DIRECT: Distributed Intelligence for REsilient Collaborative roboTics in Extreme Environments
Robots hold great potential to support humans in extreme environments—such as post-disaster zones—where conditions are hazardous, unpredictable, and often lack reliable connectivity. Their capacity to operate in dangerous or inaccessible areas makes them crucial for search and rescue, damage assessment, and emergency response. Yet, conventional robotic systems often struggle in such contexts due to limited adaptability, fragile autonomy, and insufficient coordination. Challenges like degraded sensor inputs, dynamic unstructured terrains, GPS-denied environments, and constrained communications further hinder their effectiveness. While embodied AI has significantly enhanced the perception and action capabilities of individual robots, its focus has largely remained on single-agent systems. Achieving efficient multi-robot collaboration under uncertain, real-time, and bandwidth-limited conditions remains a major research frontier. The DIRECT project tackles these challenges by developing an innovative distributed machine learning framework that enables a fleet of robots to collaboratively perceive, reason, plan, and act in extreme environments.
4/6 complete
Imported from OpenAIRE Ukraine | Grant 101299316.
Open sourceOverview
Grant agreement 101299316 | DIRECT
Open sourceRobots hold great potential to support humans in extreme environments—such as post-disaster zones—where conditions are hazardous, unpredictable, and often lack reliable connectivity. Their capacity to operate in dangerous or inaccessible areas makes them crucial for search and rescue, damage assessment, and emergency response. Yet, conventional robotic systems often struggle in such contexts due to limited adaptability, fragile autonomy, and insufficient coordination. Challenges like degraded sensor inputs, dynamic unstructured terrains, GPS-denied environments, and constrained communications further hinder their effectiveness. While embodied AI has significantly enhanced the perception and action capabilities of individual robots, its focus has largely remained on single-agent systems. Achieving efficient multi-robot collaboration under uncertain, real-time, and bandwidth-limited conditions remains a major research frontier. The DIRECT project tackles these challenges by developing an innovative distributed machine learning framework that enables a fleet of robots to collaboratively perceive, reason, plan, and act in extreme environments.