Deep AR Law Enforcement Ecosystem



Technical Approach

At the core of the DARLENE architecture lies an ecosystem of interconnected devices, known as Internet of Things (IoT), based on real-time data collected from the patrol/situation scenes. In collecting this data, the DARLENE technology adheres to EU GDPR principles of data minimisation and data security, processing data in real-time and locally as a default unless further processing is strictly required to combat crime and terrorism.

From an architectural perspective, DARLENE is divided into two sublayers representing its functions and network characteristics. This architecture is rooted in the philosophy of collaboration between application development and IT operations teams, a methodology known as DevOps. Such an architecture improves efficiency and reduces the necessary time of data processing, which can have tighter data processing constraints. A significant technical challenge in this type of architecture is matching the information from the real-world with the information produced by the smart devices. The DARLENE cloud addresses these problems, acting as a key enabler for the ad hoc provision of such harmonised transfer of data from one device to another.

The main development effort of DARLENE from a technological standpoint is to develop this IoT ecosystem, whose conceptual architecture is shown in the visual below. The two core layers that have been mentioned above are the DARLENE cloud and the wearable Augmented Reality (AR) applications. These AR applications align with the requirements of the DARLENE use case scenarios, i.e., enhancing LEA officers’ situational awareness to detect any suspicious behaviour and prevent potential acts of violence. Crucially, these use cases will only be piloted in controlled research environments with volunteer participants, and not in real-world contexts. Furthermore, the AR applications will only identify threatening movements or objects, thus not identifying personal characteristics of individuals, in accordance with GDPR principles for the necessity of data processing.

DARLENE functionality will be offered through the “DARLENE Wearable Edge Computing Node (WECN)”, i.e., data is processed on the spot by the device itself or by a local server, as shown in the visual below, to provide users with information and services relevant to their contexts and useful for their assigned tasks. They will allow a display of information in the users’ field of view, as well as capturing information from the physical world using sensing hardware such as cameras, microphones etc. This wearable device is complemented by a smart band and Wi-Fi module, enabling the system to function as a 5G access point. Optionally, additional edge computing nodes can be installed on LEA patrol cars (PCEN), as shown in the visual below, utilising multiple smart devices, similar to the ones “worn” by the officers.

Data collected from these devices will be processed on the DARLENE cloud, where the Machine Learning (ML) method, which can analyse data with minimal human intervention, will be used to classify serious cases of potential harm perpetrated by criminals or terrorists. The DARLENE project, as part of its legal and ethics efforts, will be considering mitigation measures to avoid bias in the use of these machine learning algorithms.