There are several applications that we can consider for this embedded AI lab. Within these applications, we are planning several projects that can take the form of internships, bachelor, master assignments, or independent projects. If you are a student and interested in participating in such a project, please don’t hesitate to contact us.
1. Privacy-driven smart cameras
In today's world, people are increasingly concerned about privacy in data processing. By processing data locally, users can have control over their data and reduce the risk of cyber-attacks on external servers. When data is processed close to the camera, it can be computed faster, but there are limitations in computational power. This presents an interesting challenge to maintain acceptable performance while staying within those limitations.
Projects within this application could include (but are not limited to):
- Wearable camera for scene description to aid visually impaired people
- Wearable camera for real-time object detection and collision avoidance to aid visually impaired people
- Wearable camera for on-the-fly face registration to detect new faces without saving identities
- Security camera powered by solar energy for real-time thread classification
2. Wearable Biometrics
The security of biometric data has become a significant concern, and wearable authentication devices would allow users to be in control of their biometric data, leading to better protection against cyber-attacks. However, since these devices are embedded systems, their computational power is limited, and energy consumption must be carefully managed to ensure that the device can operate for a longer period.
To address these challenges, AI-based algorithms can be used to achieve state-of-the-art performance, but have to be adapted to work with minimal available computational resources.
3. Preventive maintenance / Smart sensors
Embedded AI can enable smart sensors to perform real-time data processing, make decisions based on data patterns, communicate and share data with other devices, and operate with low power consumption. This makes them a valuable tool in various industries and applications, including environmental monitoring, agriculture, and smart homes. Local processing can help reduce the amount of data that needs to be transmitted over the radio, which can in turn help reduce the bandwidth requirements and power consumption. This is because transmitting data over a radio link consumes a significant amount of energy, particularly for wireless devices with limited battery capacity.
Such solutions are well-suited to the application of preventive maintenance. By using AI to analyze sensor data from machinery, companies can reduce downtime, lower maintenance costs, and increase operational efficiency. With the ability to perform predictive maintenance, companies can identify potential issues before they become major problems, ultimately resulting in higher productivity and greater profitability.
Embedded AI also enables predictive maintenance to be performed in environments where internet connectivity may be limited or unreliable, such as in remote or rural areas. Another advantage of embedded AI is increased security. By processing data locally on the device, sensitive information can be kept private and secure, reducing the risk of data breaches or cyberattacks.
4. Unmanned Autonomous vehicles
Embedded AI can be a solution for UAVs (e.g. drones or carts) due to its ability to process large amounts of data in real-time, reducing latency and improving response times. Additionally, AI algorithms can be optimized for low-power consumption, allowing UAVs to perform complex tasks while conserving battery life. Lower latency means that the UAV can respond more quickly to changes in its environment. This is critical for UAVs that are used in real-time applications.