The ReScaLe Project aims to address the following four research questions:
RQ1: How can we achieve active and continual learning for assistive robots?
The key question here is how can we build technology that enables a robot to actively participate in learning from demonstrations to maximize the learning result and how to learn over time without the need for complete retraining.
RQ2: Can we meta-learn efficient robot policies that generalize to new environments without retraining?
We will develop novel methods for transfer learning robot policies on new environments without additional retraining, by meta-learning such policies on a previous set of environments. In particular, we tackle the negative transfer phenomenon across unrelated environments, besides learning to generate new environments that boost the generalization performance of robot policies.
RQ3: How can we ensure the responsible and risk-sensitive development of AI-based robotic systems that are safe and aligned with human rights?
Here we address the challenge of how to develop effective and scalable frameworks for ensuring the responsible development of AI and AI-based robotic systems that are based on human rights. It includes the protection of personal data based on risk-sensitive AI, relying on transparent and efficient risk management techniques together with a risk-adaptive governance scheme.
RQ4: How can we ensure a successful mutual transfer of knowledge — from science to non-scientific actors and vice versa?
We address the question in how far a transdisciplinary and participatory design of transfer formats can facilitate a mutual transfer of knowledge and a collaborative knowledge production. Against this background, we develop a technology acceptance model for assistive robots that involves reverse communication from non-scientific actors to scientists.