high-fidelity synthetic images.  

➸ Activities: 

• State-of-the-art of surveillance and reconnaissance applications
• Methods in which autonomous drone swarm can be integrated into surveillance and reconnaissance applications
• Research and development of a prototype that would support the scope of the project.

➸ Requirements:

• Deliver monthly research reports documenting the work done
• Publish at least one research article published in a journal indexed by BDI/WoS.

➸ Apply for a patent High-level components:

• The ML Swarm Agent, utilizing a reinforcement learning model to control a group of drones in response to detected hazards or incidents.
• The Drone Controller, which operates the drones in the field by processing and transmitting commands from the ML Swarm Agent, such as movement, object tracking, and perimeter scanning.
• The Virtual Controller, which enables manual drone control via peripheral devices like joysticks or keyboards.
• The Drone Camera System, responsible for capturing and transmitting images to the Image Processing component, which extracts information and adds labels to images for analysis by the ML Swarm Agent module.
• The Image Generator, which generates synthetic images for use in training the ML algorithm.
   

Note that, for this project, the Image Processing module is separate from the ML Swarm Agent and will demonstrate the capability of training an AI actor to recognize real-life entities using synthetic images as input.
 

Project ID: 22081/05.10.2022 

Funding: European Regional Development Fund, Competitiveness Operational Program 2014-2020, CENTRIC 

Research articles

For more insight, we recommend reading these articles that our colleagues published in peer-approved publications, detailing processes, ideas, and the technology used.  

• Experimental Results on Synthetic Data Generation in Unreal Engine 5 for Real-World Object Detection 2023, 17th International Conference on Engineering of Modern Electric Systems (EMES) 

Discover an innovative breakthrough in the world of object detection (OD) algorithms! Our research delves into a groundbreaking method for creating top-notch synthetic training data that fuels the future of OD technology. By harnessing the power of photogrammetry, we seamlessly transform real-world objects into precise 3D digital replicas, all within the immersive realm of Unreal Engine 5 (UE5).