Periodic Reporting for period 1 - REXASI-PRO (REliable & eXplAinable Swarm Intelligence for People with Reduced mObility)
Periodo di rendicontazione: 2022-10-01 al 2024-03-31
AI's rapid adoption spans various domains, fueled by advancements in computational architectures, algorithms, and data availability. Despite remarkable progress, challenges persist, notably in transparency and trustworthiness due to AI's opaque nature. Addressing these concerns, guidelines like those from the European Commission emphasize human oversight, technical robustness, privacy, and fairness. However, current solutions for safety-critical AI applications face foundational hurdles. Initiatives like REXASI-PRO aim to overcome these challenges by integrating ethics, explainability, and decision science into a reliable and secure AI framework. The project focuses on developing Trustworthy Artificial Swarm Intelligence, ensuring safety, security, and ethical compliance. Specifically tailored for autonomous vehicles aiding people with reduced mobility, this framework promises a seamless, door-to-door experience.
The project aims to develop trustworthy AI systems through human-centric solutions, focusing on multi-agent navigation in crowded areas. Utilizing imitation learning and bio-inspired adaptations, the project will enhance predictability and legibility in various scenarios. Evaluation will be conducted using wheelchairs and flying robots. Additionally, emphasis will be placed on ensuring reliability in planning algorithms, obstacle detection, and swarm communication, employing novel predictive monitoring and cybersecurity measures. The project will advance TRL levels of autonomous wheelchairs and propose energy-efficient AI orchestrators and topology-based dataset optimization for ML algorithms. Furthermore, it will establish an AI ethical risk framework tailored to its use-cases, addressing societal concerns and ensuring secure, ethical cooperation among autonomous systems.
DFKI has been at the forefront of research on smart wheelchairs, focusing on safe and reliable operation for compensation, rehabilitation, assessment, and training since 2012. Their Rolland project has produced prototypes of mobility assistants for the handicapped and elderly. Moving beyond the current state of the art, DFKI plans to enhance their technology by addressing weighted error-gradients between Deep Neural Network (DNN) commands and human control behavior, implementing model-based social navigation algorithms, and analyzing interference rates of safety layer components. Hovering Solutions (HSOL) specializes in flying robots for emergency services and underground inspections. In REXASI-PRO, they aim to improve real-time decision-making capabilities, enabling efficient path planning, autonomous exploration, and obstacle avoidance. Enhanced communication security between agents, like flying robots and wheelchairs, is also a priority. King's College London (KCL) focuses on safety assurance for Cyberphysical Systems (CPSs), critical in avionics, autonomous vehicles, and biomedical devices. They plan to extend predictive monitoring to support multiple specifications and stochastic system dynamics. Additionally, they'll develop techniques to maintain guarantees in dynamic environments. SUPSI aims to develop AI algorithms for trustworthy robotics and social navigation. They'll extend pedestrian-inspired navigation rules and refine event detection methodology. Real-world demonstrators will validate their approach, advancing from proof-of-concept to real-world prototypes. CNR specializes in explainable and reliable AI (XAI and RAI). They'll develop methods for explainable trajectory planning and collision avoidance in complex scenarios involving wheelchairs and flying robots. Data augmentation techniques will enhance knowledge discovery, while topology-based analysis will optimize AI algorithms' performance. The University of South-East (USE) focuses on topology-based analysis for greener AI. They'll develop methods to simplify datasets while preserving AI performance, reducing computation requirements. These methods will be deployed on wheelchairs and flying robots, saving costs and energy. V-Research will enhance the security of CPSs through explainable cybersecurity architecture design and verification. They'll extend design verification to consider functional logic and communication protocols, improving security against logical flaws and ensuring compliance with standards. AITEK aims to improve real-time prediction capabilities of AI systems to ensure safety and reliability. Object detection algorithms will enhance safety by detecting and predicting unsafe scenarios, while tools like SHAP will improve interpretability and reliability of AI systems. SPXL focuses on orchestrating fleets of autonomous robots to increase robustness and safety. They'll develop a scheduling algorithm that's inherently risk-aware, exploiting autonomous navigation capabilities to adapt to unpredicted outcomes. This approach aims to enhance trust and reduce energy consumption.